1 2020-02-27 Andrew Stubbs <ams@codesourcery.com>
3 * config/gcn/gcn.md (mov<mode>): Add transformations for BI subregs.
5 2020-02-27 Mark Williams <mwilliams@fb.com>
7 * dwarf2out.c (file_name_acquire): Call remap_debug_filename.
8 * lto-opts.c (lto_write_options): Drop -fdebug-prefix-map,
9 -ffile-prefix-map and -fmacro-prefix-map.
10 * lto-streamer-out.c: Include file-prefix-map.h.
11 (lto_output_location): Remap the file part of locations.
13 2020-02-27 Jakub Jelinek <jakub@redhat.com>
16 * gimplify.c (gimplify_init_constructor): Don't promote readonly
17 DECL_REGISTER variables to TREE_STATIC.
19 PR tree-optimization/93582
20 PR tree-optimization/93945
21 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle memset with
22 non-zero INTEGER_CST second argument and ref->offset or ref->size
23 not a multiple of BITS_PER_UNIT.
25 2020-02-27 Jonathan Wakely <jwakely@redhat.com>
27 * doc/install.texi (Binaries): Update description of BullFreeware.
29 2020-02-26 Sandra Loosemore <sandra@codesourcery.com>
33 * doc/invoke.texi (Option Summary): Re-alphabetize warnings in
34 C++ Language Options, Warning Options, and Static Analyzer
35 Options lists. Document negative form of options enabled by
36 default. Move some things around to more accurately sort
38 (C++ Dialect Options, Warning Options, Static Analyzer
39 Options): Document negative form of options when enabled by
40 default. Move some things around to more accurately sort
41 warnings by category. Add some missing index entries.
44 2020-02-26 Carl Love <cel@us.ibm.com>
47 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
48 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
49 for the vector unsigned short arguments. It is also listed as the
50 name of the built-in for arguments vector unsigned short,
51 vector unsigned int and vector unsigned long long built-ins. The
52 name of the builtins for these arguments should be:
53 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
54 __builtin_crypto_vpmsumd respectively.
56 2020-02-26 Richard Biener <rguenther@suse.de>
58 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
61 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
64 * optabs-tree.c (supportable_convert_operation): Reject types with
67 2020-02-26 David Malcolm <dmalcolm@redhat.com>
69 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
71 2020-02-26 Jakub Jelinek <jakub@redhat.com>
73 PR tree-optimization/93820
74 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
75 argument to ALL_INTEGER_CST_P boolean.
76 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
77 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
78 adjacent INTEGER_CST store into merged_store->only_constants like
81 2020-02-25 Jakub Jelinek <jakub@redhat.com>
84 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
86 * cfghooks.c (verify_flow_info): Likewise.
87 * predict.c (combine_predictions_for_bb): Likewise.
88 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
89 sucessor -> successor.
90 (find_traces_1_round): Fix comment typo, destinarion -> destination.
91 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
93 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
94 message typo, sucessors -> successors.
96 2020-02-25 Martin Sebor <msebor@redhat.com>
98 * doc/extend.texi (attribute access): Correct an example.
100 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
102 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
104 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
106 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
107 (VD): Enable for V4BF.
109 (VQ): Enable for V8BF.
112 (VDBL, Vdbl): Add V4BF.
113 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
114 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
115 (bfloat16x8x2_t): Likewise.
116 (bfloat16x4x3_t): Likewise.
117 (bfloat16x8x3_t): Likewise.
118 (bfloat16x4x4_t): Likewise.
119 (bfloat16x8x4_t): Likewise.
120 (vcombine_bf16): New.
121 (vld1_bf16, vld1_bf16_x2): New.
122 (vld1_bf16_x3, vld1_bf16_x4): New.
123 (vld1q_bf16, vld1q_bf16_x2): New.
124 (vld1q_bf16_x3, vld1q_bf16_x4): New.
125 (vld1_lane_bf16): New.
126 (vld1q_lane_bf16): New.
127 (vld1_dup_bf16): New.
128 (vld1q_dup_bf16): New.
131 (vld2_dup_bf16): New.
132 (vld2q_dup_bf16): New.
135 (vld3_dup_bf16): New.
136 (vld3q_dup_bf16): New.
139 (vld4_dup_bf16): New.
140 (vld4q_dup_bf16): New.
141 (vst1_bf16, vst1_bf16_x2): New.
142 (vst1_bf16_x3, vst1_bf16_x4): New.
143 (vst1q_bf16, vst1q_bf16_x2): New.
144 (vst1q_bf16_x3, vst1q_bf16_x4): New.
145 (vst1_lane_bf16): New.
146 (vst1q_lane_bf16): New.
154 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
156 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
157 (VALL_F16): Likewise.
158 (VALLDI_F16): Likewise.
161 (vswap_width_name): Likewise.
162 (VSWAP_WIDTH): Likewise.
166 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
167 (vget_lane_bf16, vgetq_lane_bf16): New.
169 (vdup_n_bf16, vdupq_n_bf16): New.
170 (vdup_lane_bf16, vdup_laneq_bf16): New.
171 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
172 (vduph_lane_bf16, vduph_laneq_bf16): New.
173 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
174 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
175 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
176 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
177 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
178 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
179 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
180 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
181 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
182 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
183 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
184 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
185 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
186 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
187 (vreinterpretq_bf16_p128): New.
188 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
189 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
190 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
191 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
192 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
193 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
194 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
195 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
196 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
197 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
198 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
199 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
200 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
201 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
202 (vreinterpretq_p128_bf16): New.
204 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
206 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
207 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
208 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
209 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
210 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
211 * config/arm/iterators.md (VSF2BF): New attribute.
212 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
213 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
214 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
216 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
218 * config/arm/arm.md (required_for_purecode): New attribute.
219 (enabled): Handle required_for_purecode.
220 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
221 work with -mpure-code.
223 2020-02-25 Jakub Jelinek <jakub@redhat.com>
225 PR rtl-optimization/93908
226 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
229 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
231 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
233 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
235 * doc/install.texi (--enable-checking): Adjust wording.
237 2020-02-25 Richard Biener <rguenther@suse.de>
239 PR tree-optimization/93868
240 * tree-vect-slp.c (slp_copy_subtree): New function.
241 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
242 re-arranging stmts in it.
244 2020-02-25 Jakub Jelinek <jakub@redhat.com>
247 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
248 dummy function and remove it at the end.
251 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
252 paramter -> parameter.
253 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
254 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
256 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
258 * doc/install.texi (--enable-checking): Properly document current
260 (--enable-stage1-checking): Minor clarification about bootstrap.
262 2020-02-24 David Malcolm <dmalcolm@redhat.com>
265 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
266 -fanalyzer-checker=taint is also required.
267 (-fanalyzer-checker=): Note that providing this option enables the
268 given checker, and doing so may be required for checkers that are
271 2020-02-24 David Malcolm <dmalcolm@redhat.com>
273 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
274 significant control flow events; add a "3" which shows all
275 control flow events; the old "3" becomes "4".
277 2020-02-24 Jakub Jelinek <jakub@redhat.com>
279 PR tree-optimization/93582
280 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
281 pd.offset and pd.size to be counted in bits rather than bytes, add
282 support for maxsizei that is not a multiple of BITS_PER_UNIT and
283 handle bitfield stores and loads.
284 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
285 uncomparable quantities - bytes vs. bits. Allow push_partial_def
286 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
287 pd.offset/pd.size to be counted in bits rather than bytes.
288 Formatting fix. Rename shadowed len variable to buflen.
290 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
291 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
294 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
295 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
296 * opts-common.c (parse_options_from_collect_gcc_options): New function.
297 (prepend_xassembler_to_collect_as_options): Likewise.
298 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
299 (prepend_xassembler_to_collect_as_options): Likewise.
300 * lto-opts.c (lto_write_options): Stream assembler options
301 in COLLECT_AS_OPTIONS.
302 * lto-wrapper.c (xassembler_options_error): New static variable.
303 (get_options_from_collect_gcc_options): Move parsing options code to
304 parse_options_from_collect_gcc_options and call it.
305 (merge_and_complain): Validate -Xassembler options.
306 (append_compiler_options): Handle OPT_Xassembler.
307 (run_gcc): Append command line -Xassembler options to
309 * doc/invoke.texi: Add documentation about using Xassembler
312 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
314 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
316 (riscv_rtx_costs): Update cost model for LTGT.
318 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
320 PR rtl-optimization/93564
321 * ira-color.c (struct update_cost_queue_elem): New member start.
322 (queue_update_cost, get_next_update_cost): Add new arg start.
323 (allocnos_conflict_p): New function.
324 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
325 Add checking conflicts with allocnos_conflict_p.
326 (update_costs_from_prefs, restore_costs_from_copies): Adjust
327 update_costs_from_allocno calls.
328 (update_conflict_hard_regno_costs): Add checking conflicts with
329 allocnos_conflict_p. Adjust calls of queue_update_cost and
330 get_next_update_cost.
331 (assign_hard_reg): Adjust calls of queue_update_cost. Add
333 (bucket_allocno_compare_func): Restore previous version.
335 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
337 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
338 double-word size when handling aggregate return values.
339 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
340 that homogeneous SFmode and DFmode aggregates are passed and returned
341 in general registers.
343 2020-02-21 Jakub Jelinek <jakub@redhat.com>
346 * opts.c (print_filtered_help): Translate help before appending
347 messages to it rather than after that.
349 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
351 PR rtl-optimization/PR92989
352 * lra-lives.c (process_bb_lives): Restore the original order
353 of the bb liveness update. Call make_hard_regno_dead for each
354 register clobbered at the start of an EH receiver.
356 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
359 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
360 self-recursively generated.
362 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
365 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
368 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
370 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
371 Document new target supports option.
373 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
375 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
376 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
377 * config/arm/iterators.md (MATMUL): New iterator.
378 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
379 (mmla_sfx): New attribute.
380 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
381 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
382 (UNSPEC_MATMUL_US): New.
384 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
386 * config/arm/arm.md: Prevent scalar shifts from being used when big
389 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
390 Richard Biener <rguenther@suse.de>
392 PR tree-optimization/93586
393 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
394 after mismatched array refs; do not sure type size information to
395 recover from unmatched referneces with !flag_strict_aliasing_p.
397 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
399 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
400 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
401 (scatter_store<mode>): Rename to ...
402 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
403 (scatter<mode>_exec): Delete. Move contents ...
404 (mask_scatter_store<mode>): ... here, and rename that to ...
405 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
406 Remove mode conversion.
407 (mask_gather_load<mode>): Rename to ...
408 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
409 Remove mode conversion.
410 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
412 2020-02-21 Martin Jambor <mjambor@suse.cz>
414 PR tree-optimization/93845
415 * tree-sra.c (verify_sra_access_forest): Only test access size of
418 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
420 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
421 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
422 (addv64di3_exec): Likewise.
423 (subv64di3): Likewise.
424 (subv64di3_exec): Likewise.
425 (addv64di3_zext): Likewise.
426 (addv64di3_zext_exec): Likewise.
427 (addv64di3_zext_dup): Likewise.
428 (addv64di3_zext_dup_exec): Likewise.
429 (addv64di3_zext_dup2): Likewise.
430 (addv64di3_zext_dup2_exec): Likewise.
431 (addv64di3_sext_dup2): Likewise.
432 (addv64di3_sext_dup2_exec): Likewise.
433 (<expander>v64di3): Likewise.
434 (<expander>v64di3_exec): Likewise.
435 (*<reduc_op>_dpp_shr_v64di): Likewise.
436 (*plus_carry_dpp_shr_v64di): Likewise.
437 * config/gcn/gcn.md (adddi3): Likewise.
438 (addptrdi3): Likewise.
439 (<expander>di3): Likewise.
441 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
443 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
445 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
447 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
448 support. Use aarch64_emit_mult instead of emitting multiplication
449 instructions directly.
450 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
451 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
453 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
455 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
456 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
457 instead of emitting multiplication instructions directly.
458 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
459 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
460 (@aarch64_frecps<mode>): New expanders.
462 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
464 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
465 on and produce uint64_ts rather than ints.
466 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
467 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
469 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
471 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
472 an unused xmsk register when handling approximate rsqrt.
474 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
476 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
477 flag_finite_math_only condition.
479 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
482 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
483 to destination operand for shufps alternative.
484 (*vec_extractv2si_1): Ditto.
486 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
489 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
492 2020-02-20 Martin Liska <mliska@suse.cz>
495 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
497 2020-02-20 Martin Liska <mliska@suse.cz>
500 * common/config/avr/avr-common.c: Remote trailing "|".
502 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
504 * collect2.c (maybe_run_lto_and_relink): Fix typo in
507 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
509 PR tree-optimization/93767
510 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
511 access-size bias from the offset calculations for negative strides.
513 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
515 * collect2.c (c_file, o_file): Make const again.
516 (ldout,lderrout, dump_ld_file): Remove.
517 (tool_cleanup): Avoid calling not signal-safe functions.
518 (maybe_run_lto_and_relink): Avoid possible signal handler
519 access to unintialzed memory (lto_o_files).
520 (main): Avoid leaking temp files in $TMPDIR.
521 Initialize c_file/o_file with concat, which avoids exposing
522 uninitialized memory to signal handler, which calls unlink(!).
523 Avoid calling maybe_unlink when the main function returns,
524 since the atexit handler is already doing this.
525 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
527 2020-02-19 Martin Jambor <mjambor@suse.cz>
529 PR tree-optimization/93776
530 * tree-sra.c (create_access): Do not create zero size accesses.
531 (get_access_for_expr): Do not search for zero sized accesses.
533 2020-02-19 Martin Jambor <mjambor@suse.cz>
535 PR tree-optimization/93667
536 * tree-sra.c (scalarizable_type_p): Return false if record fields
537 do not follow wach other.
539 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
541 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
542 rather than fmv.x.s/fmv.s.x.
544 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
546 * config/aarch64/aarch64-simd-builtins.def
547 (intrinsic_vec_smult_lo_): New.
548 (intrinsic_vec_umult_lo_): Likewise.
549 (vec_widen_smult_hi_): Likewise.
550 (vec_widen_umult_hi_): Likewise.
551 * config/aarch64/aarch64-simd.md
552 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
553 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
554 (vmull_high_s16): Likewise.
555 (vmull_high_s32): Likewise.
556 (vmull_high_u8): Likewise.
557 (vmull_high_u16): Likewise.
558 (vmull_high_u32): Likewise.
559 (vmull_s8): Likewise.
560 (vmull_s16): Likewise.
561 (vmull_s32): Likewise.
562 (vmull_u8): Likewise.
563 (vmull_u16): Likewise.
564 (vmull_u32): Likewise.
566 2020-02-18 Martin Liska <mliska@suse.cz>
568 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
569 bootstrap by missing removal of invalid sanity check.
571 2020-02-18 Martin Liska <mliska@suse.cz>
574 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
575 Always compare LHS of gimple_assign.
577 2020-02-18 Martin Liska <mliska@suse.cz>
580 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
581 and return type of functions.
582 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
583 Drop MALLOC attribute for void functions.
584 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
585 malloc_state for a new VOID clone.
587 2020-02-18 Martin Liska <mliska@suse.cz>
590 * common.opt: Add -fprofile-reproducibility.
591 * doc/invoke.texi: Document it.
592 * value-prof.c (dump_histogram_value):
593 Document and support behavior for counters[0]
594 being a negative value.
595 (get_nth_most_common_value): Handle negative
596 counters[0] in respect to flag_profile_reproducible.
598 2020-02-18 Jakub Jelinek <jakub@redhat.com>
601 * cgraph.c (verify_speculative_call): Use speculative_id instead of
602 speculative_uid in messages. Remove trailing whitespace from error
603 message. Use num_speculative_call_targets instead of
604 num_speculative_targets in a message.
605 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
606 edge messages and stmt instead of cal_stmt in reference message.
608 PR tree-optimization/93780
609 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
610 before calling build_vector_type.
611 (execute_update_addresses_taken): Likewise.
614 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
615 typo, functoin -> function.
616 * tree.c (free_lang_data_in_decl): Fix comment typo,
617 functoin -> function.
618 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
620 2020-02-17 David Malcolm <dmalcolm@redhat.com>
622 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
624 (print_option_information): Don't call get_option_url if URLs
627 2020-02-17 Alexandre Oliva <oliva@adacore.com>
629 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
630 handling of register_common-less targets.
632 2020-02-17 Martin Liska <mliska@suse.cz>
635 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
637 2020-02-17 Martin Liska <mliska@suse.cz>
640 * config/rs6000/rs6000.c (rs6000_option_override_internal):
643 2020-02-17 Martin Liska <mliska@suse.cz>
646 * config/rx/elf.opt: Fix typo.
648 2020-02-17 Richard Biener <rguenther@suse.de>
651 * opts-global.c (print_ignored_options): Use inform and
654 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
657 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
659 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
662 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
663 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
665 2020-02-15 Jason Merrill <jason@redhat.com>
667 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
669 2020-02-15 Jakub Jelinek <jakub@redhat.com>
671 PR tree-optimization/93744
672 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
673 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
674 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
675 sure @2 in the first and @1 in the other patterns has no side-effects.
677 2020-02-15 David Malcolm <dmalcolm@redhat.com>
678 Bernd Edlinger <bernd.edlinger@hotmail.de>
682 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
683 * configure.ac (--with-diagnostics-urls): New configuration
684 option, based on --with-diagnostics-color.
685 (DIAGNOSTICS_URLS_DEFAULT): New define.
686 * config.h: Regenerate.
687 * configure: Regenerate.
688 * diagnostic.c (diagnostic_urls_init): Handle -1 for
689 DIAGNOSTICS_URLS_DEFAULT from configure-time
690 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
691 and TERM_URLS environment variable.
692 * diagnostic-url.h (diagnostic_url_format): New enum type.
693 (diagnostic_urls_enabled_p): rename to...
694 (determine_url_format): ... this, and change return type.
695 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
696 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
697 the linux console, and mingw.
698 (diagnostic_urls_enabled_p): rename to...
699 (determine_url_format): ... this, and adjust.
700 * pretty-print.h (pretty_printer::show_urls): rename to...
701 (pretty_printer::url_format): ... this, and change to enum.
702 * pretty-print.c (pretty_printer::pretty_printer,
703 pp_begin_url, pp_end_url, test_urls): Adjust.
704 * doc/install.texi (--with-diagnostics-urls): Document the new
705 configuration option.
706 (--with-diagnostics-color): Document the existing interaction
707 with GCC_COLORS better.
708 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
709 vindex reference. Update description of defaults based on the above.
710 (-fdiagnostics-color): Update description of how -fdiagnostics-color
711 interacts with GCC_COLORS.
713 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
716 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
717 conjunction with TARGET_GNU_TLS in early return.
719 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
721 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
722 the mode is not wider than UNITS_PER_WORD.
724 2020-02-14 Martin Jambor <mjambor@suse.cz>
726 PR tree-optimization/93516
727 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
728 access of the same type as the parent.
729 (propagate_subaccesses_from_lhs): Likewise.
731 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
734 * config/i386/avx512vbmi2intrin.h
735 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
736 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
737 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
738 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
739 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
740 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
741 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
742 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
743 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
744 of lacking a closing parenthesis.
745 * config/i386/avx512vbmi2vlintrin.h
746 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
747 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
748 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
749 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
750 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
751 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
752 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
753 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
754 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
755 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
756 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
757 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
758 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
759 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
760 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
761 _mm_shldi_epi32, _mm_mask_shldi_epi32,
762 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
763 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
765 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
768 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
769 the target function entry.
771 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
773 * common/config/arc/arc-common.c (arc_option_optimization_table):
774 Disable if-conversion step when optimized for size.
776 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
778 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
779 R12-R15 are always in ARCOMPACT16_REGS register class.
780 * config/arc/arc.opt (mq-class): Deprecate.
781 * config/arc/constraint.md ("q"): Remove dependency on mq-class
783 * doc/invoke.texi (mq-class): Update text.
784 * common/config/arc/arc-common.c (arc_option_optimization_table):
787 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
789 * config/arc/arc.c (arc_insn_cost): New function.
790 (TARGET_INSN_COST): Define.
791 * config/arc/arc.md (cost): New attribute.
792 (add_n): Use arc_nonmemory_operand.
793 (ashlsi3_insn): Likewise, also update constraints.
794 (ashrsi3_insn): Likewise.
796 (add_shift): Likewise.
797 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
799 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
801 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
803 (umulsidi_600): Likewise.
805 2020-02-13 Jakub Jelinek <jakub@redhat.com>
808 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
809 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
810 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
811 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
812 pass __A to the builtin followed by __W instead of __A followed by
814 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
815 _mm512_mask_popcnt_epi64): Likewise.
816 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
817 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
818 _mm256_mask_popcnt_epi64): Likewise.
820 PR tree-optimization/93582
821 * fold-const.h (shift_bytes_in_array_left,
822 shift_bytes_in_array_right): Declare.
823 * fold-const.c (shift_bytes_in_array_left,
824 shift_bytes_in_array_right): New function, moved from
825 gimple-ssa-store-merging.c, no longer static.
826 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
827 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
828 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
829 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
830 shift_bytes_in_array.
831 (verify_shift_bytes_in_array): Rename to ...
832 (verify_shift_bytes_in_array_left): ... this. Use
833 shift_bytes_in_array_left instead of shift_bytes_in_array.
834 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
835 instead of verify_shift_bytes_in_array.
836 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
837 / native_interpret_expr where the store covers all needed bits,
838 punt on PDP-endian, otherwise allow all involved offsets and sizes
839 not to be byte-aligned.
842 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
843 use const_0_to_255_operand predicate instead of immediate_operand.
844 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
845 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
846 vgf2p8affineinvqb_<mode><mask_name>,
847 vgf2p8affineqb_<mode><mask_name>): Drop mode from
848 const_0_to_255_operand predicated operands.
850 2020-02-12 Jeff Law <law@redhat.com>
852 * config/h8300/h8300.md (comparison shortening peepholes): Use
853 a mode iterator to merge the HImode and SImode peepholes.
855 2020-02-12 Jakub Jelinek <jakub@redhat.com>
858 * real.c (is_even): Make static. Function comment fix.
859 (is_halfway_below): Make static, don't assert R is not inf/nan,
860 instead return false for those. Small formatting fixes.
862 2020-02-12 Martin Sebor <msebor@redhat.com>
865 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
866 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
867 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
868 (strlen_check_and_optimize_call): Adjust callee name.
870 2020-02-12 Jeff Law <law@redhat.com>
872 * config/h8300/h8300.md (comparison shortening peepholes): Drop
873 (and (xor)) variant. Combine other two into single peephole.
875 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
877 PR rtl-optimization/93565
878 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
880 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
882 * config/aarch64/aarch64-simd.md
883 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
884 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
885 generating separate ADDV and zero_extend patterns.
886 * config/aarch64/iterators.md (VDQV_E): New iterator.
888 2020-02-12 Jeff Law <law@redhat.com>
890 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
891 expanders, splits, etc.
892 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
893 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
894 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
895 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
896 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
897 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
899 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
901 2020-02-12 Jakub Jelinek <jakub@redhat.com>
904 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
905 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
906 TARGET_AVX512DQ from condition.
907 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
908 instead of <mask_mode512bit_condition> in condition. If
909 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
911 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
914 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
917 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
919 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
921 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
922 where strlen is more legible.
923 (rs6000_builtin_vectorized_libmass): Ditto.
924 (rs6000_print_options_internal): Ditto.
926 2020-02-11 Martin Sebor <msebor@redhat.com>
928 PR tree-optimization/93683
929 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
931 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
933 * config/rs6000/predicates.md (cint34_operand): Rename the
934 -mprefixed-addr option to be -mprefixed.
935 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
936 the -mprefixed-addr option to be -mprefixed.
937 (OTHER_FUTURE_MASKS): Likewise.
938 (POWERPC_MASKS): Likewise.
939 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
940 the -mprefixed-addr option to be -mprefixed. Change error
941 messages to refer to -mprefixed.
942 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
944 (rs6000_legitimate_offset_address_p): Likewise.
945 (rs6000_mode_dependent_address): Likewise.
946 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
947 "-mprefixed" for target attributes and pragmas.
948 (address_to_insn_form): Rename the -mprefixed-addr option to be
950 (rs6000_adjust_insn_length): Likewise.
951 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
952 -mprefixed-addr option to be -mprefixed.
953 (ASM_OUTPUT_OPCODE): Likewise.
954 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
955 -mprefixed-addr option to be -mprefixed.
956 * config/rs6000/rs6000.opt (-mprefixed): Rename the
957 -mprefixed-addr option to be prefixed. Change the option from
958 being undocumented to being documented.
959 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
960 -mprefixed option. Update the -mpcrel documentation to mention
963 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
965 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
966 including FIRST_PSEUDO_REGISTER - 1.
967 * ira-color.c (print_hard_reg_set): Ditto.
969 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
971 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
972 (USTERNOP_QUALIFIERS): New define.
973 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
974 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
975 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
976 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
977 * config/arm/arm_neon.h (vusdot_s32): New.
978 (vusdot_lane_s32): New.
979 (vusdotq_lane_s32): New.
980 (vsudot_lane_s32): New.
981 (vsudotq_lane_s32): New.
982 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
983 * config/arm/iterators.md (DOTPROD_I8MM): New.
984 (sup, opsuffix): Add <us/su>.
985 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
986 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
988 2020-02-11 Richard Biener <rguenther@suse.de>
990 PR tree-optimization/93661
991 PR tree-optimization/93662
992 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
994 * tree-sra.c (get_access_for_expr): Likewise.
996 2020-02-10 Jakub Jelinek <jakub@redhat.com>
999 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
1000 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
1001 Change condition from TARGET_AVX2 to TARGET_AVX.
1003 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
1006 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
1007 argument of strncmp.
1009 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1011 Try to generate zero-based comparisons.
1012 * config/cris/cris.c (cris_reduce_compare): New function.
1013 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
1014 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
1015 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
1017 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
1020 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
1021 in Thumb state and also as a destination in Arm state. Add T16
1024 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
1026 * md.texi (Define Subst): Match closing paren in example.
1028 2020-02-10 Jakub Jelinek <jakub@redhat.com>
1032 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
1033 arguments of strncmp.
1035 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
1038 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
1039 but different source value.
1040 (adjust_callers_for_value_intersection): New function.
1041 (gather_edges_for_value): Adjust order of callers to let a
1042 non-self-recursive caller be the first element.
1043 (self_recursive_pass_through_p): Add a new parameter "simple", and
1044 check generalized self-recursive pass-through jump function.
1045 (self_recursive_agg_pass_through_p): Likewise.
1046 (find_more_scalar_values_for_callers_subset): Compute value from
1047 pass-through jump function for self-recursive.
1048 (intersect_with_plats): Cleanup previous implementation code for value
1049 itersection with self-recursive call edge.
1050 (intersect_with_agg_replacements): Likewise.
1051 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1052 function for self-recursive call edge. Cleanup previous implementation
1053 code for value intersection with self-recursive call edge.
1054 (decide_whether_version_node): Remove dead callers and adjust order
1055 to let a non-self-recursive caller be the first element.
1057 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1059 * recog.c: Move pass_split_before_sched2 code in front of
1060 pass_split_before_regstack.
1061 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1062 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1063 (rest_of_handle_split_before_sched2): Remove.
1064 (pass_split_before_sched2::execute): Unconditionally call
1066 (enable_split_before_sched2): New function.
1067 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1068 (pass_split_before_regstack::gate): Ditto.
1069 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1070 Update name check for renamed split4 pass.
1071 * config/sh/sh.c (register_sh_passes): Update pass insertion
1072 point for renamed split4 pass.
1074 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1076 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1077 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1078 copying them around between host and target.
1080 2020-02-08 Andrew Pinski <apinski@marvell.com>
1083 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1084 STRICT_ALIGNMENT also.
1086 2020-02-08 Jim Wilson <jimw@sifive.com>
1089 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1091 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1092 Jakub Jelinek <jakub@redhat.com>
1095 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1096 xmm16-xmm31 call-used even in 64-bit ms-abi.
1098 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1100 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1101 (simd_ummla, simd_usmmla): Likewise.
1102 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1103 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1104 (vusmmlaq_s32): New.
1106 2020-02-07 Richard Biener <rguenther@suse.de>
1109 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1110 skipping unreachable regions.
1111 (optimize_inline_calls): Skip folding stmts when we didn't
1114 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1117 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1118 Don't return aggregates with only SFmode and DFmode in SSE
1120 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1122 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1125 * config/rs6000/rs6000-logue.c
1126 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1127 if it fails, move rs into end_addr and retry. Add
1128 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1129 the insn pattern doesn't describe well what exactly happens to
1133 * config/i386/predicates.md (avx_identity_operand): Remove.
1134 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1135 (avx_<castmode><avxsizesuffix>_<castmode>,
1136 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1137 a VEC_CONCAT of the operand and UNSPEC_CAST.
1138 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1139 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1143 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1144 recog_data.insn if distance_non_agu_define changed it.
1146 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1149 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1150 we only had X-FORM (reg+reg) addressing for vectors. Also before
1151 ISA 3.0, we only had X-FORM addressing for scalars in the
1152 traditional Altivec registers.
1154 2020-02-06 <zhongyunde@huawei.com>
1155 Vladimir Makarov <vmakarov@redhat.com>
1157 PR rtl-optimization/93561
1158 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1159 hard register range.
1161 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1163 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1166 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1168 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1169 where the low and the high 32 bits are equal to each other specially,
1170 with an rldimi instruction.
1172 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1174 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1176 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1178 * config/arm/arm-tables.opt: Regenerate.
1180 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1183 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1184 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1185 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1187 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1189 PR rtl-optimization/87763
1190 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1192 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1194 * config/aarch64/aarch64-simd-builtins.def
1195 (bfmlaq): New built-in function.
1196 (bfmlalb): New built-in function.
1197 (bfmlalt): New built-in function.
1198 (bfmlalb_lane): New built-in function.
1199 (bfmlalt_lane): New built-in function.
1200 * config/aarch64/aarch64-simd.md
1201 (aarch64_bfmmlaqv4sf): New pattern.
1202 (aarch64_bfmlal<bt>v4sf): New pattern.
1203 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1204 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1205 (vbfmlalbq_f32): New intrinsic.
1206 (vbfmlaltq_f32): New intrinsic.
1207 (vbfmlalbq_lane_f32): New intrinsic.
1208 (vbfmlaltq_lane_f32): New intrinsic.
1209 (vbfmlalbq_laneq_f32): New intrinsic.
1210 (vbfmlaltq_laneq_f32): New intrinsic.
1211 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1212 (bt): New int attribute.
1214 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1216 * config/i386/i386.md (*pushtf): Emit "#" instead of
1217 calling gcc_unreachable in insn output.
1220 (*pushsf_rex64): Ditto for alternatives other than 1.
1221 (*pushsf): Ditto for alternatives other than 1.
1223 2020-02-06 Martin Liska <mliska@suse.cz>
1225 PR gcov-profile/91971
1226 PR gcov-profile/93466
1227 * coverage.c (coverage_init): Revert mangling of
1228 path into filename. It can lead to huge filename length.
1229 Creation of subfolders seem more natural.
1231 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1234 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1235 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1236 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1238 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1241 * config/i386/predicates.md (avx_identity_operand): New predicate.
1242 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1243 define_insn_and_split.
1246 * omp-low.c (use_pointer_for_field): For nested constructs, also
1247 look for map clauses on target construct.
1248 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1249 taskreg_nesting_level.
1252 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1253 shared clause, call omp_notice_variable on outer context if any.
1255 2020-02-05 Jason Merrill <jason@redhat.com>
1258 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1259 non-zero address even if weak and not yet defined.
1261 2020-02-05 Martin Sebor <msebor@redhat.com>
1263 PR tree-optimization/92765
1264 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1265 * tree-ssa-strlen.c (compute_string_length): Remove.
1266 (determine_min_objsize): Remove.
1267 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1268 Avoid using type size as the upper bound on string length.
1269 (handle_builtin_string_cmp): Add an argument. Adjust.
1270 (strlen_check_and_optimize_call): Pass additional argument to
1271 handle_builtin_string_cmp.
1273 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1275 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1276 (*pushdi2_rex64 peephole2): Unconditionally split after
1278 (*ashl<mode>3_doubleword): Ditto.
1279 (*<shift_insn><mode>3_doubleword): Ditto.
1281 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1284 * config/rs6000/rs6000.c (get_vector_offset): Fix
1286 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1288 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1290 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1293 (Special Functions for Debugging the Analyzer): Update description
1294 of __analyzer_dump_exploded_nodes.
1296 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1299 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1300 include sets and not clobbers in the vzeroupper pattern.
1301 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1302 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1303 (*avx_vzeroupper_1): New define_insn_and_split.
1306 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1307 don't run when !optimize.
1308 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1311 2020-02-05 Richard Biener <rguenther@suse.de>
1314 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1315 checks before matching calls.
1317 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1319 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1320 function comment typo.
1323 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1324 simd_clone_create failed when i == 0, adjust clone->nargs by
1327 2020-02-05 Martin Liska <mliska@suse.cz>
1330 * doc/invoke.texi: Document that one should
1331 not combine ASLR and -fpch.
1333 2020-02-04 Richard Biener <rguenther@suse.de>
1335 PR tree-optimization/93538
1336 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1338 2020-02-04 Richard Biener <rguenther@suse.de>
1340 PR tree-optimization/91123
1341 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1342 (vn_walk_cb_data::last_vuse): New member.
1343 (vn_walk_cb_data::saved_operands): Likewsie.
1344 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1345 (vn_walk_cb_data::push_partial_def): Use finish.
1346 (vn_reference_lookup_2): Update last_vuse and use finish if
1347 we've saved operands.
1348 (vn_reference_lookup_3): Use finish and update calls to
1349 push_partial_defs everywhere. When translating through
1350 memcpy or aggregate copies save off operands and alias-set.
1351 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1352 operation for redundant store removal.
1354 2020-02-04 Richard Biener <rguenther@suse.de>
1356 PR tree-optimization/92819
1357 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1358 generating more stmts than before.
1360 2020-02-04 Martin Liska <mliska@suse.cz>
1362 * config/arm/arm.c (arm_gen_far_branch): Move the function
1363 outside of selftests.
1365 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1367 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1368 function to adjust PC-relative vector addresses.
1369 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1370 handle vectors with PC-relative addresses.
1372 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1374 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1376 (hard_reg_and_mode_to_addr_mask): Delete.
1377 (rs6000_adjust_vec_address): If the original vector address
1378 was REG+REG or REG+OFFSET and the element is not zero, do the add
1379 of the elements in the original address before adding the offset
1380 for the vector element. Use address_to_insn_form to validate the
1381 address using the register being loaded, rather than guessing
1382 whether the address is a DS-FORM or DQ-FORM address.
1384 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1386 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1387 to calculate the offset in memory from the start of a vector of a
1388 particular element. Add code to keep the element number in
1389 bounds if the element number is variable.
1390 (rs6000_adjust_vec_address): Move calculation of offset of the
1391 vector element to get_vector_offset.
1392 (rs6000_split_vec_extract_var): Do not do the initial AND of
1393 element here, move the code to get_vector_offset.
1395 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1397 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1400 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1402 * config/rs6000/constraints.md: Improve documentation.
1404 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1407 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1408 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1410 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1412 * config.gcc: Remove "carrizo" support.
1413 * config/gcn/gcn-opts.h (processor_type): Likewise.
1414 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1415 * config/gcn/gcn.opt (gpu_type): Likewise.
1416 * config/gcn/t-omp-device: Likewise.
1418 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1421 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1422 * config/arm/arm.c (arm_gen_far_branch): New function
1424 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1426 2020-02-03 Julian Brown <julian@codesourcery.com>
1427 Tobias Burnus <tobias@codesourcery.com>
1429 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1431 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1434 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1435 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1438 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1440 PR rtl-optimization/91333
1441 * ira-color.c (struct allocno_color_data): Add member
1443 (init_allocno_threads): Set the member up.
1444 (bucket_allocno_compare_func): Add compare hard reg
1447 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1449 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1451 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1452 * config.in: Regenerated.
1453 * configure: Regenerated.
1454 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1455 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1456 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1458 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1460 * configure: Regenerate.
1462 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1464 PR rtl-optimization/91333
1465 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1466 reg preferences comparison up.
1468 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1470 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1471 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1472 aarch64-sve-builtins-base.h.
1473 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1474 aarch64-sve-builtins-base.cc.
1475 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1476 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1478 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1479 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1480 (svcvtnt): New functions.
1481 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1482 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1483 (svcvtnt): New functions.
1484 (svcvt): Add a form that converts f32 to bf16.
1485 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1486 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1488 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1489 Treat B as bfloat16_t.
1490 (ternary_bfloat_lane_base): New class.
1491 (ternary_bfloat_def): Likewise.
1492 (ternary_bfloat): New shape.
1493 (ternary_bfloat_lane_def): New class.
1494 (ternary_bfloat_lane): New shape.
1495 (ternary_bfloat_lanex2_def): New class.
1496 (ternary_bfloat_lanex2): New shape.
1497 (ternary_bfloat_opt_n_def): New class.
1498 (ternary_bfloat_opt_n): New shape.
1499 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1500 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1501 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1502 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1503 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1504 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1505 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1506 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1507 the pattern off the narrow mode instead of the wider one.
1508 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1509 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1510 (sve_fp_op): Handle them.
1511 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1512 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1514 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1516 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1517 * config/aarch64/aarch64-modes.def (BF): Move definition before
1518 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1519 (SVE_MODES): Handle BF modes.
1520 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1522 (aarch64_full_sve_mode): Likewise.
1523 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1525 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1526 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1527 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1528 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1530 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1532 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1533 (TYPES_all_data): Add bf16.
1534 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1535 (register_tuple_type): Increase buffer size.
1536 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1537 (bf16): New type suffix.
1538 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1539 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1540 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1541 Change type from all_data to all_arith.
1542 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1545 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1546 Matthew Malcomson <matthew.malcomson@arm.com>
1547 Richard Sandiford <richard.sandiford@arm.com>
1549 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1550 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1551 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1552 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1553 __ARM_FEATURE_MATMUL_FP64.
1554 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1555 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1556 be disabled at the same time.
1557 (f32mm): New extension.
1558 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1559 (AARCH64_FL_F64MM): Bump to the next bit up.
1560 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1561 (TARGET_SVE_F64MM): New macros.
1562 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1563 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1564 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1565 (UNSPEC_ZIP2Q): New unspeccs.
1566 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1567 (optab, sur, perm_insn): Handle the new unspecs.
1568 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1569 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1570 TARGET_SVE_F64MM instead of separate tests.
1571 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1572 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1573 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1574 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1575 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1576 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1577 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1578 (TYPES_s_signed): New macro.
1579 (TYPES_s_integer): Use it.
1580 (TYPES_d_float): New macro.
1581 (TYPES_d_data): Use it.
1582 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1583 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1584 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1585 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1586 (svmmla): New shape.
1587 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1588 template parameters.
1589 (ternary_resize2_lane_base): Likewise.
1590 (ternary_resize2_base): New class.
1591 (ternary_qq_lane_base): Likewise.
1592 (ternary_intq_uintq_lane_def): Likewise.
1593 (ternary_intq_uintq_lane): New shape.
1594 (ternary_intq_uintq_opt_n_def): New class
1595 (ternary_intq_uintq_opt_n): New shape.
1596 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1597 (ternary_uintq_intq_def): New class.
1598 (ternary_uintq_intq): New shape.
1599 (ternary_uintq_intq_lane_def): New class.
1600 (ternary_uintq_intq_lane): New shape.
1601 (ternary_uintq_intq_opt_n_def): New class.
1602 (ternary_uintq_intq_opt_n): New shape.
1603 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1604 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1605 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1606 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1608 (svdotprod_lane_impl): ...this new class.
1609 (svmmla_impl, svusdot_impl): New classes.
1610 (svdot_lane): Update to use svdotprod_lane_impl.
1611 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1612 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1614 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1615 function, with no types defined.
1616 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1617 AARCH64_FL_I8MM functions.
1618 (svmmla): New AARCH64_FL_F32MM function.
1619 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1620 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1621 AARCH64_FL_F64MM function.
1622 (REQUIRED_EXTENSIONS):
1624 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1626 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1629 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1631 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1632 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1633 (*movti_internal): Do not check for
1634 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1635 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1636 just after check for TARGET_AVX.
1637 (*movdf_internal): Ditto.
1638 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1639 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1640 * config/i386/sse.md (mov<mode>_internal): Only check
1641 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1642 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1643 (<sse>_andnot<mode>3<mask_name>): Move check for
1644 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1645 (<code><mode>3<mask_name>): Ditto.
1646 (*andnot<mode>3): Ditto.
1647 (*andnottf3): Ditto.
1648 (*<code><mode>3): Ditto.
1649 (*<code>tf3): Ditto.
1650 (*andnot<VI:mode>3): Remove
1651 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1652 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1653 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1654 (sse4_1_blendv<ssemodesuffix>): Ditto.
1655 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1656 Explain that tune applies to 128bit instructions only.
1658 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1660 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1661 to definition of hsa_kernel_description. Parse assembly to find SGPR
1662 and VGPR count of kernel and store in hsa_kernel_description.
1664 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1666 PR rtl-optimization/91838
1667 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1668 to truncate if allowed or reject combination.
1670 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1672 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1673 (find_inv_vars_cb): Likewise.
1675 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1677 * calls.c (special_function_p): Split out the check for DECL_NAME
1678 being non-NULL and fndecl being extern at file scope into a
1679 new maybe_special_function_p and call it. Drop check for fndecl
1680 being non-NULL that was after a usage of DECL_NAME (fndecl).
1681 * tree.h (maybe_special_function_p): New inline function.
1683 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1685 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1686 (mask_gather_load<mode>): ... here, and zero-initialize the
1688 (maskload<mode>di): Zero-initialize the destination.
1691 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1694 * doc/analyzer.texi (Limitations): Note that constraints on
1695 floating-point values are currently ignored.
1697 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1700 * symtab.c (symtab_node::noninterposable_alias): If localalias
1701 already exists, but is not usable, append numbers after it until
1702 a unique name is found. Formatting fix.
1705 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1708 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1710 * config/gcn/gcn.c (print_operand): Handle LTGT.
1711 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1713 2020-01-30 Richard Biener <rguenther@suse.de>
1715 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1716 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1718 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1720 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1721 without a DECL in .data.rel.ro.local.
1723 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1726 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1730 * config/i386/sse.md
1731 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1732 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1733 any_extend code iterator instead of always zero_extend.
1734 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1735 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1736 Use any_extend code iterator instead of always zero_extend.
1737 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1738 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1739 Use any_extend code iterator instead of always zero_extend.
1740 (*sse2_pmovmskb_ext): New define_insn.
1741 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1744 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1745 (*popcountsi2_zext_falsedep): New define_insn.
1747 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1749 * config.in: Regenerated.
1750 * configure: Regenerated.
1752 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1755 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1756 LLVM's assembler changed the default in version 9.
1758 2020-01-24 Jeff Law <law@redhat.com>
1760 PR tree-optimization/89689
1761 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1763 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1767 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1769 PR rtl-optimization/87763
1770 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1771 simplification to handle subregs as well as bare regs.
1772 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1774 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1777 * ira.c (ira): Revert use of simplified LRA algorithm.
1779 2020-01-29 Martin Jambor <mjambor@suse.cz>
1781 PR tree-optimization/92706
1782 * tree-sra.c (struct access): Fields first_link, last_link,
1783 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1784 next_rhs_queued and grp_rhs_queued respectively, new fields
1785 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1786 (struct assign_link): Field next renamed to next_rhs, new field
1787 next_lhs. Updated comment.
1788 (work_queue_head): Renamed to rhs_work_queue_head.
1789 (lhs_work_queue_head): New variable.
1790 (add_link_to_lhs): New function.
1791 (relink_to_new_repr): Also relink LHS lists.
1792 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1793 (add_access_to_lhs_work_queue): New function.
1794 (pop_access_from_work_queue): Renamed to
1795 pop_access_from_rhs_work_queue.
1796 (pop_access_from_lhs_work_queue): New function.
1797 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1799 (child_would_conflict_in_lacc): Renamed to
1800 child_would_conflict_in_acc. Adjusted parameter names.
1801 (create_artificial_child_access): New parameter set_grp_read, use it.
1802 (subtree_mark_written_and_enqueue): Renamed to
1803 subtree_mark_written_and_rhs_enqueue.
1804 (propagate_subaccesses_across_link): Renamed to
1805 propagate_subaccesses_from_rhs.
1806 (propagate_subaccesses_from_lhs): New function.
1807 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1810 2020-01-29 Martin Jambor <mjambor@suse.cz>
1812 PR tree-optimization/92706
1813 * tree-sra.c (struct access): Adjust comment of
1814 grp_total_scalarization.
1815 (find_access_in_subtree): Look for single children spanning an entire
1817 (scalarizable_type_p): Allow register accesses, adjust callers.
1818 (completely_scalarize): Remove function.
1819 (scalarize_elem): Likewise.
1820 (create_total_scalarization_access): Likewise.
1821 (sort_and_splice_var_accesses): Do not track total scalarization
1823 (analyze_access_subtree): New parameter totally, adjust to new meaning
1824 of grp_total_scalarization.
1825 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1826 (can_totally_scalarize_forest_p): New function.
1827 (create_total_scalarization_access): Likewise.
1828 (create_total_access_and_reshape): Likewise.
1829 (total_should_skip_creating_access): Likewise.
1830 (totally_scalarize_subtree): Likewise.
1831 (analyze_all_variable_accesses): Perform total scalarization after
1832 subaccess propagation using the new functions above.
1833 (initialize_constant_pool_replacements): Output initializers by
1834 traversing the access tree.
1836 2020-01-29 Martin Jambor <mjambor@suse.cz>
1838 * tree-sra.c (verify_sra_access_forest): New function.
1839 (verify_all_sra_access_forests): Likewise.
1840 (create_artificial_child_access): Set parent.
1841 (analyze_all_variable_accesses): Call the verifier.
1843 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1845 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1846 if called on indirect edge.
1847 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1848 speculative call if needed.
1850 2020-01-29 Richard Biener <rguenther@suse.de>
1852 PR tree-optimization/93428
1853 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1854 permutation when the load node is created.
1855 (vect_analyze_slp_instance): Re-use it here.
1857 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1859 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1861 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1863 PR rtl-optimization/93272
1864 * ira-lives.c (process_out_of_region_eh_regs): New function.
1865 (process_bb_node_lives): Call it.
1867 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1869 * coverage.c (read_counts_file): Make error message lowercase.
1871 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1873 * profile-count.c (profile_quality_display_names): Fix ordering.
1875 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1878 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1879 hash only when edge is first within the sequence.
1880 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1881 (symbol_table::create_edge): Do not set target_prob.
1882 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1884 (cgraph_edge::make_speculative): Drop target_prob parameter.
1885 (cgraph_edge::speculative_call_info): Remove.
1886 (cgraph_edge::first_speculative_call_target): New member function.
1887 (update_call_stmt_hash_for_removing_direct_edge): New function.
1888 (cgraph_edge::resolve_speculation): Rewrite to new API.
1889 (cgraph_edge::speculative_call_for_target): New member function.
1890 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1891 multiple speculation targets.
1892 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1894 (verify_speculative_call): Verify that targets form an interval.
1895 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1896 (cgraph_edge::first_speculative_call_target): New member function.
1897 (cgraph_edge::next_speculative_call_target): New member function.
1898 (cgraph_edge::speculative_call_target_ref): New member function.
1899 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1900 (cgraph_edge): Remove target_prob.
1901 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1902 Fix handling of speculative calls.
1903 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1904 * ipa-fnsummary.c (analyze_function_body): Likewise.
1905 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1906 * ipa-profile.c (dump_histogram): Fix formating.
1907 (ipa_profile_generate_summary): Watch for overflows.
1908 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1909 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1910 (update_indirect_edges_after_inlining): Update to new API.
1911 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1913 * profile-count.h: (profile_probability::adjusted): New.
1914 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1915 updating of profile.
1916 * value-prof.c (gimple_ic_transform): Rename to ...
1917 (dump_ic_profile): ... this one; update dumping.
1918 (stream_in_histogram_value): Fix formating.
1919 (gimple_value_profile_transformations): Update.
1921 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1924 * config/i386/i386.md (*movoi_internal_avx): Remove
1925 TARGET_SSE_TYPELESS_STORES check.
1926 (*movti_internal): Prefer TARGET_AVX over
1927 TARGET_SSE_TYPELESS_STORES.
1928 (*movtf_internal): Likewise.
1929 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1930 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1931 from TARGET_SSE_TYPELESS_STORES.
1933 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1935 * diagnostic-core.h (warning_at): Rename overload to...
1936 (warning_meta): ...this.
1937 (emit_diagnostic_valist): Delete decl of overload taking
1938 diagnostic_metadata.
1939 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1940 (warning_at): Rename overload taking diagnostic_metadata to...
1941 (warning_meta): ...this.
1943 2020-01-28 Richard Biener <rguenther@suse.de>
1945 PR tree-optimization/93439
1946 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1947 * tree-cfg.c (move_sese_region_to_fn): ... here.
1948 (verify_types_in_gimple_reference): Verify used cliques are
1951 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1954 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1955 argument of a pointer to struct gcc_options and pass it to
1956 parse_mtune_ctrl_str.
1957 (ix86_function_specific_restore): Pass opts to
1958 set_ix86_tune_features.
1959 (ix86_option_override_internal): Likewise.
1960 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1961 gcc_options and use it for x_ix86_tune_ctrl_string.
1963 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1965 PR rtl-optimization/87763
1966 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1967 simplification to handle subregs as well as bare regs.
1968 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1970 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1972 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1973 for reduction chains that (now) include a call.
1975 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1977 PR tree-optimization/92822
1978 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1979 out the don't-care elements of a vector whose significant elements
1980 are duplicates, make the don't-care elements duplicates too.
1982 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1984 PR tree-optimization/93434
1985 * tree-predcom.c (split_data_refs_to_components): Record which
1986 components have had aliasing loads removed. Prevent store-store
1987 commoning for all such components.
1989 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1992 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1993 -1 or is_vshift is true, use new_vector with number of elts npatterns
1994 rather than new_unary_operation.
1996 PR tree-optimization/93454
1997 * gimple-fold.c (fold_array_ctor_reference): Perform
1998 elt_size.to_uhwi () just once, instead of calling it in every
1999 iteration. Punt if that value is above size of the temporary
2000 buffer. Decrease third native_encode_expr argument when
2001 bufoff + elt_sz is above size of buf.
2003 2020-01-27 Joseph Myers <joseph@codesourcery.com>
2005 * config/mips/mips.c (mips_declare_object_name)
2006 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
2008 2020-01-27 Martin Liska <mliska@suse.cz>
2010 PR gcov-profile/93403
2011 * tree-profile.c (gimple_init_gcov_profiler): Generate
2012 both __gcov_indirect_call_profiler_v4 and
2013 __gcov_indirect_call_profiler_v4_atomic.
2015 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2018 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
2020 (@aarch64_split_simd_mov<mode>): Use it.
2021 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
2022 Leave the vec_extract patterns to handle 2-element vectors.
2023 (aarch64_simd_mov_from_<mode>high): Likewise.
2024 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
2025 (vec_extractv2dfv1df): Likewise.
2027 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2029 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
2030 jump conditions for *compare_condjump<GPI:mode>.
2032 2020-01-27 David Malcolm <dmalcolm@redhat.com>
2035 * digraph.cc (test_edge::test_edge): Specify template for base
2038 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2040 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
2042 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2044 * config/arc/arc-protos.h (gen_mlo): Remove.
2045 (gen_mhi): Likewise.
2046 * config/arc/arc.c (AUX_MULHI): Define.
2047 (arc_must_save_reister): Special handling for r58/59.
2048 (arc_compute_frame_size): Consider mlo/mhi registers.
2049 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2051 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2052 mlo/mhi name selection.
2053 (arc_restore_callee_saves): Don't early restore blink when ISR.
2054 (arc_expand_prologue): Add mlo/mhi saving.
2055 (arc_expand_epilogue): Add mlo/mhi restoring.
2058 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2059 numbering when MUL64 option is used.
2060 (DWARF2_FRAME_REG_OUT): Define.
2061 * config/arc/arc.md (arc600_stall): New pattern.
2062 (VUNSPEC_ARC_ARC600_STALL): Define.
2063 (mulsi64): Use correct mlo/mhi registers.
2064 (mulsi_600): Clean it up.
2065 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2067 (mhi_operand): Likewise.
2069 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2070 Petro Karashchenko <petro.karashchenko@ring.com>
2072 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2073 attributes if needed.
2074 (prepare_move_operands): Generate special unspec instruction for
2076 (arc_isuncached_mem_p): Propagate uncached attribute to each
2078 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2079 (VUNSPEC_ARC_STDI): Likewise.
2080 (ALLI): New mode iterator.
2081 (mALLI): New mode attribute.
2082 (lddi): New instruction pattern.
2084 (stdidi_split): Split instruction for architectures which are not
2085 supporting ll64 option.
2086 (lddidi_split): Likewise.
2088 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2090 PR rtl-optimization/92989
2091 * lra-lives.c (process_bb_lives): Update the live-in set before
2092 processing additional clobbers.
2094 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2096 PR rtl-optimization/93170
2097 * cselib.c (cselib_invalidate_regno_val): New function, split out
2099 (cselib_invalidate_regno): ...here.
2100 (cselib_invalidated_by_call_p): New function.
2101 (cselib_process_insn): Iterate over all the hard-register entries in
2102 REG_VALUES and invalidate any that cross call-clobbered registers.
2104 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2106 * dojump.c (split_comparison): Use HONOR_NANS rather than
2107 HONOR_SNANS when splitting LTGT.
2109 2020-01-27 Martin Liska <mliska@suse.cz>
2112 * opts.c (print_filtered_help): Exclude language-specific
2113 options from --help=common unless enabled in all FEs.
2115 2020-01-27 Martin Liska <mliska@suse.cz>
2117 * opts.c (print_help): Exclude params from
2118 all except --help=param.
2120 2020-01-27 Martin Liska <mliska@suse.cz>
2123 * config/i386/i386-features.c (make_resolver_func):
2124 Align the code with ppc64 target implementation.
2125 Do not generate a unique name for resolver function.
2127 2020-01-27 Richard Biener <rguenther@suse.de>
2129 PR tree-optimization/93397
2130 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2131 converted reduction chain SLP graph adjustment.
2133 2020-01-26 Marek Polacek <polacek@redhat.com>
2136 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2139 2020-01-26 Jason Merrill <jason@redhat.com>
2142 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2145 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2147 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2148 (rx_setmem): Likewise.
2150 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2153 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2154 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2155 drop <di> from constraint of last operand.
2158 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2159 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2160 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2162 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2165 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2168 2020-01-24 Jeff Law <law@redhat.com>
2170 PR tree-optimization/92788
2171 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2174 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2177 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2178 *avx_vperm_broadcast_<mode>,
2179 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2180 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2181 Move before avx2_perm<mode>/avx512f_perm<mode>.
2184 * simplify-rtx.c (simplify_const_unary_operation,
2185 simplify_const_binary_operation): Punt for mode precision above
2186 MAX_BITSIZE_MODE_ANY_INT.
2188 2020-01-24 Andrew Pinski <apinski@marvell.com>
2190 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2193 2020-01-24 Jeff Law <law@redhat.com>
2196 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2197 for REGs. Call output_operand_lossage to get more reasonable
2200 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2202 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2203 gcn_fp_compare_operator.
2204 (vec_cmpu<mode>di): Use gcn_compare_operator.
2205 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2206 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2207 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2208 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2209 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2210 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2211 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2212 gcn_fp_compare_operator.
2213 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2214 gcn_fp_compare_operator.
2215 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2216 gcn_fp_compare_operator.
2217 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2218 gcn_fp_compare_operator.
2220 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2222 * doc/install.texi (Cross-Compiler-Specific Options): Document
2223 `--with-toolexeclibdir' option.
2225 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2227 * target.def (flags_regnum): Also mention effect on delay slot filling.
2228 * doc/tm.texi: Regenerate.
2230 2020-01-23 Jeff Law <law@redhat.com>
2232 PR translation/90162
2233 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2235 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2238 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2241 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2243 PR rtl-optimization/93402
2244 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2247 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2249 * config.in: Regenerated.
2250 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2251 for TARGET_LIBC_GNUSTACK.
2252 * configure: Regenerated.
2253 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2254 found to be 2.31 or greater.
2256 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2258 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2260 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2261 (mips_asm_file_end): New function. Delegate to
2262 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2263 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2265 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2268 * config/i386/i386-modes.def (POImode): New mode.
2269 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2270 * config/i386/i386.md (DPWI): New mode attribute.
2271 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2273 (QPWI): ... this. Use POI instead of OI for TImode.
2274 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2275 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2278 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2281 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2283 (speculation_tracker_rev): New pattern.
2284 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2285 Use speculation_tracker_rev to track the inverse condition.
2287 2020-01-23 Richard Biener <rguenther@suse.de>
2289 PR tree-optimization/93381
2290 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2291 alias-set of the def as argument and record the first one.
2292 (vn_walk_cb_data::first_set): New member.
2293 (vn_reference_lookup_3): Pass the alias-set of the current def
2294 to push_partial_def. Fix alias-set used in the aggregate copy
2296 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2297 * real.c (clear_significand_below): Fix out-of-bound access.
2299 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2302 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2303 New define_insn patterns.
2305 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2307 * doc/sourcebuild.texi (check-function-bodies): Add an
2308 optional target/xfail selector.
2310 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2312 PR rtl-optimization/93124
2313 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2314 bare USE and CLOBBER insns.
2316 2020-01-22 Andrew Pinski <apinski@marvell.com>
2318 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2320 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2323 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2324 diagnostic_manager into "ana" namespace.
2325 * selftest-run-tests.c (selftest::run_tests): Update for move of
2326 selftest::run_analyzer_selftests to
2327 ana::selftest::run_analyzer_selftests.
2329 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2331 * cfgexpand.c (union_stack_vars): Update the size.
2333 2020-01-22 Richard Biener <rguenther@suse.de>
2335 PR tree-optimization/93381
2336 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2337 throughout, handle all conversions the same.
2339 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2342 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2343 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2344 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2345 Call force_reg on high_in2 unconditionally.
2347 2020-01-22 Martin Liska <mliska@suse.cz>
2349 PR tree-optimization/92924
2350 * profile.c (compute_value_histograms): Divide
2353 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2356 * output.h (assemble_name_resolve): Declare.
2357 * varasm.c (assemble_name_resolve): New function.
2358 (assemble_name): Use it.
2359 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2361 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2363 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2364 update_web_docs_git instead of update_web_docs_svn.
2366 2020-01-21 Andrew Pinski <apinski@marvell.com>
2369 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2370 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2371 (*tlsgd_small_<mode>): Likewise.
2372 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2373 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2374 register. Convert that register back to dest using convert_mode.
2376 2020-01-21 Jim Wilson <jimw@sifive.com>
2378 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2381 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2382 Uros Bizjak <ubizjak@gmail.com>
2385 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2387 (legitimize_tls_address): Do GNU2 TLS address computation in
2388 ptr_mode and zero-extend result to Pmode.
2389 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2390 :P with :PTR and Pmode with ptr_mode.
2391 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2392 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2393 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2395 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2398 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2399 the last two operands are CONST_INT_P before using them as such.
2401 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2403 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2404 to get the integer element types.
2406 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2408 * config/aarch64/aarch64-sve-builtins.h
2409 (function_expander::convert_to_pmode): Declare.
2410 * config/aarch64/aarch64-sve-builtins.cc
2411 (function_expander::convert_to_pmode): New function.
2412 (function_expander::get_contiguous_base): Use it.
2413 (function_expander::prepare_gather_address_operands): Likewise.
2414 * config/aarch64/aarch64-sve-builtins-sve2.cc
2415 (svwhilerw_svwhilewr_impl::expand): Likewise.
2417 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2420 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2421 cfun->machine->label_is_assembled.
2422 (aarch64_print_patchable_function_entry): New.
2423 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2424 * config/aarch64/aarch64.h (struct machine_function): New field,
2427 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2430 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2433 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2436 * cgraph.c (cgraph_edge::resolve_speculation,
2437 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2438 call_stmt_site_hash.
2440 2020-01-21 Martin Liska <mliska@suse.cz>
2442 * config/rs6000/rs6000.c (common_mode_defined): Remove
2445 2020-01-21 Richard Biener <rguenther@suse.de>
2447 PR tree-optimization/92328
2448 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2449 type when value-numbering same-sized store by inserting a
2451 (eliminate_dom_walker::eliminate_stmt): When eliminating
2452 a redundant store handle bit-reinterpretation of the same value.
2454 2020-01-21 Andrew Pinski <apinski@marvel.com>
2457 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2459 (prepare_block_for_update): This. Use a worklist instead of
2462 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2464 * gcc/config/arm/arm.c (clear_operation_p):
2465 Initialise last_regno, skip first iteration
2466 based on the first_set value and use ints instead
2467 of the unnecessary HOST_WIDE_INTs.
2469 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2472 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2473 compare_mode other than SFmode or DFmode.
2475 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2478 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2479 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2480 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2482 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2484 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2486 2020-01-20 Andrew Pinski <apinski@marvell.com>
2489 * targhooks.c (default_print_patchable_function_entry): Use
2490 output_asm_insn to emit the nop instruction.
2492 2020-01-20 Fangrui Song <maskray@google.com>
2495 * targhooks.c (default_print_patchable_function_entry): Align to
2498 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2501 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2502 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2503 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2504 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2505 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2506 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2507 Remove the {q} suffix from lea.
2508 (*tls_dynamic_gnu2_call_64): Renamed to ...
2509 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2510 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2511 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2512 Pass Pmode to gen_tls_dynamic_gnu2_64.
2514 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2516 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2518 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2520 * config/aarch64/aarch64-sve-builtins-base.cc
2521 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2523 2020-01-20 Richard Biener <rguenther@suse.de>
2526 * dwarf2out.c (prune_unused_types): Unconditionally mark
2527 called function DIEs.
2529 2020-01-20 Martin Liska <mliska@suse.cz>
2531 PR tree-optimization/93199
2532 * tree-eh.c (struct leh_state): Add
2533 new field outer_non_cleanup.
2534 (cleanup_is_dead_in): Pass leh_state instead
2535 of eh_region. Add a checking that state->outer_non_cleanup
2536 points to outer non-clean up region.
2537 (lower_try_finally): Record outer_non_cleanup
2539 (lower_catch): Likewise.
2540 (lower_eh_filter): Likewise.
2541 (lower_eh_must_not_throw): Likewise.
2542 (lower_cleanup): Likewise.
2544 2020-01-20 Richard Biener <rguenther@suse.de>
2546 PR tree-optimization/93094
2547 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2548 (vect_transform_loop): Likewise.
2549 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2550 loop_vectorized_call to vect_transform_loop.
2551 * tree-vect-loop.c (vect_transform_loop): Pass down
2552 loop_vectorized_call to vect_loop_versioning.
2553 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2554 the earlier discovered loop_vectorized_call.
2556 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2558 * doc/contribute.texi: Update for SVN -> Git transition.
2559 * doc/install.texi: Likewise.
2561 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2564 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2565 speculative targets.
2566 (verify_speculative_call): New function
2567 (cgraph_node::verify_node): Use it.
2568 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2571 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2574 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2575 (cgraph_edge::make_direct): Remove all indirect targets.
2576 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2577 (cgraph_node::verify_node): Verify that only one call_stmt or
2578 lto_stmt_uid is set.
2579 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2581 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2582 (lto_output_ref): Simplify streaming of stmt.
2583 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2585 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2587 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2588 Mark parameter unused.
2590 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2592 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2594 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2596 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2598 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2600 * Makefile.in: Add coroutine-passes.o.
2601 * builtin-types.def (BT_CONST_SIZE): New.
2602 (BT_FN_BOOL_PTR): New.
2603 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2604 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2605 * coroutine-builtins.def: New file.
2606 * coroutine-passes.cc: New file.
2607 * function.h (struct GTY function): Add a bit to indicate that the
2608 function is a coroutine component.
2609 * internal-fn.c (expand_CO_FRAME): New.
2610 (expand_CO_YIELD): New.
2611 (expand_CO_SUSPN): New.
2612 (expand_CO_ACTOR): New.
2613 * internal-fn.def (CO_ACTOR): New.
2617 * passes.def: Add pass_coroutine_lower_builtins,
2618 pass_coroutine_early_expand_ifns.
2619 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2620 (make_pass_coroutine_early_expand_ifns): New.
2621 * doc/invoke.texi: Document the fcoroutines command line
2624 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2626 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2629 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2630 after checking the argument is a REG. Don't use REGNO (reg)
2631 again to set last_regno, reuse regno variable instead.
2633 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2635 * doc/analyzer.texi (Limitations): Add note about NaN.
2637 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2638 Sudakshina Das <sudi.das@arm.com>
2640 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2641 and valid immediate.
2642 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2643 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2644 * config/arm/constraints.md (Pg): New.
2645 * config/arm/predicates.md (long_shift_imm): New.
2646 (arm_reg_or_long_shift_imm): Likewise.
2647 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2648 (thumb2_lsll): Likewise.
2651 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2652 Sudakshina Das <sudi.das@arm.com>
2654 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2655 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2656 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2657 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2658 * config/arm/thumb2.md (thumb2_asrl): New.
2659 (thumb2_lsll): Likewise.
2661 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2663 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2666 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2668 * gdbinit.in (help-gcc-hooks): New command.
2669 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2670 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2673 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2675 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2676 correct target macro.
2678 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2680 * config/aarch64/aarch64-protos.h
2681 (aarch64_sve_ld1ro_operand_p): New.
2682 * config/aarch64/aarch64-sve-builtins-base.cc
2683 (class load_replicate): New.
2684 (class svld1ro_impl): New.
2685 (class svld1rq_impl): Change to inherit from load_replicate.
2686 (svld1ro): New sve intrinsic function base.
2687 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2688 New DEF_SVE_FUNCTION.
2689 * config/aarch64/aarch64-sve-builtins-base.h
2690 (svld1ro): New decl.
2691 * config/aarch64/aarch64-sve-builtins.cc
2692 (function_expander::add_mem_operand): Modify assert to allow
2694 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2696 * config/aarch64/aarch64.c
2697 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2698 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2699 (aarch64_sve_ld1ro_operand_p): New.
2700 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2701 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2702 * config/aarch64/predicates.md
2703 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2705 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2707 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2708 Introduce this ACLE specified predefined macro.
2709 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2710 (fp): Disabling this disables f64mm.
2711 (simd): Disabling this disables f64mm.
2712 (fp16): Disabling this disables f64mm.
2713 (sve): Disabling this disables f64mm.
2714 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2715 (AARCH64_ISA_F64MM): New.
2716 (TARGET_F64MM): New.
2717 * doc/invoke.texi (f64mm): Document new option.
2719 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2721 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2722 (neoversen1_tunings): Likewise.
2724 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2727 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2728 Add assert to ensure prolog has been emitted.
2729 (aarch64_split_atomic_op): Likewise.
2730 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2731 Use epilogue_completed rather than reload_completed.
2732 (aarch64_atomic_exchange<mode>): Likewise.
2733 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2734 (atomic_nand<mode>): Likewise.
2735 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2736 (atomic_fetch_nand<mode>): Likewise.
2737 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2738 (atomic_nand_fetch<mode>): Likewise.
2740 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2743 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2745 (REVERSE_CONDITION): Delete.
2746 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2747 (CCFP_CCFPE): Likewise.
2748 (e): New mode attribute.
2749 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2750 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2751 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2752 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2753 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2754 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2755 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2756 name of generator from gen_ccmpdi to gen_ccmpccdi.
2757 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2758 the previous comparison but aren't able to, use the new ccmp_rev
2761 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2763 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2764 than testing directly for INTEGER_CST.
2765 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2767 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2769 PR tree-optimization/93292
2770 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2771 get_vectype_for_scalar_type returns NULL.
2773 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2775 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2776 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2778 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2780 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2782 * params.opt: (max-predicted-iterations): Set bounds.
2783 * predict.c (real_almost_one, real_br_prob_base,
2784 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2785 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2786 probabilities; do not truncate to reg_br_prob_bases.
2787 (estimate_loops_at_level): Pass max_cyclic_prob.
2788 (estimate_loops): Compute max_cyclic_prob.
2789 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2791 * profile-count.c (profile_probability::to_sreal): New.
2792 * profile-count.h (class sreal): Move up in file.
2793 (profile_probability::to_sreal): Declare.
2795 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2798 (arm_invalid_conversion): New function for target hook.
2799 (arm_invalid_unary_op): New function for target hook.
2800 (arm_invalid_binary_op): New function for target hook.
2802 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2804 * config.gcc: Add arm_bf16.h.
2805 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2806 (arm_simd_builtin_std_type): Add BFmode.
2807 (arm_init_simd_builtin_types): Define element types for vector types.
2808 (arm_init_bf16_types): New function.
2809 (arm_init_builtins): Add arm_init_bf16_types function call.
2810 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2811 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2812 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2813 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2814 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2815 (arm_mangle_type): Add __bf16.
2816 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2817 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2818 arm_bf16_ptr_type_node.
2819 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2820 define_split between ARM registers.
2821 * config/arm/arm_bf16.h: New file.
2822 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2823 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2825 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2826 * config/arm/vfp.md: Add BFmode to movhf patterns.
2828 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2829 Andre Vieira <andre.simoesdiasvieira@arm.com>
2831 * config/arm/arm-cpus.in (mve, mve_float): New features.
2832 (dsp, mve, mve.fp): New options.
2833 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2834 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2835 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2837 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2838 Thomas Preud'homme <thomas.preudhomme@arm.com>
2840 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2842 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2843 error for using -mcmse when targeting Armv8.1-M Mainline.
2845 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2846 Thomas Preud'homme <thomas.preudhomme@arm.com>
2848 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2849 address in r4 when targeting Armv8.1-M Mainline.
2850 (nonsecure_call_value_internal): Likewise.
2851 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2852 a register match_operand again. Emit BLXNS when targeting
2854 (nonsecure_call_value_reg_thumb2): Likewise.
2856 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2857 Thomas Preud'homme <thomas.preudhomme@arm.com>
2859 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2860 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2861 variable as true when floating-point ABI is not hard. Replace
2862 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2863 Generate VLSTM and VLLDM instruction respectively before and
2864 after a function call to cmse_nonsecure_call function.
2865 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2866 (VUNSPEC_VLLDM): Likewise.
2867 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2868 (lazy_load_multiple_insn): Likewise.
2870 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2871 Thomas Preud'homme <thomas.preudhomme@arm.com>
2873 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2874 (arm_emit_vfp_multi_reg_pop): Likewise.
2875 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2876 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2877 restore callee-saved VFP registers.
2879 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2880 Thomas Preud'homme <thomas.preudhomme@arm.com>
2882 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2883 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2884 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2885 callee-saved GPRs as well as clear ip register before doing a nonsecure
2886 call then restore callee-saved GPRs after it when targeting
2888 (arm_reorg): Adapt to function rename.
2890 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2891 Thomas Preud'homme <thomas.preudhomme@arm.com>
2893 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2894 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2895 clear_vfp_multiple pattern based on a new vfp parameter.
2896 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2897 targeting Armv8.1-M Mainline.
2898 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2899 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2900 whether VFP registers are available before looking call_used_regs for a
2902 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2903 of prototype of clear_operation_p.
2904 (clear_vfp_multiple_operation): New predicate.
2905 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2906 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2908 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2909 Thomas Preud'homme <thomas.preudhomme@arm.com>
2911 * config/arm/arm-protos.h (clear_operation_p): Declare.
2912 * config/arm/arm.c (clear_operation_p): New function.
2913 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2914 targeting Armv8.1-M Mainline or successor.
2915 (output_return_instruction): Only output APSR register clearing if
2916 Armv8.1-M Mainline instructions not available.
2917 (thumb_exit): Likewise.
2918 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2919 * config/arm/thumb2.md (clear_apsr): New define_insn.
2920 (clear_multiple): Likewise.
2921 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2923 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2924 Thomas Preud'homme <thomas.preudhomme@arm.com>
2926 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2927 (use_return_insn): Also return false for Armv8.1-M Mainline.
2928 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2929 Mainline instructions are available.
2930 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2931 when targeting Armv8.1-M Mainline Security Extensions.
2932 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2933 Mainline entry function.
2934 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2935 targeting Armv8.1-M Mainline or successor.
2936 (arm_expand_epilogue): Fix indentation of caller-saved register
2937 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2939 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2940 (FP_SYSREGS): Likewise.
2941 (enum vfp_sysregs_encoding): Define enum.
2942 (fp_sysreg_names): Declare.
2943 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2944 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2945 (pop_fpsysreg_insn): Likewise.
2947 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2948 Thomas Preud'homme <thomas.preudhomme@arm.com>
2950 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2951 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2952 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2953 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2954 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2955 (ARMv8_1m_main): New feature group.
2956 (armv8.1-m.main): New architecture.
2957 * config/arm/arm-tables.opt: Regenerate.
2958 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2959 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2960 (arm_options_perform_arch_sanity_checks): Error out when targeting
2961 Armv8.1-M Mainline Security Extensions.
2962 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2964 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2966 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2967 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2968 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2969 aarch64_bfdot_laneq): New.
2970 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2971 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2972 vbfdotq_laneq_f32): New.
2973 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2974 VBFMLA_W, VBF): New.
2975 (isquadop): Add V4BF, V8BF.
2977 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2979 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2980 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2981 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2982 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2983 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2984 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2985 usdot_laneq, sudot_lane,sudot_laneq): New.
2986 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2987 (aarch64_<sur>dot_lane): New.
2988 * config/aarch64/arm_neon.h (vusdot_s32): New.
2990 (vusdot_lane_s32): New.
2991 (vsudot_lane_s32): New.
2992 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2993 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2995 2020-01-16 Martin Liska <mliska@suse.cz>
2997 * value-prof.c (dump_histogram_value): Fix
2998 obvious spacing issue.
3000 2020-01-16 Andrew Pinski <apinski@marvell.com>
3002 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
3003 !storage_order_barrier_p.
3005 2020-01-16 Andrew Pinski <apinski@marvell.com>
3007 * sched-int.h (_dep): Add unused bit-field field for the padding.
3008 * sched-deps.c (init_dep_1): Init unused field.
3010 2020-01-16 Andrew Pinski <apinski@marvell.com>
3012 * optabs.h (create_expand_operand): Initialize target field also.
3014 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
3016 PR tree-optimization/92429
3017 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
3018 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
3020 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
3023 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
3025 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
3026 aarch64_sve_int_mode to each mode.
3028 2020-01-15 David Malcolm <dmalcolm@redhat.com>
3030 * doc/analyzer.texi (Overview): Add note about
3031 -fdump-ipa-analyzer.
3033 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
3035 PR tree-optimization/93231
3036 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
3037 input_type is unsigned. Use tree_to_shwi for shift constant.
3038 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
3039 (simplify_count_trailing_zeroes): Add test to handle known non-zero
3040 inputs more efficiently.
3042 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3044 * config/i386/i386.md (*movsf_internal): Do not require
3045 SSE2 ISA for alternatives 14 and 15.
3047 2020-01-15 Richard Biener <rguenther@suse.de>
3050 * tree-eh.c (sink_clobbers): If we already visited the destination
3051 block do not defer insertion.
3052 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3053 the purpose of defered insertion.
3055 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3057 * BASE-VER: Bump to 10.0.1.
3059 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3061 PR tree-optimization/93247
3062 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3063 type of the stmt that we're going to vectorize.
3065 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3067 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3068 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3071 2020-01-15 Martin Liska <mliska@suse.cz>
3073 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3074 2 calls of streamer_read_hwi in a function call.
3076 2020-01-15 Richard Biener <rguenther@suse.de>
3078 * alias.c (record_alias_subset): Avoid redundant work when
3079 subset is already recorded.
3081 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3083 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3084 the analyzer options provide CWE identifiers.
3086 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3088 * tree-diagnostic-path.cc (path_summary::event_range::print):
3089 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3090 using get_pure_location.
3092 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3094 PR tree-optimization/93262
3095 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3096 perform head trimming only if the last argument is constant,
3097 either all ones, or larger or equal to head trim, in the latter
3098 case decrease the last argument by head_trim.
3100 PR tree-optimization/93249
3101 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3102 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3103 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3104 perform head trim unless we can prove there are no '\0' chars
3105 from the source among the first head_trim chars.
3107 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3109 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3111 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3114 * config/i386/sse.md
3115 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3116 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3117 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3118 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3119 just a single alternative instead of two, make operands 1 and 2
3122 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3125 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3128 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3130 * Makefile.in (lang_opt_files): Add analyzer.opt.
3131 (ANALYZER_OBJS): New.
3132 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3133 tristate.o and ANALYZER_OBJS.
3134 (TEXI_GCCINT_FILES): Add analyzer.texi.
3135 * common.opt (-fanalyzer): New driver option.
3136 * config.in: Regenerate.
3137 * configure: Regenerate.
3138 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3139 (gccdepdir): Also create depdir for "analyzer" subdir.
3140 * digraph.cc: New file.
3141 * digraph.h: New file.
3142 * doc/analyzer.texi: New file.
3143 * doc/gccint.texi ("Static Analyzer") New menu item.
3144 (analyzer.texi): Include it.
3145 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3146 ("Warning Options"): Add static analysis warnings to the list.
3147 (-Wno-analyzer-double-fclose): New option.
3148 (-Wno-analyzer-double-free): New option.
3149 (-Wno-analyzer-exposure-through-output-file): New option.
3150 (-Wno-analyzer-file-leak): New option.
3151 (-Wno-analyzer-free-of-non-heap): New option.
3152 (-Wno-analyzer-malloc-leak): New option.
3153 (-Wno-analyzer-possible-null-argument): New option.
3154 (-Wno-analyzer-possible-null-dereference): New option.
3155 (-Wno-analyzer-null-argument): New option.
3156 (-Wno-analyzer-null-dereference): New option.
3157 (-Wno-analyzer-stale-setjmp-buffer): New option.
3158 (-Wno-analyzer-tainted-array-index): New option.
3159 (-Wno-analyzer-use-after-free): New option.
3160 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3161 (-Wno-analyzer-use-of-uninitialized-value): New option.
3162 (-Wanalyzer-too-complex): New option.
3163 (-fanalyzer-call-summaries): New warning.
3164 (-fanalyzer-checker=): New warning.
3165 (-fanalyzer-fine-grained): New warning.
3166 (-fno-analyzer-state-merge): New warning.
3167 (-fno-analyzer-state-purge): New warning.
3168 (-fanalyzer-transitivity): New warning.
3169 (-fanalyzer-verbose-edges): New warning.
3170 (-fanalyzer-verbose-state-changes): New warning.
3171 (-fanalyzer-verbosity=): New warning.
3172 (-fdump-analyzer): New warning.
3173 (-fdump-analyzer-callgraph): New warning.
3174 (-fdump-analyzer-exploded-graph): New warning.
3175 (-fdump-analyzer-exploded-nodes): New warning.
3176 (-fdump-analyzer-exploded-nodes-2): New warning.
3177 (-fdump-analyzer-exploded-nodes-3): New warning.
3178 (-fdump-analyzer-supergraph): New warning.
3179 * doc/sourcebuild.texi (dg-require-dot): New.
3180 (dg-check-dot): New.
3181 * gdbinit.in (break-on-saved-diagnostic): New command.
3182 * graphviz.cc: New file.
3183 * graphviz.h: New file.
3184 * ordered-hash-map-tests.cc: New file.
3185 * ordered-hash-map.h: New file.
3186 * passes.def (pass_analyzer): Add before
3187 pass_ipa_whole_program_visibility.
3188 * selftest-run-tests.c (selftest::run_tests): Call
3189 selftest::ordered_hash_map_tests_cc_tests.
3190 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3192 * shortest-paths.h: New file.
3193 * timevar.def (TV_ANALYZER): New timevar.
3194 (TV_ANALYZER_SUPERGRAPH): Likewise.
3195 (TV_ANALYZER_STATE_PURGE): Likewise.
3196 (TV_ANALYZER_PLAN): Likewise.
3197 (TV_ANALYZER_SCC): Likewise.
3198 (TV_ANALYZER_WORKLIST): Likewise.
3199 (TV_ANALYZER_DUMP): Likewise.
3200 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3201 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3202 * tree-pass.h (make_pass_analyzer): New decl.
3203 * tristate.cc: New file.
3204 * tristate.h: New file.
3206 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3209 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3210 alternatives 9 and 10.
3212 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3214 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3215 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3216 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3217 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3218 (selftest::hash_map_tests_c_tests): Call it.
3219 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3220 New static constant, using the value of = H::empty_zero_p.
3221 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3222 from default_hash_traits <Value>.
3223 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3225 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3226 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3227 calls to mark_empty with !Descriptor::empty_zero_p.
3228 (hash_table::empty_slow): Conditionalize the memset call with a
3229 check that Descriptor::empty_zero_p; otherwise, loop through the
3230 entries calling mark_empty on them.
3231 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3232 (pointer_hash::empty_zero_p): Likewise.
3233 (pair_hash::empty_zero_p): Likewise.
3234 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3236 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3237 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3238 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3239 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3240 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3241 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3242 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3243 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3245 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3248 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3250 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3251 fix typo on return value.
3253 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3256 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3258 (cgraph_edge::make_speculative): Add param for setting speculative_id
3260 (cgraph_edge::speculative_call_info): Update comments and find reference
3261 by speculative_id for multiple indirect targets.
3262 (cgraph_edge::resolve_speculation): Decrease the speculations
3263 for indirect edge, drop it's speculative if not direct target
3264 left. Update comments.
3265 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3266 (cgraph_node::dump): Print num_speculative_call_targets.
3267 (cgraph_node::verify_node): Don't report error if speculative
3268 edge not include statement.
3269 (cgraph_edge::num_speculative_call_targets_p): New function.
3270 * cgraph.h (int common_target_id): Remove.
3271 (int common_target_probability): Remove.
3272 (num_speculative_call_targets): New variable.
3273 (make_speculative): Add param for setting speculative_id.
3274 (cgraph_edge::num_speculative_call_targets_p): New declare.
3275 (target_prob): New variable.
3276 (speculative_id): New variable.
3277 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3278 call summaries for multiple speculative call targets.
3279 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3280 * ipa-profile.c (struct speculative_call_target): New struct.
3281 (class speculative_call_summary): New class.
3282 (class speculative_call_summaries): New class.
3283 (call_sums): New variable.
3284 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3285 (ipa_profile_write_edge_summary): New function.
3286 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3287 (ipa_profile_dump_all_summaries): New function.
3288 (ipa_profile_read_edge_summary): New function.
3289 (ipa_profile_read_summary_section): New function.
3290 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3291 (ipa_profile): Generate num_speculative_call_targets from
3293 * ipa-ref.h (speculative_id): New variable.
3294 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3295 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3296 common_target_probability. Stream out speculative_id and
3297 num_speculative_call_targets.
3298 (input_edge): Likewise.
3299 * predict.c (dump_prediction): Remove edges count assert to be
3301 * symtab.c (symtab_node::create_reference): Init speculative_id.
3302 (symtab_node::clone_references): Clone speculative_id.
3303 (symtab_node::clone_referring): Clone speculative_id.
3304 (symtab_node::clone_reference): Clone speculative_id.
3305 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3306 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3307 if indirect call contains multiple speculative targets.
3308 * value-prof.h (check_ic_target): Remove.
3309 * value-prof.c (gimple_value_profile_transformations):
3310 Use void function gimple_ic_transform.
3311 * value-prof.c (gimple_ic_transform): Handle topn case.
3312 Fix comment typos. Change it to a void function.
3314 2020-01-13 Andrew Pinski <apinski@marvell.com>
3316 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3317 (octeontx2t98): New define.
3318 (octeontx2t96): New define.
3319 (octeontx2t93): New define.
3320 (octeontx2f95): New define.
3321 (octeontx2f95n): New define.
3322 (octeontx2f95mm): New define.
3323 * config/aarch64/aarch64-tune.md: Regenerate.
3324 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3326 2020-01-13 Jason Merrill <jason@redhat.com>
3328 PR c++/33799 - destroy return value if local cleanup throws.
3329 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3331 2020-01-13 Martin Liska <mliska@suse.cz>
3333 * ipa-cp.c (get_max_overall_size): Use newly
3334 renamed param param_ipa_cp_unit_growth.
3335 * params.opt: Remove legacy param name.
3337 2020-01-13 Martin Sebor <msebor@redhat.com>
3339 PR tree-optimization/93213
3340 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3341 stores to be eliminated.
3343 2020-01-13 Martin Liska <mliska@suse.cz>
3345 * opts.c (print_help): Do not print CL_PARAM
3346 and CL_WARNING for CL_OPTIMIZATION.
3348 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3351 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3352 depending on optimization settings.
3354 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3356 PR tree-optimization/90838
3357 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3358 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3359 argument rather than to initialize temporary for targets that
3360 don't use the mode argument at all. Initialize ctzval to avoid
3363 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3365 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3366 * tree-core.h: Document it.
3367 * gimplify.c (gimplify_omp_workshare): Set it.
3368 * omp-low.c (lower_omp_target): Use it.
3369 * tree-pretty-print.c (dump_omp_clause): Print it.
3371 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3372 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3374 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3376 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3377 * common.opt (fdiagnostics-path-format=): New option.
3378 (diagnostic_path_format): New enum.
3379 (fdiagnostics-show-path-depths): New option.
3380 * coretypes.h (diagnostic_event_id_t): New forward decl.
3381 * diagnostic-color.c (color_dict): Add "path".
3382 * diagnostic-event-id.h: New file.
3383 * diagnostic-format-json.cc (json_from_expanded_location): Make
3385 (json_end_diagnostic): Call context->make_json_for_path if it
3386 exists and the diagnostic has a path.
3387 (diagnostic_output_format_init): Clear context->print_path.
3388 * diagnostic-path.h: New file.
3389 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3390 when printing a run of events in a diagnostic_path so that they
3391 all get the same color.
3392 (layout::m_diagnostic_path_p): New field.
3393 (layout::layout): Initialize it.
3394 (layout::print_any_labels): Don't colorize the label text for an
3395 event in a diagnostic_path.
3396 (gcc_rich_location::add_location_if_nearby): Add
3397 "restrict_to_current_line_spans" and "label" params. Pass the
3398 former to layout.maybe_add_location_range; pass the latter
3399 when calling add_range.
3400 * diagnostic.c: Include "diagnostic-path.h".
3401 (diagnostic_initialize): Initialize context->path_format and
3402 context->show_path_depths.
3403 (diagnostic_show_any_path): New function.
3404 (diagnostic_path::interprocedural_p): New function.
3405 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3406 (simple_diagnostic_path::num_events): New function.
3407 (simple_diagnostic_path::get_event): New function.
3408 (simple_diagnostic_path::add_event): New function.
3409 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3410 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3411 (debug): New overload taking a diagnostic_path *.
3412 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3413 * diagnostic.h (enum diagnostic_path_format): New enum.
3414 (json::value): New forward decl.
3415 (diagnostic_context::path_format): New field.
3416 (diagnostic_context::show_path_depths): New field.
3417 (diagnostic_context::print_path): New callback field.
3418 (diagnostic_context::make_json_for_path): New callback field.
3419 (diagnostic_show_any_path): New decl.
3420 (json_from_expanded_location): New decl.
3421 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3422 (-fdiagnostics-show-path-depths): New option.
3423 (-fdiagnostics-color): Add "path" to description of default
3424 GCC_COLORS; describe it.
3425 (-fdiagnostics-format=json): Document how diagnostic paths are
3426 represented in the JSON output format.
3427 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3428 Add optional params "restrict_to_current_line_spans" and "label".
3429 * opts.c (common_handle_option): Handle
3430 OPT_fdiagnostics_path_format_ and
3431 OPT_fdiagnostics_show_path_depths.
3432 * pretty-print.c: Include "diagnostic-event-id.h".
3433 (pp_format): Implement "%@" format code for printing
3434 diagnostic_event_id_t *.
3435 (selftest::test_pp_format): Add tests for "%@".
3436 * selftest-run-tests.c (selftest::run_tests): Call
3437 selftest::tree_diagnostic_path_cc_tests.
3438 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3439 * toplev.c (general_init): Initialize global_dc->path_format and
3440 global_dc->show_path_depths.
3441 * tree-diagnostic-path.cc: New file.
3442 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3443 non-static. Drop "diagnostic" param in favor of storing the
3444 original value of "where" and re-using it.
3445 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3446 maybe_unwind_expanded_macro_loc.
3447 (tree_diagnostics_defaults): Initialize context->print_path and
3448 context->make_json_for_path.
3449 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3451 (default_tree_make_json_for_path): New decl.
3452 (maybe_unwind_expanded_macro_loc): New decl.
3454 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3456 PR tree-optimization/93210
3457 * fold-const.h (native_encode_initializer,
3458 can_native_interpret_type_p): Declare.
3459 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3461 (native_encode_initializer): New function, moved from dwarf2out.c.
3462 Adjust to native_encode_expr compatible arguments, including dry-run
3463 and partial extraction modes. Don't handle STRING_CST.
3464 (can_native_interpret_type_p): No longer static.
3465 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3466 offset / BITS_PER_UNIT fits into int and don't call it if
3467 can_native_interpret_type_p fails. If suboff is NULL and for
3468 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3469 native_encode_initializer.
3470 (fold_const_aggregate_ref_1): Formatting fix.
3471 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3472 (tree_add_const_value_attribute): Adjust caller.
3474 PR tree-optimization/90838
3475 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3476 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3477 CTZ_DEFINED_VALUE_AT_ZERO.
3479 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3482 * lra-constraints.c (match_reload): Permit input operands have the
3483 same mode as output while other input operands have a different
3486 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3488 PR tree-optimization/90838
3489 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3490 (check_ctz_string): Likewise.
3491 (optimize_count_trailing_zeroes): Likewise.
3492 (simplify_count_trailing_zeroes): Likewise.
3493 (pass_forwprop::execute): Try ctz simplification.
3494 * match.pd: Add matching for ctz idioms.
3496 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3498 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3500 (aarch64_invalid_unary_op): New function for target hook.
3501 (aarch64_invalid_binary_op): New function for target hook.
3503 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3505 * config.gcc: Add arm_bf16.h.
3506 * config/aarch64/aarch64-builtins.c
3507 (aarch64_simd_builtin_std_type): Add BFmode.
3508 (aarch64_init_simd_builtin_types): Define element types for vector
3510 (aarch64_init_bf16_types): New function.
3511 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3512 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3514 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3515 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3517 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3518 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3519 * config/aarch64/aarch64.c
3520 (aarch64_classify_vector_mode): Add support for BF types.
3521 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3522 (aarch64_vq_mode): Add support for BF types.
3523 (aarch64_simd_container_mode): Add support for BF types.
3524 (aarch64_mangle_type): Add support for BF scalar type.
3525 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3526 * config/aarch64/arm_bf16.h: New file.
3527 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3528 * config/aarch64/iterators.md: Add BF types to mode attributes.
3529 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3531 2020-01-10 Jason Merrill <jason@redhat.com>
3533 PR c++/93173 - incorrect tree sharing.
3534 * gimplify.c (copy_if_shared): No longer static.
3535 * gimplify.h: Declare it.
3537 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3539 * doc/invoke.texi (-msve-vector-bits=): Document that
3540 -msve-vector-bits=128 now generates VL-specific code for
3541 little-endian targets.
3542 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3543 build_vector_type_for_mode to construct the data vector types.
3544 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3545 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3546 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3547 for 128-bit vectors.
3549 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3551 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3554 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3556 * config/aarch64/aarch64-builtins.c
3557 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3558 rather than checking the number of elements and the element mode.
3560 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3562 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3563 get_related_vectype_for_scalar_type rather than build_vector_type
3564 to create the index type for a conditional reduction.
3566 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3568 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3569 for any type of gather or scatter, including strided accesses.
3571 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3573 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3576 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3578 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3580 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3581 parameter and its use to reset DR_OFFSET's.
3582 (vect_transform_loop): Remove orig_drs_init argument.
3583 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3584 member of dr_vec_info rather than the offset of the associated
3585 data_reference's innermost_loop_behavior.
3586 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3587 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3588 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3589 get_dr_vinfo_offset.
3590 (vectorizable_store): Likewise.
3591 (vectorizable_load): Likewise.
3593 2020-01-10 Richard Biener <rguenther@suse.de>
3595 * gimple-ssa-store-merging
3596 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3598 2020-01-10 Martin Liska <mliska@suse.cz>
3601 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3602 encapsulation that was there before r280040.
3604 2020-01-10 Richard Biener <rguenther@suse.de>
3607 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3608 sequences to avoid walking them again for secondary opportunities.
3609 (pass_lower_eh_dispatch::execute): Instead actually insert
3612 2020-01-10 Richard Biener <rguenther@suse.de>
3615 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3616 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3617 avoid quadraticness.
3619 2020-01-10 Martin Jambor <mjambor@suse.cz>
3621 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3622 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3623 to get param_ipa_sra_max_replacements.
3624 (param_splitting_across_edge): Pass the caller to
3625 pull_accesses_from_callee.
3627 2020-01-10 Martin Jambor <mjambor@suse.cz>
3629 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3630 * ipa-cp.c (max_new_size): Removed.
3631 (orig_overall_size): New variable.
3632 (get_max_overall_size): New function.
3633 (estimate_local_effects): Use it. Adjust dump.
3634 (decide_about_value): Likewise.
3635 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3636 orig_overall_size. Adjust dump.
3637 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3639 2020-01-10 Martin Jambor <mjambor@suse.cz>
3641 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3642 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3643 instead of param_ipa_max_agg_items.
3644 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3645 optimization info for the callee.
3647 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3649 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3650 markers if debug_inline_points is false.
3652 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3654 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3656 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3657 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3658 aarch64-sve-builtins-sve2.h.
3659 (aarch64-sve-builtins-sve2.o): New rule.
3660 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3661 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3662 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3663 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3664 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3665 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3667 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3668 instructions that are handled here.
3669 (@cond_asrd<mode>): Generalize to...
3670 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3671 (*cond_asrd<mode>_2): Generalize to...
3672 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3673 (*cond_asrd<mode>_z): Generalize to...
3674 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3675 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3676 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3677 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3678 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3680 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3681 (@aarch64_scatter_stnt<mode>): Likewise.
3682 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3683 (@aarch64_mul_lane_<mode>): Likewise.
3684 (@aarch64_sve_suqadd<mode>_const): Likewise.
3685 (*<sur>h<addsub><mode>): Generalize to...
3686 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3688 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3689 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3690 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3691 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3692 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3693 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3694 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3695 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3696 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3697 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3698 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3699 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3700 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3701 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3702 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3703 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3704 (@aarch64_sve2_xar<mode>): Likewise.
3705 (@aarch64_sve2_bcax<mode>): Likewise.
3706 (*aarch64_sve2_eor3<mode>): Rename to...
3707 (@aarch64_sve2_eor3<mode>): ...this.
3708 (@aarch64_sve2_bsl<mode>): New expander.
3709 (@aarch64_sve2_nbsl<mode>): Likewise.
3710 (@aarch64_sve2_bsl1n<mode>): Likewise.
3711 (@aarch64_sve2_bsl2n<mode>): Likewise.
3712 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3713 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3714 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3715 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3716 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3717 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3718 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3719 (<su>mull<bt><Vwide>): Generalize to...
3720 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3722 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3723 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3724 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3725 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3726 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3727 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3728 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3729 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3730 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3731 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3732 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3733 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3734 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3735 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3736 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3737 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3738 (<SHRNB:r>shrnb<mode>): Generalize to...
3739 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3741 (<SHRNT:r>shrnt<mode>): Generalize to...
3742 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3744 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3745 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3746 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3747 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3748 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3749 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3750 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3751 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3752 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3753 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3754 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3755 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3756 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3757 (@aarch64_sve2_cvtnt<mode>): Likewise.
3758 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3759 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3760 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3761 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3762 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3763 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3764 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3765 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3766 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3767 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3768 (@aarch64_sve2_pmul<mode>): Likewise.
3769 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3770 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3771 (@aarch64_sve2_tbl2<mode>): Likewise.
3772 (@aarch64_sve2_tbx<mode>): Likewise.
3773 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3774 (@aarch64_sve2_histcnt<mode>): Likewise.
3775 (@aarch64_sve2_histseg<mode>): Likewise.
3776 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3777 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3778 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3779 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3780 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3781 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3782 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3783 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3784 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3785 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3786 (SVE2_PMULL_PAIR_I): New mode iterators.
3787 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3788 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3789 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3790 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3791 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3792 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3793 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3794 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3795 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3796 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3797 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3798 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3799 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3800 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3801 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3802 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3803 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3804 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3805 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3806 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3807 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3808 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3809 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3810 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3811 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3812 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3813 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3814 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3815 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3816 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3817 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3819 (VNARROW, Ventype): New mode attributes.
3820 (Vewtype): Handle VNx2DI. Fix typo in comment.
3821 (VDOUBLE): New mode attribute.
3822 (sve_lane_con): Handle VNx8HI.
3823 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3824 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3825 (sve_int_op, sve_int_op_rev): Handle the above codes.
3826 (sve_pred_int_rhs2_operand): Likewise.
3827 (MULLBT, SHRNB, SHRNT): Delete.
3828 (SVE_INT_SHIFT_IMM): New int iterator.
3829 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3830 and UNSPEC_WHILEHS for TARGET_SVE2.
3831 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3832 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3833 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3834 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3835 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3836 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3837 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3838 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3839 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3840 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3841 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3842 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3843 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3844 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3845 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3846 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3847 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3848 (optab): Handle the new unspecs.
3849 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3851 (lr): Handle the new unspecs.
3853 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3854 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3855 (sve_int_qsub_op): New int attributes.
3856 (sve_fp_op, rot): Handle the new unspecs.
3857 * config/aarch64/aarch64-sve-builtins.h
3858 (function_resolver::require_matching_pointer_type): Declare.
3859 (function_resolver::resolve_unary): Add an optional boolean argument.
3860 (function_resolver::finish_opt_n_resolution): Add an optional
3861 type_suffix_index argument.
3862 (gimple_folder::redirect_call): Declare.
3863 (gimple_expander::prepare_gather_address_operands): Add an optional
3865 * config/aarch64/aarch64-sve-builtins.cc: Include
3866 aarch64-sve-builtins-sve2.h.
3867 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3868 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3869 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3870 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3871 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3872 (TYPES_s_unsigned): Likewise.
3873 (TYPES_s_integer): Use TYPES_s_unsigned.
3874 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3875 (TYPES_sd_integer): Use them.
3876 (TYPES_d_unsigned): New macro.
3877 (TYPES_d_integer): Use it.
3878 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3879 (TYPES_cvt_narrow): Likewise.
3880 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3881 (preds_mx): New variable.
3882 (function_builder::add_overloaded_function): Allow the new feature
3883 set to be more restrictive than the original one.
3884 (function_resolver::infer_pointer_type): Remove qualifiers from
3885 the pointer type before printing it.
3886 (function_resolver::require_matching_pointer_type): New function.
3887 (function_resolver::resolve_sv_displacement): Handle functions
3888 that don't support 32-bit vector indices or svint32_t vector offsets.
3889 (function_resolver::finish_opt_n_resolution): Take the inferred type
3890 as a separate argument.
3891 (function_resolver::resolve_unary): Optionally treat all forms in
3892 the same way as normal merging functions.
3893 (gimple_folder::redirect_call): New function.
3894 (function_expander::prepare_gather_address_operands): Add an argument
3895 that says whether scaled forms are available. If they aren't,
3896 handle scaling of vector indices and don't add the extension and
3898 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3899 fall back to using cond_* instead.
3900 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3901 Split out the member variables into...
3902 (rtx_code_function_base): ...this new base class.
3903 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3904 (unspec_based_function): Split out the member variables into...
3905 (unspec_based_function_base): ...this new base class.
3906 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3907 (unspec_based_function_exact_insn): New class.
3908 (unspec_based_add_function, unspec_based_add_lane_function)
3909 (unspec_based_lane_function, unspec_based_pred_function)
3910 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3911 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3912 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3914 (unspec_based_fused_function): New class.
3915 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3916 (unspec_based_fused_lane_function): New class.
3917 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3919 (CODE_FOR_MODE1): New macro.
3920 (fixed_insn_function): New class.
3921 (while_comparison): Likewise.
3922 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3923 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3924 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3925 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3926 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3927 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3928 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3929 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3930 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3931 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3932 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3933 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3934 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3935 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3936 Also add an initial argument for unary_convert_narrowt, regardless
3937 of the predication type.
3938 (build_32_64): Allow loads and stores to specify MODE_none.
3939 (build_sv_index64, build_sv_uint_offset): New functions.
3940 (long_type_suffix): New function.
3941 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3942 (binary_imm_long_base, load_gather_sv_base): Likewise.
3943 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3944 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3945 (unary_narrowb_base, unary_narrowt_base): Likewise.
3946 (binary_long_lane_def, binary_long_lane): New shape.
3947 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3948 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3949 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3950 (binary_to_uint_def, binary_to_uint): Likewise.
3951 (binary_wide_def, binary_wide): Likewise.
3952 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3953 (compare_def, compare): Likewise.
3954 (compare_ptr_def, compare_ptr): Likewise.
3955 (load_ext_gather_index_restricted_def,
3956 load_ext_gather_index_restricted): Likewise.
3957 (load_ext_gather_offset_restricted_def,
3958 load_ext_gather_offset_restricted): Likewise.
3959 (load_gather_sv_def): Inherit from load_gather_sv_base.
3960 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3961 (shift_left_imm_def, shift_left_imm): Likewise.
3962 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3963 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3964 (store_scatter_index_restricted_def,
3965 store_scatter_index_restricted): Likewise.
3966 (store_scatter_offset_restricted_def,
3967 store_scatter_offset_restricted): Likewise.
3968 (tbl_tuple_def, tbl_tuple): Likewise.
3969 (ternary_long_lane_def, ternary_long_lane): Likewise.
3970 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3971 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3972 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3973 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3974 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3975 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3976 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3977 (ternary_uint_def, ternary_uint): Likewise.
3978 (unary_convert): Fix typo in comment.
3979 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3980 (unary_long_def, unary_long): Likewise.
3981 (unary_narrowb_def, unary_narrowb): Likewise.
3982 (unary_narrowt_def, unary_narrowt): Likewise.
3983 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3984 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3985 (unary_to_int_def, unary_to_int): Likewise.
3986 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3987 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3988 (svasrd_impl): Delete.
3989 (svcadd_impl::expand): Handle integer operations too.
3990 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3991 new functions to derive the unspec numbers.
3992 (svmla_svmls_lane_impl): Replace with...
3993 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3994 integer operations too.
3995 (svwhile_impl): Rename to...
3996 (svwhilelx_impl): ...this and inherit from while_comparison.
3997 (svasrd): Use unspec_based_function.
3998 (svmla_lane): Use svmla_lane_impl.
3999 (svmls_lane): Use svmls_lane_impl.
4000 (svrecpe, svrsqrte): Handle unsigned integer operations too.
4001 (svwhilele, svwhilelt): Use svwhilelx_impl.
4002 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
4003 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
4004 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
4005 * config/aarch64/aarch64-sve-builtins.def: Include
4006 aarch64-sve-builtins-sve2.def.
4008 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4010 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
4011 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
4012 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
4013 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
4014 immediates as well as vector ones.
4015 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
4016 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
4017 (aarch64_sve_qsub_immediate): Update calls accordingly.
4019 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4021 * config/aarch64/aarch64-sve2.md: Add banner comments.
4022 (<su>mulh<r>s<mode>3): Move further up file.
4023 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
4024 (*aarch64_sve2_sra<mode>): Move further down file.
4025 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
4027 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4029 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
4031 (while_optab_cmp): Handle them.
4032 * config/aarch64/aarch64-sve.md
4033 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
4034 and add a "@" marker.
4035 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
4036 instead of gen_aarch64_sve2_while_ptest.
4037 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
4039 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4041 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
4042 (UNSPEC_WHILELE): ...this.
4043 (UNSPEC_WHILE_LO): Rename to...
4044 (UNSPEC_WHILELO): ...this.
4045 (UNSPEC_WHILE_LS): Rename to...
4046 (UNSPEC_WHILELS): ...this.
4047 (UNSPEC_WHILE_LT): Rename to...
4048 (UNSPEC_WHILELT): ...this.
4049 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4050 (cmp_op, while_optab_cmp): Likewise.
4051 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4052 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4053 (svwhilelt): Likewise.
4055 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4057 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4058 (unary_to_uint): Define.
4059 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4060 (unary_count): Rename to...
4061 (unary_to_uint_def, unary_to_uint): ...this.
4062 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4064 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4066 * config/aarch64/aarch64-sve-builtins-functions.h
4067 (code_for_mode_function): New class.
4068 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4069 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4070 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4071 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4072 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4074 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4076 * config/aarch64/iterators.md (addsub): New code attribute.
4077 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4079 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4080 in the asm string and attributes. Fix indentation.
4081 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4083 (@aarch64_sve_<optab><mode>): ...this.
4084 * config/aarch64/aarch64-sve-builtins.h
4085 (function_expander::expand_signed_unpred_op): Delete.
4086 * config/aarch64/aarch64-sve-builtins.cc
4087 (function_expander::expand_signed_unpred_op): Likewise.
4088 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4089 try using code_for_aarch64_sve instead.
4090 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4091 (svqsub_impl): Likewise.
4092 (svqadd, svqsub): Use rtx_code_function instead.
4094 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4096 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4097 (HADDSUB, sur, addsub): Remove them.
4099 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4101 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4102 functions the same way, rather than singling out those that
4103 aren't mapped directly to optabs.
4105 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4107 * target.def (compatible_vector_types_p): New target hook.
4108 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4109 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4110 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4111 * doc/tm.texi: Regenerate.
4112 * gimple-expr.c: Include target.h.
4113 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4114 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4116 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4117 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4118 Use the original predicate if it already has a suitable type.
4120 2020-01-09 Martin Jambor <mjambor@suse.cz>
4122 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4123 resolve_speculation and redirect_call_stmt_to_callee static. Change
4124 return type of set_call_stmt to cgraph_edge *.
4125 * auto-profile.c (afdo_indirect_call): Adjust call to
4126 redirect_call_stmt_to_callee.
4127 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4128 make the this pointer explicit, adjust self-recursive calls and the
4129 call top make_direct. Return the resulting edge.
4130 (cgraph_edge::remove): Make this pointer explicit.
4131 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4132 (cgraph_edge::make_direct): Likewise, adjust call to
4133 resolve_speculation.
4134 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4135 call to set_call_stmt.
4136 (cgraph_update_edges_for_call_stmt_node): Update call to
4137 set_call_stmt and remove.
4138 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4139 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4140 (cgraph_node::create_edge_including_clones): Moved "first" definition
4141 of edge to the block where it was used. Adjusted calls to
4143 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4144 cgraph_edge::remove.
4145 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4146 make_direct and redirect_call_stmt_to_callee.
4147 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4148 resolve_speculation and make_direct.
4149 * ipa-inline-transform.c (inline_transform): Adjust call to
4150 redirect_call_stmt_to_callee.
4151 (check_speculations_1):: Adjust call to resolve_speculation.
4152 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4153 resolve-speculation.
4154 (inline_small_functions): Adjust call to resolve_speculation.
4155 (ipa_inline): Likewise.
4156 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4158 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4159 safe with regards to edge removal, adjust calls to
4160 redirect_call_stmt_to_callee.
4161 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4162 and redirect_call_stmt_to_callee.
4163 * multiple_target.c (create_dispatcher_calls): Adjust call to
4164 redirect_call_stmt_to_callee
4165 (redirect_to_specific_clone): Likewise.
4166 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4167 Adjust calls to cgraph_edge::remove.
4168 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4169 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4170 (expand_call_inline): Adjust call to cgraph_edge::remove.
4172 2020-01-09 Martin Liska <mliska@suse.cz>
4174 * params.opt: Set Optimization for
4175 param_max_speculative_devirt_maydefs.
4177 2020-01-09 Martin Sebor <msebor@redhat.com>
4181 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4183 2020-01-09 Martin Liska <mliska@suse.cz>
4185 * auto-profile.c (auto_profile): Use opt_for_fn
4187 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4188 (propagate_vals_across_arith_jfunc): Likewise.
4189 (hint_time_bonus): Likewise.
4190 (incorporate_penalties): Likewise.
4191 (good_cloning_opportunity_p): Likewise.
4192 (perform_estimation_of_a_value): Likewise.
4193 (estimate_local_effects): Likewise.
4194 (ipcp_propagate_stage): Likewise.
4195 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4196 (set_switch_stmt_execution_predicate): Likewise.
4197 (analyze_function_body): Likewise.
4198 * ipa-inline-analysis.c (offline_size): Likewise.
4199 * ipa-inline.c (early_inliner): Likewise.
4200 * ipa-prop.c (ipa_analyze_node): Likewise.
4201 (ipcp_transform_function): Likewise.
4202 * ipa-sra.c (process_scan_results): Likewise.
4203 (ipa_sra_summarize_function): Likewise.
4204 * params.opt: Rename ipcp-unit-growth to
4205 ipa-cp-unit-growth. Add Optimization for various
4206 IPA-related parameters.
4208 2020-01-09 Richard Biener <rguenther@suse.de>
4211 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4213 2020-01-09 Richard Biener <rguenther@suse.de>
4215 PR tree-optimization/93040
4216 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4218 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4220 * common/config/avr/avr-common.c (avr_option_optimization_table)
4221 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4223 2020-01-09 Martin Liska <mliska@suse.cz>
4225 * cgraphclones.c (symbol_table::materialize_all_clones):
4226 Use cgraph_node::dump_name.
4228 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4231 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4232 output_operand_lossage instead of gcc_unreachable.
4233 * doc/md.texi (riscv f constraint): Fix typo.
4236 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4237 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4238 CONST_SCALAR_INT_P instead of CONST_INT_P.
4239 (*subv<mode>4_1): Rename to ...
4240 (subv<mode>4_1): ... this.
4241 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4242 define_insn_and_split patterns.
4243 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4246 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4248 * vec.c (class selftest::count_dtor): New class.
4249 (selftest::test_auto_delete_vec): New test.
4250 (selftest::vec_c_tests): Call it.
4251 * vec.h (class auto_delete_vec): New class template.
4252 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4254 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4256 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4258 2020-01-08 Jim Wilson <jimw@sifive.com>
4260 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4261 use of TLS_MODEL_LOCAL_EXEC when not pic.
4263 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4265 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4268 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4271 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4272 *stack_protect_set_3 peephole2): Also check that the second
4273 insns source is general_operand.
4276 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4277 predicate for output operand instead of register_operand.
4278 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4279 memory destination and non-memory operands[2].
4281 2020-01-08 Martin Liska <mliska@suse.cz>
4283 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4284 ::dump_asm_name instead of (::name or ::asm_name).
4285 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4286 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4287 (analyze_functions): Likewise.
4288 (expand_all_functions): Likewise.
4289 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4290 (propagate_bits_across_jump_function): Likewise.
4291 (dump_profile_updates): Likewise.
4292 (ipcp_store_bits_results): Likewise.
4293 (ipcp_store_vr_results): Likewise.
4294 * ipa-devirt.c (dump_targets): Likewise.
4295 * ipa-fnsummary.c (analyze_function_body): Likewise.
4296 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4297 (process_hsa_functions): Likewise.
4298 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4299 (set_alias_uids): Likewise.
4300 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4301 * ipa-inline.c (recursive_inlining): Likewise.
4302 (inline_to_all_callers_1): Likewise.
4303 (ipa_inline): Likewise.
4304 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4305 (ipa_propagate_frequency): Likewise.
4306 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4307 (remove_described_reference): Likewise.
4308 * ipa-pure-const.c (worse_state): Likewise.
4309 (check_retval_uses): Likewise.
4310 (analyze_function): Likewise.
4311 (propagate_pure_const): Likewise.
4312 (propagate_nothrow): Likewise.
4313 (dump_malloc_lattice): Likewise.
4314 (propagate_malloc): Likewise.
4315 (pass_local_pure_const::execute): Likewise.
4316 * ipa-visibility.c (optimize_weakref): Likewise.
4317 (function_and_variable_visibility): Likewise.
4318 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4319 (ipa_discover_variable_flags): Likewise.
4320 * lto-streamer-out.c (output_function): Likewise.
4321 (output_constructor): Likewise.
4322 * tree-inline.c (copy_bb): Likewise.
4323 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4324 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4326 2020-01-08 Richard Biener <rguenther@suse.de>
4329 * tree-eh.c (sink_clobbers): Update virtual operands for
4330 the first and last stmt only. Add a dry-run capability.
4331 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4332 after CFG manipulations and in RPO order to catch all
4333 secondary opportunities reliably.
4335 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4338 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4340 2019-01-08 Richard Biener <rguenther@suse.de>
4343 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4344 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4345 virtual operand, also updating SSA use.
4346 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4347 Update stmt after resetting virtual operand.
4348 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4349 * gimple-iterator.c (gsi_remove): When not removing the stmt
4350 permanently do not delink immediate uses or mark the stmt modified.
4352 2020-01-08 Martin Liska <mliska@suse.cz>
4354 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4355 (ipa_call_context::estimate_size_and_time): Likewise.
4356 (inline_analyze_function): Likewise.
4358 2020-01-08 Martin Liska <mliska@suse.cz>
4360 * cgraph.c (cgraph_node::dump): Use systematically
4363 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4365 Add -nodevicespecs option for avr.
4368 * config/avr/avr.opt (-nodevicespecs): New driver option.
4369 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4370 "-specs=device-specs/..." if that option is not set.
4371 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4373 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4375 Implement 64-bit double functions for avr.
4378 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4379 --with-double-comparison.
4380 * doc/install.texi: Document them.
4381 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4382 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4383 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4384 * doc/invoke.texi (AVR Built-in Macros): Document them.
4385 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4386 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4387 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4389 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4392 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4393 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4394 when only building rm-profile multilibs.
4396 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4399 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4400 lattice for a value to check.
4401 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4402 finite propagation in self-recursive scc.
4404 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4406 * ipa-inline.c (caller_growth_limits): Restore the AND.
4408 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4410 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4411 (VEC_ALLREG_ALT): New iterator.
4412 (VEC_ALLREG_INT_MODE): New iterator.
4413 (VCMP_MODE): New iterator.
4414 (VCMP_MODE_INT): New iterator.
4415 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4416 (vec_cmp<u>v64qidi): New define_expand.
4417 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4418 (vec_cmpu<mode>di_exec): New define_expand.
4419 (vec_cmp<u>v64qidi_exec): New define_expand.
4420 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4421 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4422 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4423 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4424 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4425 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4426 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4427 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4428 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4429 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4431 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4432 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4434 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4436 * config/gcn/constraints.md (DA): Update description and match.
4438 (Db): New constraint.
4439 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4441 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4442 Implement 'Db' mixed immediate type.
4443 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4444 (addcv64si3_dup<exec_vcc>): Delete.
4445 (subcv64si3<exec_vcc>): Rework constraints.
4446 (addv64di3): Rework constraints.
4447 (addv64di3_exec): Rework constraints.
4448 (subv64di3): Rework constraints.
4449 (addv64di3_dup): Delete.
4450 (addv64di3_dup_exec): Delete.
4451 (addv64di3_zext): Rework constraints.
4452 (addv64di3_zext_exec): Rework constraints.
4453 (addv64di3_zext_dup): Rework constraints.
4454 (addv64di3_zext_dup_exec): Rework constraints.
4455 (addv64di3_zext_dup2): Rework constraints.
4456 (addv64di3_zext_dup2_exec): Rework constraints.
4457 (addv64di3_sext_dup2): Rework constraints.
4458 (addv64di3_sext_dup2_exec): Rework constraints.
4460 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4462 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4463 existing target checks.
4465 2020-01-07 Richard Biener <rguenther@suse.de>
4467 * doc/install.texi: Bump minimal supported MPC version.
4469 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4471 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4472 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4473 * langhooks.c: Include stor-layout.h.
4474 (lhd_simulate_enum_decl): New function.
4475 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4476 handle_arm_sve_h for the LTO frontend.
4477 (register_vector_type): Cope with null returns from pushdecl.
4479 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4481 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4482 (aarch64_sve::nvectors_if_data_type): Replace with...
4483 (aarch64_sve::builtin_type_p): ...this.
4484 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4485 (find_vector_type): Delete.
4486 (add_sve_type_attribute): New function.
4487 (lookup_sve_type_attribute): Likewise.
4488 (register_builtin_types): Add an "SVE type" attribute to each type.
4489 (register_tuple_type): Likewise.
4490 (svbool_type_p, nvectors_if_data_type): Delete.
4491 (mangle_builtin_type): Use lookup_sve_type_attribute.
4492 (builtin_type_p): Likewise. Add an overload that returns the
4493 number of constituent vector and predicate registers.
4494 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4495 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4496 instead of aarch64_sve_argument_p.
4497 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4498 (aarch64_pass_by_reference): Likewise.
4499 (aarch64_function_value_1): Likewise.
4500 (aarch64_return_in_memory): Likewise.
4501 (aarch64_layout_arg): Likewise.
4503 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4505 PR tree-optimization/93156
4506 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4507 least significant bit is always clear.
4509 PR tree-optimization/93118
4510 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4511 simplifier with two intermediate conversions.
4513 2020-01-07 Martin Liska <mliska@suse.cz>
4515 * params.opt: Add Optimization for various parameters.
4517 2020-01-07 Martin Liska <mliska@suse.cz>
4520 * doc/extend.texi: Explain cloning for target_clone
4523 2020-01-07 Martin Liska <mliska@suse.cz>
4525 PR tree-optimization/92860
4526 * common.opt: Make in Optimization option
4527 as it is affected by -O0, which is an Optimization
4529 * tree-inline.c (tree_inlinable_function_p):
4530 Use opt_for_fn for warn_inline.
4531 (expand_call_inline): Likewise.
4533 2020-01-07 Martin Liska <mliska@suse.cz>
4535 PR tree-optimization/92860
4536 * common.opt: Make flag_ree as optimization
4539 2020-01-07 Martin Liska <mliska@suse.cz>
4541 PR optimization/92860
4542 * params.opt: Mark param_min_crossjump_insns with Optimization
4545 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4547 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4548 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4550 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4552 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4553 helper function to return the valid addressing formats for a given
4554 hard register and mode.
4555 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4557 * config/rs6000/constraints.md (Q constraint): Update
4559 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4562 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4563 Use 'Q' for doing vector extract from memory.
4564 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4566 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4567 doing vector extract from memory.
4568 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4569 extract from memory.
4571 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4572 for the offset being 34-bits when -mcpu=future is used.
4574 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4576 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4577 instead of cmpib_comparison_operator in cmpib patterns.
4578 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4579 of cmpib_comparison_operator. Revise comment.
4581 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4583 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4584 in an IFN_DIV_POW2 node to be equal.
4586 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4588 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4589 (vect_check_scalar_mask): ...this.
4590 (vectorizable_store, vectorizable_load): Update call accordingly.
4591 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4592 argument in calls to conditional internal functions.
4594 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4596 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4597 '0' matching inputs.
4598 (subv64di3_exec): Likewise.
4600 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4602 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4603 * doc/md.texi (movstr): Likewise.
4605 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4607 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4610 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4612 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4614 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4615 to a temporary file and use move-if-change to update the real
4616 file where necessary.
4618 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4620 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4621 rather than Upa for CPY /M.
4623 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4625 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4628 2020-01-06 Martin Liska <mliska@suse.cz>
4630 PR tree-optimization/92860
4631 * params.opt: Mark param_max_combine_insns with Optimization
4634 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4637 * config/i386/i386.md (SWIDWI): New mode iterator.
4638 (DWI, dwi): Add TImode variants.
4639 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4640 <general_hilo_operand> instead of <general_operand>. Use
4641 CONST_SCALAR_INT_P instead of CONST_INT_P.
4642 (*addv<mode>4_1): Rename to ...
4643 (addv<mode>4_1): ... this.
4644 (QWI): New mode attribute.
4645 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4646 define_insn_and_split patterns.
4647 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4649 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4650 <general_hilo_operand> instead of <general_operand>.
4651 (*addcarry<mode>_1): New define_insn.
4652 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4654 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4656 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4657 Use "call" instead of "set".
4659 2020-01-03 Martin Jambor <mjambor@suse.cz>
4662 * ipa-cp.c (print_all_lattices): Skip functions without info.
4664 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4667 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4668 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4669 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4670 for 'e' simd clones.
4673 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4675 (mprefer-vector-width=): Add Save.
4676 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4677 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4678 (ix86_debug_options, ix86_function_specific_print): Adjust
4679 ix86_target_string callers.
4680 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4681 (ix86_valid_target_attribute_tree): Likewise.
4682 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4683 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4684 ix86_target_string caller.
4687 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4688 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4689 instead of gen_int_shift_amount + convert_modes.
4691 PR rtl-optimization/93088
4692 * loop-iv.c (find_single_def_src): Punt after looking through
4693 128 reg copies for regs with single definitions. Move definitions
4696 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4698 * config/arm/arm-c.c (arm_cpu_builtins): Define
4699 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4700 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4701 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4702 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4703 * config/arm/arm-tables.opt: Regenerated.
4704 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4705 arm_arch_i8mm and arm_arch_bf16 when enabled.
4706 * config/arm/arm.h (TARGET_I8MM): New macro.
4707 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4708 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4709 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4710 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4711 (v8_6_a_simd_variants): New.
4712 (v8_*_a_simd_variants): Add i8mm and bf16.
4713 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4715 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4718 * predict.c (compute_function_frequency): Don't call
4719 warn_function_cold on functions that already have cold attribute.
4721 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4724 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4725 COMDAT group function labels in .data.rel.ro.local section.
4726 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4729 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4730 comparison_operator in B and S integer comparisons. Likewise, use
4731 ordered_comparison_operator instead of cmpib_comparison_operator in
4733 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4735 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4737 Update copyright years.
4739 * gcc.c (process_command): Update copyright notice dates.
4740 * gcov-dump.c (print_version): Ditto.
4741 * gcov.c (print_version): Ditto.
4742 * gcov-tool.c (print_version): Ditto.
4743 * gengtype.c (create_file): Ditto.
4744 * doc/cpp.texi: Bump @copying's copyright year.
4745 * doc/cppinternals.texi: Ditto.
4746 * doc/gcc.texi: Ditto.
4747 * doc/gccint.texi: Ditto.
4748 * doc/gcov.texi: Ditto.
4749 * doc/install.texi: Ditto.
4750 * doc/invoke.texi: Ditto.
4752 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4754 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4757 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4759 PR tree-optimization/93098
4760 * match.pd (popcount): For shift amounts, use integer_onep
4761 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4762 tests. Make sure that precision is power of two larger than or equal
4763 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4764 instead of ULL suffixed constants. Formatting fixes.
4766 Copyright (C) 2020 Free Software Foundation, Inc.
4768 Copying and distribution of this file, with or without modification,
4769 are permitted in any medium without royalty provided the copyright
4770 notice and this notice are preserved.