1 2020-02-26 Carl Love <cel@us.ibm.com>
4 * doc/extend.texi (PowerPC AltiVec Built-in Functions available on
5 ISA 2.07): The builtin-function name __builtin_crypto_vpmsumb is only
6 for the vector unsigned short arguments. It is also listed as the
7 name of the built-in for arguments vector unsigned short,
8 vector unsigned int and vector unsigned long long built-ins. The
9 name of the builtins for these arguments should be:
10 __builtin_crypto_vpmsumh, __builtin_crypto_vpmsumw and
11 __builtin_crypto_vpmsumd respectively.
13 2020-02-26 Richard Biener <rguenther@suse.de>
15 * tree-vect-slp.c (vect_print_slp_tree): Also dump ref count
18 2020-02-26 Richard Sandiford <richard.sandiford@arm.com>
21 * optabs-tree.c (supportable_convert_operation): Reject types with
24 2020-02-26 David Malcolm <dmalcolm@redhat.com>
26 * Makefile.in (ANALYZER_OBJS): Add analyzer/bar-chart.o.
28 2020-02-26 Jakub Jelinek <jakub@redhat.com>
30 PR tree-optimization/93820
31 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
32 argument to ALL_INTEGER_CST_P boolean.
33 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
34 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
35 adjacent INTEGER_CST store into merged_store->only_constants like
38 2020-02-25 Jakub Jelinek <jakub@redhat.com>
41 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
43 * cfghooks.c (verify_flow_info): Likewise.
44 * predict.c (combine_predictions_for_bb): Likewise.
45 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
46 sucessor -> successor.
47 (find_traces_1_round): Fix comment typo, destinarion -> destination.
48 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
50 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
51 message typo, sucessors -> successors.
53 2020-02-25 Martin Sebor <msebor@redhat.com>
55 * doc/extend.texi (attribute access): Correct an example.
57 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
59 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
61 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
63 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
64 (VD): Enable for V4BF.
66 (VQ): Enable for V8BF.
69 (VDBL, Vdbl): Add V4BF.
70 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
71 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
72 (bfloat16x8x2_t): Likewise.
73 (bfloat16x4x3_t): Likewise.
74 (bfloat16x8x3_t): Likewise.
75 (bfloat16x4x4_t): Likewise.
76 (bfloat16x8x4_t): Likewise.
78 (vld1_bf16, vld1_bf16_x2): New.
79 (vld1_bf16_x3, vld1_bf16_x4): New.
80 (vld1q_bf16, vld1q_bf16_x2): New.
81 (vld1q_bf16_x3, vld1q_bf16_x4): New.
82 (vld1_lane_bf16): New.
83 (vld1q_lane_bf16): New.
85 (vld1q_dup_bf16): New.
89 (vld2q_dup_bf16): New.
93 (vld3q_dup_bf16): New.
97 (vld4q_dup_bf16): New.
98 (vst1_bf16, vst1_bf16_x2): New.
99 (vst1_bf16_x3, vst1_bf16_x4): New.
100 (vst1q_bf16, vst1q_bf16_x2): New.
101 (vst1q_bf16_x3, vst1q_bf16_x4): New.
102 (vst1_lane_bf16): New.
103 (vst1q_lane_bf16): New.
111 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
113 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
114 (VALL_F16): Likewise.
115 (VALLDI_F16): Likewise.
118 (vswap_width_name): Likewise.
119 (VSWAP_WIDTH): Likewise.
123 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
124 (vget_lane_bf16, vgetq_lane_bf16): New.
126 (vdup_n_bf16, vdupq_n_bf16): New.
127 (vdup_lane_bf16, vdup_laneq_bf16): New.
128 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
129 (vduph_lane_bf16, vduph_laneq_bf16): New.
130 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
131 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
132 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
133 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
134 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
135 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
136 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
137 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
138 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
139 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
140 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
141 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
142 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
143 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
144 (vreinterpretq_bf16_p128): New.
145 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
146 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
147 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
148 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
149 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
150 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
151 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
152 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
153 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
154 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
155 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
156 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
157 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
158 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
159 (vreinterpretq_p128_bf16): New.
161 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
163 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
164 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
165 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
166 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
167 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
168 * config/arm/iterators.md (VSF2BF): New attribute.
169 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
170 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
171 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
173 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
175 * config/arm/arm.md (required_for_purecode): New attribute.
176 (enabled): Handle required_for_purecode.
177 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
178 work with -mpure-code.
180 2020-02-25 Jakub Jelinek <jakub@redhat.com>
182 PR rtl-optimization/93908
183 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
186 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
188 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
190 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
192 * doc/install.texi (--enable-checking): Adjust wording.
194 2020-02-25 Richard Biener <rguenther@suse.de>
196 PR tree-optimization/93868
197 * tree-vect-slp.c (slp_copy_subtree): New function.
198 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
199 re-arranging stmts in it.
201 2020-02-25 Jakub Jelinek <jakub@redhat.com>
204 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
205 dummy function and remove it at the end.
208 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
209 paramter -> parameter.
210 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
211 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
213 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
215 * doc/install.texi (--enable-checking): Properly document current
217 (--enable-stage1-checking): Minor clarification about bootstrap.
219 2020-02-24 David Malcolm <dmalcolm@redhat.com>
222 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
223 -fanalyzer-checker=taint is also required.
224 (-fanalyzer-checker=): Note that providing this option enables the
225 given checker, and doing so may be required for checkers that are
228 2020-02-24 David Malcolm <dmalcolm@redhat.com>
230 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
231 significant control flow events; add a "3" which shows all
232 control flow events; the old "3" becomes "4".
234 2020-02-24 Jakub Jelinek <jakub@redhat.com>
236 PR tree-optimization/93582
237 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
238 pd.offset and pd.size to be counted in bits rather than bytes, add
239 support for maxsizei that is not a multiple of BITS_PER_UNIT and
240 handle bitfield stores and loads.
241 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
242 uncomparable quantities - bytes vs. bits. Allow push_partial_def
243 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
244 pd.offset/pd.size to be counted in bits rather than bytes.
245 Formatting fix. Rename shadowed len variable to buflen.
247 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
248 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
251 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
252 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
253 * opts-common.c (parse_options_from_collect_gcc_options): New function.
254 (prepend_xassembler_to_collect_as_options): Likewise.
255 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
256 (prepend_xassembler_to_collect_as_options): Likewise.
257 * lto-opts.c (lto_write_options): Stream assembler options
258 in COLLECT_AS_OPTIONS.
259 * lto-wrapper.c (xassembler_options_error): New static variable.
260 (get_options_from_collect_gcc_options): Move parsing options code to
261 parse_options_from_collect_gcc_options and call it.
262 (merge_and_complain): Validate -Xassembler options.
263 (append_compiler_options): Handle OPT_Xassembler.
264 (run_gcc): Append command line -Xassembler options to
266 * doc/invoke.texi: Add documentation about using Xassembler
269 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
271 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
273 (riscv_rtx_costs): Update cost model for LTGT.
275 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
277 PR rtl-optimization/93564
278 * ira-color.c (struct update_cost_queue_elem): New member start.
279 (queue_update_cost, get_next_update_cost): Add new arg start.
280 (allocnos_conflict_p): New function.
281 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
282 Add checking conflicts with allocnos_conflict_p.
283 (update_costs_from_prefs, restore_costs_from_copies): Adjust
284 update_costs_from_allocno calls.
285 (update_conflict_hard_regno_costs): Add checking conflicts with
286 allocnos_conflict_p. Adjust calls of queue_update_cost and
287 get_next_update_cost.
288 (assign_hard_reg): Adjust calls of queue_update_cost. Add
290 (bucket_allocno_compare_func): Restore previous version.
292 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
294 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
295 double-word size when handling aggregate return values.
296 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
297 that homogeneous SFmode and DFmode aggregates are passed and returned
298 in general registers.
300 2020-02-21 Jakub Jelinek <jakub@redhat.com>
303 * opts.c (print_filtered_help): Translate help before appending
304 messages to it rather than after that.
306 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
308 PR rtl-optimization/PR92989
309 * lra-lives.c (process_bb_lives): Restore the original order
310 of the bb liveness update. Call make_hard_regno_dead for each
311 register clobbered at the start of an EH receiver.
313 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
316 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
317 self-recursively generated.
319 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
322 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
325 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
327 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
328 Document new target supports option.
330 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
332 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
333 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
334 * config/arm/iterators.md (MATMUL): New iterator.
335 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
336 (mmla_sfx): New attribute.
337 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
338 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
339 (UNSPEC_MATMUL_US): New.
341 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
343 * config/arm/arm.md: Prevent scalar shifts from being used when big
346 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
347 Richard Biener <rguenther@suse.de>
349 PR tree-optimization/93586
350 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
351 after mismatched array refs; do not sure type size information to
352 recover from unmatched referneces with !flag_strict_aliasing_p.
354 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
356 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
357 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
358 (scatter_store<mode>): Rename to ...
359 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
360 (scatter<mode>_exec): Delete. Move contents ...
361 (mask_scatter_store<mode>): ... here, and rename that to ...
362 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
363 Remove mode conversion.
364 (mask_gather_load<mode>): Rename to ...
365 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
366 Remove mode conversion.
367 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
369 2020-02-21 Martin Jambor <mjambor@suse.cz>
371 PR tree-optimization/93845
372 * tree-sra.c (verify_sra_access_forest): Only test access size of
375 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
377 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
378 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
379 (addv64di3_exec): Likewise.
380 (subv64di3): Likewise.
381 (subv64di3_exec): Likewise.
382 (addv64di3_zext): Likewise.
383 (addv64di3_zext_exec): Likewise.
384 (addv64di3_zext_dup): Likewise.
385 (addv64di3_zext_dup_exec): Likewise.
386 (addv64di3_zext_dup2): Likewise.
387 (addv64di3_zext_dup2_exec): Likewise.
388 (addv64di3_sext_dup2): Likewise.
389 (addv64di3_sext_dup2_exec): Likewise.
390 (<expander>v64di3): Likewise.
391 (<expander>v64di3_exec): Likewise.
392 (*<reduc_op>_dpp_shr_v64di): Likewise.
393 (*plus_carry_dpp_shr_v64di): Likewise.
394 * config/gcn/gcn.md (adddi3): Likewise.
395 (addptrdi3): Likewise.
396 (<expander>di3): Likewise.
398 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
400 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
402 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
404 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
405 support. Use aarch64_emit_mult instead of emitting multiplication
406 instructions directly.
407 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
408 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
410 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
412 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
413 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
414 instead of emitting multiplication instructions directly.
415 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
416 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
417 (@aarch64_frecps<mode>): New expanders.
419 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
421 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
422 on and produce uint64_ts rather than ints.
423 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
424 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
426 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
428 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
429 an unused xmsk register when handling approximate rsqrt.
431 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
433 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
434 flag_finite_math_only condition.
436 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
439 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
440 to destination operand for shufps alternative.
441 (*vec_extractv2si_1): Ditto.
443 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
446 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
449 2020-02-20 Martin Liska <mliska@suse.cz>
452 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
454 2020-02-20 Martin Liska <mliska@suse.cz>
457 * common/config/avr/avr-common.c: Remote trailing "|".
459 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
461 * collect2.c (maybe_run_lto_and_relink): Fix typo in
464 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
466 PR tree-optimization/93767
467 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
468 access-size bias from the offset calculations for negative strides.
470 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
472 * collect2.c (c_file, o_file): Make const again.
473 (ldout,lderrout, dump_ld_file): Remove.
474 (tool_cleanup): Avoid calling not signal-safe functions.
475 (maybe_run_lto_and_relink): Avoid possible signal handler
476 access to unintialzed memory (lto_o_files).
477 (main): Avoid leaking temp files in $TMPDIR.
478 Initialize c_file/o_file with concat, which avoids exposing
479 uninitialized memory to signal handler, which calls unlink(!).
480 Avoid calling maybe_unlink when the main function returns,
481 since the atexit handler is already doing this.
482 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
484 2020-02-19 Martin Jambor <mjambor@suse.cz>
486 PR tree-optimization/93776
487 * tree-sra.c (create_access): Do not create zero size accesses.
488 (get_access_for_expr): Do not search for zero sized accesses.
490 2020-02-19 Martin Jambor <mjambor@suse.cz>
492 PR tree-optimization/93667
493 * tree-sra.c (scalarizable_type_p): Return false if record fields
494 do not follow wach other.
496 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
498 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
499 rather than fmv.x.s/fmv.s.x.
501 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
503 * config/aarch64/aarch64-simd-builtins.def
504 (intrinsic_vec_smult_lo_): New.
505 (intrinsic_vec_umult_lo_): Likewise.
506 (vec_widen_smult_hi_): Likewise.
507 (vec_widen_umult_hi_): Likewise.
508 * config/aarch64/aarch64-simd.md
509 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
510 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
511 (vmull_high_s16): Likewise.
512 (vmull_high_s32): Likewise.
513 (vmull_high_u8): Likewise.
514 (vmull_high_u16): Likewise.
515 (vmull_high_u32): Likewise.
516 (vmull_s8): Likewise.
517 (vmull_s16): Likewise.
518 (vmull_s32): Likewise.
519 (vmull_u8): Likewise.
520 (vmull_u16): Likewise.
521 (vmull_u32): Likewise.
523 2020-02-18 Martin Liska <mliska@suse.cz>
525 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
526 bootstrap by missing removal of invalid sanity check.
528 2020-02-18 Martin Liska <mliska@suse.cz>
531 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
532 Always compare LHS of gimple_assign.
534 2020-02-18 Martin Liska <mliska@suse.cz>
537 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
538 and return type of functions.
539 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
540 Drop MALLOC attribute for void functions.
541 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
542 malloc_state for a new VOID clone.
544 2020-02-18 Martin Liska <mliska@suse.cz>
547 * common.opt: Add -fprofile-reproducibility.
548 * doc/invoke.texi: Document it.
549 * value-prof.c (dump_histogram_value):
550 Document and support behavior for counters[0]
551 being a negative value.
552 (get_nth_most_common_value): Handle negative
553 counters[0] in respect to flag_profile_reproducible.
555 2020-02-18 Jakub Jelinek <jakub@redhat.com>
558 * cgraph.c (verify_speculative_call): Use speculative_id instead of
559 speculative_uid in messages. Remove trailing whitespace from error
560 message. Use num_speculative_call_targets instead of
561 num_speculative_targets in a message.
562 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
563 edge messages and stmt instead of cal_stmt in reference message.
565 PR tree-optimization/93780
566 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
567 before calling build_vector_type.
568 (execute_update_addresses_taken): Likewise.
571 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
572 typo, functoin -> function.
573 * tree.c (free_lang_data_in_decl): Fix comment typo,
574 functoin -> function.
575 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
577 2020-02-17 David Malcolm <dmalcolm@redhat.com>
579 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
581 (print_option_information): Don't call get_option_url if URLs
584 2020-02-17 Alexandre Oliva <oliva@adacore.com>
586 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
587 handling of register_common-less targets.
589 2020-02-17 Martin Liska <mliska@suse.cz>
592 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
594 2020-02-17 Martin Liska <mliska@suse.cz>
597 * config/rs6000/rs6000.c (rs6000_option_override_internal):
600 2020-02-17 Martin Liska <mliska@suse.cz>
603 * config/rx/elf.opt: Fix typo.
605 2020-02-17 Richard Biener <rguenther@suse.de>
608 * opts-global.c (print_ignored_options): Use inform and
611 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
614 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
616 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
619 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
620 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
622 2020-02-15 Jason Merrill <jason@redhat.com>
624 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
626 2020-02-15 Jakub Jelinek <jakub@redhat.com>
628 PR tree-optimization/93744
629 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
630 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
631 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
632 sure @2 in the first and @1 in the other patterns has no side-effects.
634 2020-02-15 David Malcolm <dmalcolm@redhat.com>
635 Bernd Edlinger <bernd.edlinger@hotmail.de>
639 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
640 * configure.ac (--with-diagnostics-urls): New configuration
641 option, based on --with-diagnostics-color.
642 (DIAGNOSTICS_URLS_DEFAULT): New define.
643 * config.h: Regenerate.
644 * configure: Regenerate.
645 * diagnostic.c (diagnostic_urls_init): Handle -1 for
646 DIAGNOSTICS_URLS_DEFAULT from configure-time
647 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
648 and TERM_URLS environment variable.
649 * diagnostic-url.h (diagnostic_url_format): New enum type.
650 (diagnostic_urls_enabled_p): rename to...
651 (determine_url_format): ... this, and change return type.
652 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
653 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
654 the linux console, and mingw.
655 (diagnostic_urls_enabled_p): rename to...
656 (determine_url_format): ... this, and adjust.
657 * pretty-print.h (pretty_printer::show_urls): rename to...
658 (pretty_printer::url_format): ... this, and change to enum.
659 * pretty-print.c (pretty_printer::pretty_printer,
660 pp_begin_url, pp_end_url, test_urls): Adjust.
661 * doc/install.texi (--with-diagnostics-urls): Document the new
662 configuration option.
663 (--with-diagnostics-color): Document the existing interaction
664 with GCC_COLORS better.
665 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
666 vindex reference. Update description of defaults based on the above.
667 (-fdiagnostics-color): Update description of how -fdiagnostics-color
668 interacts with GCC_COLORS.
670 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
673 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
674 conjunction with TARGET_GNU_TLS in early return.
676 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
678 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
679 the mode is not wider than UNITS_PER_WORD.
681 2020-02-14 Martin Jambor <mjambor@suse.cz>
683 PR tree-optimization/93516
684 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
685 access of the same type as the parent.
686 (propagate_subaccesses_from_lhs): Likewise.
688 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
691 * config/i386/avx512vbmi2intrin.h
692 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
693 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
694 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
695 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
696 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
697 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
698 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
699 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
700 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
701 of lacking a closing parenthesis.
702 * config/i386/avx512vbmi2vlintrin.h
703 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
704 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
705 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
706 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
707 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
708 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
709 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
710 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
711 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
712 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
713 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
714 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
715 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
716 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
717 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
718 _mm_shldi_epi32, _mm_mask_shldi_epi32,
719 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
720 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
722 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
725 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
726 the target function entry.
728 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
730 * common/config/arc/arc-common.c (arc_option_optimization_table):
731 Disable if-conversion step when optimized for size.
733 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
735 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
736 R12-R15 are always in ARCOMPACT16_REGS register class.
737 * config/arc/arc.opt (mq-class): Deprecate.
738 * config/arc/constraint.md ("q"): Remove dependency on mq-class
740 * doc/invoke.texi (mq-class): Update text.
741 * common/config/arc/arc-common.c (arc_option_optimization_table):
744 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
746 * config/arc/arc.c (arc_insn_cost): New function.
747 (TARGET_INSN_COST): Define.
748 * config/arc/arc.md (cost): New attribute.
749 (add_n): Use arc_nonmemory_operand.
750 (ashlsi3_insn): Likewise, also update constraints.
751 (ashrsi3_insn): Likewise.
753 (add_shift): Likewise.
754 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
756 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
758 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
760 (umulsidi_600): Likewise.
762 2020-02-13 Jakub Jelinek <jakub@redhat.com>
765 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
766 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
767 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
768 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
769 pass __A to the builtin followed by __W instead of __A followed by
771 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
772 _mm512_mask_popcnt_epi64): Likewise.
773 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
774 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
775 _mm256_mask_popcnt_epi64): Likewise.
777 PR tree-optimization/93582
778 * fold-const.h (shift_bytes_in_array_left,
779 shift_bytes_in_array_right): Declare.
780 * fold-const.c (shift_bytes_in_array_left,
781 shift_bytes_in_array_right): New function, moved from
782 gimple-ssa-store-merging.c, no longer static.
783 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
784 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
785 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
786 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
787 shift_bytes_in_array.
788 (verify_shift_bytes_in_array): Rename to ...
789 (verify_shift_bytes_in_array_left): ... this. Use
790 shift_bytes_in_array_left instead of shift_bytes_in_array.
791 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
792 instead of verify_shift_bytes_in_array.
793 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
794 / native_interpret_expr where the store covers all needed bits,
795 punt on PDP-endian, otherwise allow all involved offsets and sizes
796 not to be byte-aligned.
799 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
800 use const_0_to_255_operand predicate instead of immediate_operand.
801 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
802 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
803 vgf2p8affineinvqb_<mode><mask_name>,
804 vgf2p8affineqb_<mode><mask_name>): Drop mode from
805 const_0_to_255_operand predicated operands.
807 2020-02-12 Jeff Law <law@redhat.com>
809 * config/h8300/h8300.md (comparison shortening peepholes): Use
810 a mode iterator to merge the HImode and SImode peepholes.
812 2020-02-12 Jakub Jelinek <jakub@redhat.com>
815 * real.c (is_even): Make static. Function comment fix.
816 (is_halfway_below): Make static, don't assert R is not inf/nan,
817 instead return false for those. Small formatting fixes.
819 2020-02-12 Martin Sebor <msebor@redhat.com>
822 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
823 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
824 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
825 (strlen_check_and_optimize_call): Adjust callee name.
827 2020-02-12 Jeff Law <law@redhat.com>
829 * config/h8300/h8300.md (comparison shortening peepholes): Drop
830 (and (xor)) variant. Combine other two into single peephole.
832 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
834 PR rtl-optimization/93565
835 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
837 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
839 * config/aarch64/aarch64-simd.md
840 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
841 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
842 generating separate ADDV and zero_extend patterns.
843 * config/aarch64/iterators.md (VDQV_E): New iterator.
845 2020-02-12 Jeff Law <law@redhat.com>
847 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
848 expanders, splits, etc.
849 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
850 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
851 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
852 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
853 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
854 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
856 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
858 2020-02-12 Jakub Jelinek <jakub@redhat.com>
861 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
862 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
863 TARGET_AVX512DQ from condition.
864 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
865 instead of <mask_mode512bit_condition> in condition. If
866 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
868 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
871 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
874 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
876 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
878 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
879 where strlen is more legible.
880 (rs6000_builtin_vectorized_libmass): Ditto.
881 (rs6000_print_options_internal): Ditto.
883 2020-02-11 Martin Sebor <msebor@redhat.com>
885 PR tree-optimization/93683
886 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
888 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
890 * config/rs6000/predicates.md (cint34_operand): Rename the
891 -mprefixed-addr option to be -mprefixed.
892 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
893 the -mprefixed-addr option to be -mprefixed.
894 (OTHER_FUTURE_MASKS): Likewise.
895 (POWERPC_MASKS): Likewise.
896 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
897 the -mprefixed-addr option to be -mprefixed. Change error
898 messages to refer to -mprefixed.
899 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
901 (rs6000_legitimate_offset_address_p): Likewise.
902 (rs6000_mode_dependent_address): Likewise.
903 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
904 "-mprefixed" for target attributes and pragmas.
905 (address_to_insn_form): Rename the -mprefixed-addr option to be
907 (rs6000_adjust_insn_length): Likewise.
908 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
909 -mprefixed-addr option to be -mprefixed.
910 (ASM_OUTPUT_OPCODE): Likewise.
911 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
912 -mprefixed-addr option to be -mprefixed.
913 * config/rs6000/rs6000.opt (-mprefixed): Rename the
914 -mprefixed-addr option to be prefixed. Change the option from
915 being undocumented to being documented.
916 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
917 -mprefixed option. Update the -mpcrel documentation to mention
920 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
922 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
923 including FIRST_PSEUDO_REGISTER - 1.
924 * ira-color.c (print_hard_reg_set): Ditto.
926 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
928 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
929 (USTERNOP_QUALIFIERS): New define.
930 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
931 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
932 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
933 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
934 * config/arm/arm_neon.h (vusdot_s32): New.
935 (vusdot_lane_s32): New.
936 (vusdotq_lane_s32): New.
937 (vsudot_lane_s32): New.
938 (vsudotq_lane_s32): New.
939 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
940 * config/arm/iterators.md (DOTPROD_I8MM): New.
941 (sup, opsuffix): Add <us/su>.
942 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
943 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
945 2020-02-11 Richard Biener <rguenther@suse.de>
947 PR tree-optimization/93661
948 PR tree-optimization/93662
949 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
951 * tree-sra.c (get_access_for_expr): Likewise.
953 2020-02-10 Jakub Jelinek <jakub@redhat.com>
956 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
957 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
958 Change condition from TARGET_AVX2 to TARGET_AVX.
960 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
963 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
966 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
968 Try to generate zero-based comparisons.
969 * config/cris/cris.c (cris_reduce_compare): New function.
970 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
971 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
972 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
974 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
977 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
978 in Thumb state and also as a destination in Arm state. Add T16
981 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
983 * md.texi (Define Subst): Match closing paren in example.
985 2020-02-10 Jakub Jelinek <jakub@redhat.com>
989 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
990 arguments of strncmp.
992 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
995 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
996 but different source value.
997 (adjust_callers_for_value_intersection): New function.
998 (gather_edges_for_value): Adjust order of callers to let a
999 non-self-recursive caller be the first element.
1000 (self_recursive_pass_through_p): Add a new parameter "simple", and
1001 check generalized self-recursive pass-through jump function.
1002 (self_recursive_agg_pass_through_p): Likewise.
1003 (find_more_scalar_values_for_callers_subset): Compute value from
1004 pass-through jump function for self-recursive.
1005 (intersect_with_plats): Cleanup previous implementation code for value
1006 itersection with self-recursive call edge.
1007 (intersect_with_agg_replacements): Likewise.
1008 (intersect_aggregates_with_edge): Deduce value from pass-through jump
1009 function for self-recursive call edge. Cleanup previous implementation
1010 code for value intersection with self-recursive call edge.
1011 (decide_whether_version_node): Remove dead callers and adjust order
1012 to let a non-self-recursive caller be the first element.
1014 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
1016 * recog.c: Move pass_split_before_sched2 code in front of
1017 pass_split_before_regstack.
1018 (pass_data_split_before_sched2): Rename pass to split3 from split4.
1019 (pass_data_split_before_regstack): Rename pass to split4 from split3.
1020 (rest_of_handle_split_before_sched2): Remove.
1021 (pass_split_before_sched2::execute): Unconditionally call
1023 (enable_split_before_sched2): New function.
1024 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
1025 (pass_split_before_regstack::gate): Ditto.
1026 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1027 Update name check for renamed split4 pass.
1028 * config/sh/sh.c (register_sh_passes): Update pass insertion
1029 point for renamed split4 pass.
1031 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1033 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1034 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1035 copying them around between host and target.
1037 2020-02-08 Andrew Pinski <apinski@marvell.com>
1040 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1041 STRICT_ALIGNMENT also.
1043 2020-02-08 Jim Wilson <jimw@sifive.com>
1046 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1048 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1049 Jakub Jelinek <jakub@redhat.com>
1052 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1053 xmm16-xmm31 call-used even in 64-bit ms-abi.
1055 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1057 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1058 (simd_ummla, simd_usmmla): Likewise.
1059 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1060 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1061 (vusmmlaq_s32): New.
1063 2020-02-07 Richard Biener <rguenther@suse.de>
1066 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1067 skipping unreachable regions.
1068 (optimize_inline_calls): Skip folding stmts when we didn't
1071 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1074 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1075 Don't return aggregates with only SFmode and DFmode in SSE
1077 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1079 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1082 * config/rs6000/rs6000-logue.c
1083 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1084 if it fails, move rs into end_addr and retry. Add
1085 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1086 the insn pattern doesn't describe well what exactly happens to
1090 * config/i386/predicates.md (avx_identity_operand): Remove.
1091 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1092 (avx_<castmode><avxsizesuffix>_<castmode>,
1093 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1094 a VEC_CONCAT of the operand and UNSPEC_CAST.
1095 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1096 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1100 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1101 recog_data.insn if distance_non_agu_define changed it.
1103 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1106 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1107 we only had X-FORM (reg+reg) addressing for vectors. Also before
1108 ISA 3.0, we only had X-FORM addressing for scalars in the
1109 traditional Altivec registers.
1111 2020-02-06 <zhongyunde@huawei.com>
1112 Vladimir Makarov <vmakarov@redhat.com>
1114 PR rtl-optimization/93561
1115 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1116 hard register range.
1118 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1120 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1123 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1125 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1126 where the low and the high 32 bits are equal to each other specially,
1127 with an rldimi instruction.
1129 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1131 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1133 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1135 * config/arm/arm-tables.opt: Regenerate.
1137 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1140 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1141 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1142 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1144 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1146 PR rtl-optimization/87763
1147 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1149 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1151 * config/aarch64/aarch64-simd-builtins.def
1152 (bfmlaq): New built-in function.
1153 (bfmlalb): New built-in function.
1154 (bfmlalt): New built-in function.
1155 (bfmlalb_lane): New built-in function.
1156 (bfmlalt_lane): New built-in function.
1157 * config/aarch64/aarch64-simd.md
1158 (aarch64_bfmmlaqv4sf): New pattern.
1159 (aarch64_bfmlal<bt>v4sf): New pattern.
1160 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1161 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1162 (vbfmlalbq_f32): New intrinsic.
1163 (vbfmlaltq_f32): New intrinsic.
1164 (vbfmlalbq_lane_f32): New intrinsic.
1165 (vbfmlaltq_lane_f32): New intrinsic.
1166 (vbfmlalbq_laneq_f32): New intrinsic.
1167 (vbfmlaltq_laneq_f32): New intrinsic.
1168 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1169 (bt): New int attribute.
1171 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1173 * config/i386/i386.md (*pushtf): Emit "#" instead of
1174 calling gcc_unreachable in insn output.
1177 (*pushsf_rex64): Ditto for alternatives other than 1.
1178 (*pushsf): Ditto for alternatives other than 1.
1180 2020-02-06 Martin Liska <mliska@suse.cz>
1182 PR gcov-profile/91971
1183 PR gcov-profile/93466
1184 * coverage.c (coverage_init): Revert mangling of
1185 path into filename. It can lead to huge filename length.
1186 Creation of subfolders seem more natural.
1188 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1191 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1192 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1193 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1195 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1198 * config/i386/predicates.md (avx_identity_operand): New predicate.
1199 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1200 define_insn_and_split.
1203 * omp-low.c (use_pointer_for_field): For nested constructs, also
1204 look for map clauses on target construct.
1205 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1206 taskreg_nesting_level.
1209 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1210 shared clause, call omp_notice_variable on outer context if any.
1212 2020-02-05 Jason Merrill <jason@redhat.com>
1215 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1216 non-zero address even if weak and not yet defined.
1218 2020-02-05 Martin Sebor <msebor@redhat.com>
1220 PR tree-optimization/92765
1221 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1222 * tree-ssa-strlen.c (compute_string_length): Remove.
1223 (determine_min_objsize): Remove.
1224 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1225 Avoid using type size as the upper bound on string length.
1226 (handle_builtin_string_cmp): Add an argument. Adjust.
1227 (strlen_check_and_optimize_call): Pass additional argument to
1228 handle_builtin_string_cmp.
1230 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1232 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1233 (*pushdi2_rex64 peephole2): Unconditionally split after
1235 (*ashl<mode>3_doubleword): Ditto.
1236 (*<shift_insn><mode>3_doubleword): Ditto.
1238 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1241 * config/rs6000/rs6000.c (get_vector_offset): Fix
1243 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1245 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1247 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1250 (Special Functions for Debugging the Analyzer): Update description
1251 of __analyzer_dump_exploded_nodes.
1253 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1256 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1257 include sets and not clobbers in the vzeroupper pattern.
1258 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1259 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1260 (*avx_vzeroupper_1): New define_insn_and_split.
1263 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1264 don't run when !optimize.
1265 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1268 2020-02-05 Richard Biener <rguenther@suse.de>
1271 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1272 checks before matching calls.
1274 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1276 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1277 function comment typo.
1280 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1281 simd_clone_create failed when i == 0, adjust clone->nargs by
1284 2020-02-05 Martin Liska <mliska@suse.cz>
1287 * doc/invoke.texi: Document that one should
1288 not combine ASLR and -fpch.
1290 2020-02-04 Richard Biener <rguenther@suse.de>
1292 PR tree-optimization/93538
1293 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1295 2020-02-04 Richard Biener <rguenther@suse.de>
1297 PR tree-optimization/91123
1298 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1299 (vn_walk_cb_data::last_vuse): New member.
1300 (vn_walk_cb_data::saved_operands): Likewsie.
1301 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1302 (vn_walk_cb_data::push_partial_def): Use finish.
1303 (vn_reference_lookup_2): Update last_vuse and use finish if
1304 we've saved operands.
1305 (vn_reference_lookup_3): Use finish and update calls to
1306 push_partial_defs everywhere. When translating through
1307 memcpy or aggregate copies save off operands and alias-set.
1308 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1309 operation for redundant store removal.
1311 2020-02-04 Richard Biener <rguenther@suse.de>
1313 PR tree-optimization/92819
1314 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1315 generating more stmts than before.
1317 2020-02-04 Martin Liska <mliska@suse.cz>
1319 * config/arm/arm.c (arm_gen_far_branch): Move the function
1320 outside of selftests.
1322 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1324 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1325 function to adjust PC-relative vector addresses.
1326 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1327 handle vectors with PC-relative addresses.
1329 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1331 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1333 (hard_reg_and_mode_to_addr_mask): Delete.
1334 (rs6000_adjust_vec_address): If the original vector address
1335 was REG+REG or REG+OFFSET and the element is not zero, do the add
1336 of the elements in the original address before adding the offset
1337 for the vector element. Use address_to_insn_form to validate the
1338 address using the register being loaded, rather than guessing
1339 whether the address is a DS-FORM or DQ-FORM address.
1341 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1343 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1344 to calculate the offset in memory from the start of a vector of a
1345 particular element. Add code to keep the element number in
1346 bounds if the element number is variable.
1347 (rs6000_adjust_vec_address): Move calculation of offset of the
1348 vector element to get_vector_offset.
1349 (rs6000_split_vec_extract_var): Do not do the initial AND of
1350 element here, move the code to get_vector_offset.
1352 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1354 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1357 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1359 * config/rs6000/constraints.md: Improve documentation.
1361 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1364 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1365 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1367 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1369 * config.gcc: Remove "carrizo" support.
1370 * config/gcn/gcn-opts.h (processor_type): Likewise.
1371 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1372 * config/gcn/gcn.opt (gpu_type): Likewise.
1373 * config/gcn/t-omp-device: Likewise.
1375 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1378 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1379 * config/arm/arm.c (arm_gen_far_branch): New function
1381 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1383 2020-02-03 Julian Brown <julian@codesourcery.com>
1384 Tobias Burnus <tobias@codesourcery.com>
1386 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1388 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1391 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1392 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1395 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1397 PR rtl-optimization/91333
1398 * ira-color.c (struct allocno_color_data): Add member
1400 (init_allocno_threads): Set the member up.
1401 (bucket_allocno_compare_func): Add compare hard reg
1404 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1406 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1408 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1409 * config.in: Regenerated.
1410 * configure: Regenerated.
1411 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1412 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1413 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1415 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1417 * configure: Regenerate.
1419 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1421 PR rtl-optimization/91333
1422 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1423 reg preferences comparison up.
1425 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1427 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1428 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1429 aarch64-sve-builtins-base.h.
1430 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1431 aarch64-sve-builtins-base.cc.
1432 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1433 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1435 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1436 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1437 (svcvtnt): New functions.
1438 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1439 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1440 (svcvtnt): New functions.
1441 (svcvt): Add a form that converts f32 to bf16.
1442 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1443 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1445 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1446 Treat B as bfloat16_t.
1447 (ternary_bfloat_lane_base): New class.
1448 (ternary_bfloat_def): Likewise.
1449 (ternary_bfloat): New shape.
1450 (ternary_bfloat_lane_def): New class.
1451 (ternary_bfloat_lane): New shape.
1452 (ternary_bfloat_lanex2_def): New class.
1453 (ternary_bfloat_lanex2): New shape.
1454 (ternary_bfloat_opt_n_def): New class.
1455 (ternary_bfloat_opt_n): New shape.
1456 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1457 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1458 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1459 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1460 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1461 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1462 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1463 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1464 the pattern off the narrow mode instead of the wider one.
1465 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1466 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1467 (sve_fp_op): Handle them.
1468 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1469 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1471 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1473 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1474 * config/aarch64/aarch64-modes.def (BF): Move definition before
1475 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1476 (SVE_MODES): Handle BF modes.
1477 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1479 (aarch64_full_sve_mode): Likewise.
1480 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1482 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1483 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1484 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1485 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1487 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1489 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1490 (TYPES_all_data): Add bf16.
1491 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1492 (register_tuple_type): Increase buffer size.
1493 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1494 (bf16): New type suffix.
1495 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1496 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1497 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1498 Change type from all_data to all_arith.
1499 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1502 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1503 Matthew Malcomson <matthew.malcomson@arm.com>
1504 Richard Sandiford <richard.sandiford@arm.com>
1506 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1507 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1508 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1509 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1510 __ARM_FEATURE_MATMUL_FP64.
1511 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1512 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1513 be disabled at the same time.
1514 (f32mm): New extension.
1515 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1516 (AARCH64_FL_F64MM): Bump to the next bit up.
1517 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1518 (TARGET_SVE_F64MM): New macros.
1519 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1520 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1521 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1522 (UNSPEC_ZIP2Q): New unspeccs.
1523 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1524 (optab, sur, perm_insn): Handle the new unspecs.
1525 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1526 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1527 TARGET_SVE_F64MM instead of separate tests.
1528 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1529 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1530 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1531 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1532 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1533 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1534 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1535 (TYPES_s_signed): New macro.
1536 (TYPES_s_integer): Use it.
1537 (TYPES_d_float): New macro.
1538 (TYPES_d_data): Use it.
1539 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1540 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1541 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1542 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1543 (svmmla): New shape.
1544 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1545 template parameters.
1546 (ternary_resize2_lane_base): Likewise.
1547 (ternary_resize2_base): New class.
1548 (ternary_qq_lane_base): Likewise.
1549 (ternary_intq_uintq_lane_def): Likewise.
1550 (ternary_intq_uintq_lane): New shape.
1551 (ternary_intq_uintq_opt_n_def): New class
1552 (ternary_intq_uintq_opt_n): New shape.
1553 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1554 (ternary_uintq_intq_def): New class.
1555 (ternary_uintq_intq): New shape.
1556 (ternary_uintq_intq_lane_def): New class.
1557 (ternary_uintq_intq_lane): New shape.
1558 (ternary_uintq_intq_opt_n_def): New class.
1559 (ternary_uintq_intq_opt_n): New shape.
1560 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1561 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1562 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1563 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1565 (svdotprod_lane_impl): ...this new class.
1566 (svmmla_impl, svusdot_impl): New classes.
1567 (svdot_lane): Update to use svdotprod_lane_impl.
1568 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1569 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1571 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1572 function, with no types defined.
1573 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1574 AARCH64_FL_I8MM functions.
1575 (svmmla): New AARCH64_FL_F32MM function.
1576 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1577 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1578 AARCH64_FL_F64MM function.
1579 (REQUIRED_EXTENSIONS):
1581 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1583 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1586 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1588 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1589 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1590 (*movti_internal): Do not check for
1591 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1592 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1593 just after check for TARGET_AVX.
1594 (*movdf_internal): Ditto.
1595 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1596 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1597 * config/i386/sse.md (mov<mode>_internal): Only check
1598 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1599 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1600 (<sse>_andnot<mode>3<mask_name>): Move check for
1601 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1602 (<code><mode>3<mask_name>): Ditto.
1603 (*andnot<mode>3): Ditto.
1604 (*andnottf3): Ditto.
1605 (*<code><mode>3): Ditto.
1606 (*<code>tf3): Ditto.
1607 (*andnot<VI:mode>3): Remove
1608 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1609 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1610 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1611 (sse4_1_blendv<ssemodesuffix>): Ditto.
1612 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1613 Explain that tune applies to 128bit instructions only.
1615 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1617 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1618 to definition of hsa_kernel_description. Parse assembly to find SGPR
1619 and VGPR count of kernel and store in hsa_kernel_description.
1621 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1623 PR rtl-optimization/91838
1624 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1625 to truncate if allowed or reject combination.
1627 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1629 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1630 (find_inv_vars_cb): Likewise.
1632 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1634 * calls.c (special_function_p): Split out the check for DECL_NAME
1635 being non-NULL and fndecl being extern at file scope into a
1636 new maybe_special_function_p and call it. Drop check for fndecl
1637 being non-NULL that was after a usage of DECL_NAME (fndecl).
1638 * tree.h (maybe_special_function_p): New inline function.
1640 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1642 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1643 (mask_gather_load<mode>): ... here, and zero-initialize the
1645 (maskload<mode>di): Zero-initialize the destination.
1648 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1651 * doc/analyzer.texi (Limitations): Note that constraints on
1652 floating-point values are currently ignored.
1654 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1657 * symtab.c (symtab_node::noninterposable_alias): If localalias
1658 already exists, but is not usable, append numbers after it until
1659 a unique name is found. Formatting fix.
1662 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1665 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1667 * config/gcn/gcn.c (print_operand): Handle LTGT.
1668 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1670 2020-01-30 Richard Biener <rguenther@suse.de>
1672 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1673 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1675 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1677 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1678 without a DECL in .data.rel.ro.local.
1680 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1683 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1687 * config/i386/sse.md
1688 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1689 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1690 any_extend code iterator instead of always zero_extend.
1691 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1692 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1693 Use any_extend code iterator instead of always zero_extend.
1694 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1695 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1696 Use any_extend code iterator instead of always zero_extend.
1697 (*sse2_pmovmskb_ext): New define_insn.
1698 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1701 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1702 (*popcountsi2_zext_falsedep): New define_insn.
1704 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1706 * config.in: Regenerated.
1707 * configure: Regenerated.
1709 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1712 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1713 LLVM's assembler changed the default in version 9.
1715 2020-01-24 Jeff Law <law@redhat.com>
1717 PR tree-optimization/89689
1718 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1720 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1724 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1726 PR rtl-optimization/87763
1727 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1728 simplification to handle subregs as well as bare regs.
1729 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1731 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1734 * ira.c (ira): Revert use of simplified LRA algorithm.
1736 2020-01-29 Martin Jambor <mjambor@suse.cz>
1738 PR tree-optimization/92706
1739 * tree-sra.c (struct access): Fields first_link, last_link,
1740 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1741 next_rhs_queued and grp_rhs_queued respectively, new fields
1742 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1743 (struct assign_link): Field next renamed to next_rhs, new field
1744 next_lhs. Updated comment.
1745 (work_queue_head): Renamed to rhs_work_queue_head.
1746 (lhs_work_queue_head): New variable.
1747 (add_link_to_lhs): New function.
1748 (relink_to_new_repr): Also relink LHS lists.
1749 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1750 (add_access_to_lhs_work_queue): New function.
1751 (pop_access_from_work_queue): Renamed to
1752 pop_access_from_rhs_work_queue.
1753 (pop_access_from_lhs_work_queue): New function.
1754 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1756 (child_would_conflict_in_lacc): Renamed to
1757 child_would_conflict_in_acc. Adjusted parameter names.
1758 (create_artificial_child_access): New parameter set_grp_read, use it.
1759 (subtree_mark_written_and_enqueue): Renamed to
1760 subtree_mark_written_and_rhs_enqueue.
1761 (propagate_subaccesses_across_link): Renamed to
1762 propagate_subaccesses_from_rhs.
1763 (propagate_subaccesses_from_lhs): New function.
1764 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1767 2020-01-29 Martin Jambor <mjambor@suse.cz>
1769 PR tree-optimization/92706
1770 * tree-sra.c (struct access): Adjust comment of
1771 grp_total_scalarization.
1772 (find_access_in_subtree): Look for single children spanning an entire
1774 (scalarizable_type_p): Allow register accesses, adjust callers.
1775 (completely_scalarize): Remove function.
1776 (scalarize_elem): Likewise.
1777 (create_total_scalarization_access): Likewise.
1778 (sort_and_splice_var_accesses): Do not track total scalarization
1780 (analyze_access_subtree): New parameter totally, adjust to new meaning
1781 of grp_total_scalarization.
1782 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1783 (can_totally_scalarize_forest_p): New function.
1784 (create_total_scalarization_access): Likewise.
1785 (create_total_access_and_reshape): Likewise.
1786 (total_should_skip_creating_access): Likewise.
1787 (totally_scalarize_subtree): Likewise.
1788 (analyze_all_variable_accesses): Perform total scalarization after
1789 subaccess propagation using the new functions above.
1790 (initialize_constant_pool_replacements): Output initializers by
1791 traversing the access tree.
1793 2020-01-29 Martin Jambor <mjambor@suse.cz>
1795 * tree-sra.c (verify_sra_access_forest): New function.
1796 (verify_all_sra_access_forests): Likewise.
1797 (create_artificial_child_access): Set parent.
1798 (analyze_all_variable_accesses): Call the verifier.
1800 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1802 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1803 if called on indirect edge.
1804 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1805 speculative call if needed.
1807 2020-01-29 Richard Biener <rguenther@suse.de>
1809 PR tree-optimization/93428
1810 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1811 permutation when the load node is created.
1812 (vect_analyze_slp_instance): Re-use it here.
1814 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1816 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1818 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1820 PR rtl-optimization/93272
1821 * ira-lives.c (process_out_of_region_eh_regs): New function.
1822 (process_bb_node_lives): Call it.
1824 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1826 * coverage.c (read_counts_file): Make error message lowercase.
1828 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1830 * profile-count.c (profile_quality_display_names): Fix ordering.
1832 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1835 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1836 hash only when edge is first within the sequence.
1837 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1838 (symbol_table::create_edge): Do not set target_prob.
1839 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1841 (cgraph_edge::make_speculative): Drop target_prob parameter.
1842 (cgraph_edge::speculative_call_info): Remove.
1843 (cgraph_edge::first_speculative_call_target): New member function.
1844 (update_call_stmt_hash_for_removing_direct_edge): New function.
1845 (cgraph_edge::resolve_speculation): Rewrite to new API.
1846 (cgraph_edge::speculative_call_for_target): New member function.
1847 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1848 multiple speculation targets.
1849 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1851 (verify_speculative_call): Verify that targets form an interval.
1852 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1853 (cgraph_edge::first_speculative_call_target): New member function.
1854 (cgraph_edge::next_speculative_call_target): New member function.
1855 (cgraph_edge::speculative_call_target_ref): New member function.
1856 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1857 (cgraph_edge): Remove target_prob.
1858 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1859 Fix handling of speculative calls.
1860 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1861 * ipa-fnsummary.c (analyze_function_body): Likewise.
1862 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1863 * ipa-profile.c (dump_histogram): Fix formating.
1864 (ipa_profile_generate_summary): Watch for overflows.
1865 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1866 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1867 (update_indirect_edges_after_inlining): Update to new API.
1868 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1870 * profile-count.h: (profile_probability::adjusted): New.
1871 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1872 updating of profile.
1873 * value-prof.c (gimple_ic_transform): Rename to ...
1874 (dump_ic_profile): ... this one; update dumping.
1875 (stream_in_histogram_value): Fix formating.
1876 (gimple_value_profile_transformations): Update.
1878 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1881 * config/i386/i386.md (*movoi_internal_avx): Remove
1882 TARGET_SSE_TYPELESS_STORES check.
1883 (*movti_internal): Prefer TARGET_AVX over
1884 TARGET_SSE_TYPELESS_STORES.
1885 (*movtf_internal): Likewise.
1886 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1887 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1888 from TARGET_SSE_TYPELESS_STORES.
1890 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1892 * diagnostic-core.h (warning_at): Rename overload to...
1893 (warning_meta): ...this.
1894 (emit_diagnostic_valist): Delete decl of overload taking
1895 diagnostic_metadata.
1896 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1897 (warning_at): Rename overload taking diagnostic_metadata to...
1898 (warning_meta): ...this.
1900 2020-01-28 Richard Biener <rguenther@suse.de>
1902 PR tree-optimization/93439
1903 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1904 * tree-cfg.c (move_sese_region_to_fn): ... here.
1905 (verify_types_in_gimple_reference): Verify used cliques are
1908 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1911 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1912 argument of a pointer to struct gcc_options and pass it to
1913 parse_mtune_ctrl_str.
1914 (ix86_function_specific_restore): Pass opts to
1915 set_ix86_tune_features.
1916 (ix86_option_override_internal): Likewise.
1917 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1918 gcc_options and use it for x_ix86_tune_ctrl_string.
1920 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1922 PR rtl-optimization/87763
1923 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1924 simplification to handle subregs as well as bare regs.
1925 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1927 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1929 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1930 for reduction chains that (now) include a call.
1932 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1934 PR tree-optimization/92822
1935 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1936 out the don't-care elements of a vector whose significant elements
1937 are duplicates, make the don't-care elements duplicates too.
1939 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1941 PR tree-optimization/93434
1942 * tree-predcom.c (split_data_refs_to_components): Record which
1943 components have had aliasing loads removed. Prevent store-store
1944 commoning for all such components.
1946 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1949 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1950 -1 or is_vshift is true, use new_vector with number of elts npatterns
1951 rather than new_unary_operation.
1953 PR tree-optimization/93454
1954 * gimple-fold.c (fold_array_ctor_reference): Perform
1955 elt_size.to_uhwi () just once, instead of calling it in every
1956 iteration. Punt if that value is above size of the temporary
1957 buffer. Decrease third native_encode_expr argument when
1958 bufoff + elt_sz is above size of buf.
1960 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1962 * config/mips/mips.c (mips_declare_object_name)
1963 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1965 2020-01-27 Martin Liska <mliska@suse.cz>
1967 PR gcov-profile/93403
1968 * tree-profile.c (gimple_init_gcov_profiler): Generate
1969 both __gcov_indirect_call_profiler_v4 and
1970 __gcov_indirect_call_profiler_v4_atomic.
1972 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1975 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1977 (@aarch64_split_simd_mov<mode>): Use it.
1978 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1979 Leave the vec_extract patterns to handle 2-element vectors.
1980 (aarch64_simd_mov_from_<mode>high): Likewise.
1981 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1982 (vec_extractv2dfv1df): Likewise.
1984 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1986 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1987 jump conditions for *compare_condjump<GPI:mode>.
1989 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1992 * digraph.cc (test_edge::test_edge): Specify template for base
1995 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1997 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1999 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2001 * config/arc/arc-protos.h (gen_mlo): Remove.
2002 (gen_mhi): Likewise.
2003 * config/arc/arc.c (AUX_MULHI): Define.
2004 (arc_must_save_reister): Special handling for r58/59.
2005 (arc_compute_frame_size): Consider mlo/mhi registers.
2006 (arc_save_callee_saves): Emit fp/sp move only when emit_move
2008 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
2009 mlo/mhi name selection.
2010 (arc_restore_callee_saves): Don't early restore blink when ISR.
2011 (arc_expand_prologue): Add mlo/mhi saving.
2012 (arc_expand_epilogue): Add mlo/mhi restoring.
2015 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
2016 numbering when MUL64 option is used.
2017 (DWARF2_FRAME_REG_OUT): Define.
2018 * config/arc/arc.md (arc600_stall): New pattern.
2019 (VUNSPEC_ARC_ARC600_STALL): Define.
2020 (mulsi64): Use correct mlo/mhi registers.
2021 (mulsi_600): Clean it up.
2022 * config/arc/predicates.md (mlo_operand): Remove any dependency on
2024 (mhi_operand): Likewise.
2026 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2027 Petro Karashchenko <petro.karashchenko@ring.com>
2029 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2030 attributes if needed.
2031 (prepare_move_operands): Generate special unspec instruction for
2033 (arc_isuncached_mem_p): Propagate uncached attribute to each
2035 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2036 (VUNSPEC_ARC_STDI): Likewise.
2037 (ALLI): New mode iterator.
2038 (mALLI): New mode attribute.
2039 (lddi): New instruction pattern.
2041 (stdidi_split): Split instruction for architectures which are not
2042 supporting ll64 option.
2043 (lddidi_split): Likewise.
2045 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2047 PR rtl-optimization/92989
2048 * lra-lives.c (process_bb_lives): Update the live-in set before
2049 processing additional clobbers.
2051 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2053 PR rtl-optimization/93170
2054 * cselib.c (cselib_invalidate_regno_val): New function, split out
2056 (cselib_invalidate_regno): ...here.
2057 (cselib_invalidated_by_call_p): New function.
2058 (cselib_process_insn): Iterate over all the hard-register entries in
2059 REG_VALUES and invalidate any that cross call-clobbered registers.
2061 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2063 * dojump.c (split_comparison): Use HONOR_NANS rather than
2064 HONOR_SNANS when splitting LTGT.
2066 2020-01-27 Martin Liska <mliska@suse.cz>
2069 * opts.c (print_filtered_help): Exclude language-specific
2070 options from --help=common unless enabled in all FEs.
2072 2020-01-27 Martin Liska <mliska@suse.cz>
2074 * opts.c (print_help): Exclude params from
2075 all except --help=param.
2077 2020-01-27 Martin Liska <mliska@suse.cz>
2080 * config/i386/i386-features.c (make_resolver_func):
2081 Align the code with ppc64 target implementation.
2082 Do not generate a unique name for resolver function.
2084 2020-01-27 Richard Biener <rguenther@suse.de>
2086 PR tree-optimization/93397
2087 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2088 converted reduction chain SLP graph adjustment.
2090 2020-01-26 Marek Polacek <polacek@redhat.com>
2093 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2096 2020-01-26 Jason Merrill <jason@redhat.com>
2099 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2102 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2104 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2105 (rx_setmem): Likewise.
2107 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2110 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2111 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2112 drop <di> from constraint of last operand.
2115 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2116 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2117 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2119 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2122 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2125 2020-01-24 Jeff Law <law@redhat.com>
2127 PR tree-optimization/92788
2128 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2131 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2134 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2135 *avx_vperm_broadcast_<mode>,
2136 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2137 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2138 Move before avx2_perm<mode>/avx512f_perm<mode>.
2141 * simplify-rtx.c (simplify_const_unary_operation,
2142 simplify_const_binary_operation): Punt for mode precision above
2143 MAX_BITSIZE_MODE_ANY_INT.
2145 2020-01-24 Andrew Pinski <apinski@marvell.com>
2147 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2150 2020-01-24 Jeff Law <law@redhat.com>
2153 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2154 for REGs. Call output_operand_lossage to get more reasonable
2157 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2159 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2160 gcn_fp_compare_operator.
2161 (vec_cmpu<mode>di): Use gcn_compare_operator.
2162 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2163 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2164 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2165 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2166 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2167 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2168 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2169 gcn_fp_compare_operator.
2170 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2171 gcn_fp_compare_operator.
2172 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2173 gcn_fp_compare_operator.
2174 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2175 gcn_fp_compare_operator.
2177 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2179 * doc/install.texi (Cross-Compiler-Specific Options): Document
2180 `--with-toolexeclibdir' option.
2182 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2184 * target.def (flags_regnum): Also mention effect on delay slot filling.
2185 * doc/tm.texi: Regenerate.
2187 2020-01-23 Jeff Law <law@redhat.com>
2189 PR translation/90162
2190 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2192 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2195 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2198 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2200 PR rtl-optimization/93402
2201 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2204 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2206 * config.in: Regenerated.
2207 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2208 for TARGET_LIBC_GNUSTACK.
2209 * configure: Regenerated.
2210 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2211 found to be 2.31 or greater.
2213 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2215 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2217 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2218 (mips_asm_file_end): New function. Delegate to
2219 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2220 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2222 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2225 * config/i386/i386-modes.def (POImode): New mode.
2226 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2227 * config/i386/i386.md (DPWI): New mode attribute.
2228 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2230 (QPWI): ... this. Use POI instead of OI for TImode.
2231 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2232 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2235 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2238 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2240 (speculation_tracker_rev): New pattern.
2241 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2242 Use speculation_tracker_rev to track the inverse condition.
2244 2020-01-23 Richard Biener <rguenther@suse.de>
2246 PR tree-optimization/93381
2247 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2248 alias-set of the def as argument and record the first one.
2249 (vn_walk_cb_data::first_set): New member.
2250 (vn_reference_lookup_3): Pass the alias-set of the current def
2251 to push_partial_def. Fix alias-set used in the aggregate copy
2253 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2254 * real.c (clear_significand_below): Fix out-of-bound access.
2256 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2259 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2260 New define_insn patterns.
2262 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2264 * doc/sourcebuild.texi (check-function-bodies): Add an
2265 optional target/xfail selector.
2267 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2269 PR rtl-optimization/93124
2270 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2271 bare USE and CLOBBER insns.
2273 2020-01-22 Andrew Pinski <apinski@marvell.com>
2275 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2277 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2280 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2281 diagnostic_manager into "ana" namespace.
2282 * selftest-run-tests.c (selftest::run_tests): Update for move of
2283 selftest::run_analyzer_selftests to
2284 ana::selftest::run_analyzer_selftests.
2286 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2288 * cfgexpand.c (union_stack_vars): Update the size.
2290 2020-01-22 Richard Biener <rguenther@suse.de>
2292 PR tree-optimization/93381
2293 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2294 throughout, handle all conversions the same.
2296 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2299 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2300 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2301 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2302 Call force_reg on high_in2 unconditionally.
2304 2020-01-22 Martin Liska <mliska@suse.cz>
2306 PR tree-optimization/92924
2307 * profile.c (compute_value_histograms): Divide
2310 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2313 * output.h (assemble_name_resolve): Declare.
2314 * varasm.c (assemble_name_resolve): New function.
2315 (assemble_name): Use it.
2316 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2318 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2320 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2321 update_web_docs_git instead of update_web_docs_svn.
2323 2020-01-21 Andrew Pinski <apinski@marvell.com>
2326 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2327 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2328 (*tlsgd_small_<mode>): Likewise.
2329 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2330 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2331 register. Convert that register back to dest using convert_mode.
2333 2020-01-21 Jim Wilson <jimw@sifive.com>
2335 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2338 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2339 Uros Bizjak <ubizjak@gmail.com>
2342 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2344 (legitimize_tls_address): Do GNU2 TLS address computation in
2345 ptr_mode and zero-extend result to Pmode.
2346 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2347 :P with :PTR and Pmode with ptr_mode.
2348 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2349 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2350 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2352 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2355 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2356 the last two operands are CONST_INT_P before using them as such.
2358 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2360 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2361 to get the integer element types.
2363 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2365 * config/aarch64/aarch64-sve-builtins.h
2366 (function_expander::convert_to_pmode): Declare.
2367 * config/aarch64/aarch64-sve-builtins.cc
2368 (function_expander::convert_to_pmode): New function.
2369 (function_expander::get_contiguous_base): Use it.
2370 (function_expander::prepare_gather_address_operands): Likewise.
2371 * config/aarch64/aarch64-sve-builtins-sve2.cc
2372 (svwhilerw_svwhilewr_impl::expand): Likewise.
2374 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2377 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2378 cfun->machine->label_is_assembled.
2379 (aarch64_print_patchable_function_entry): New.
2380 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2381 * config/aarch64/aarch64.h (struct machine_function): New field,
2384 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2387 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2390 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2393 * cgraph.c (cgraph_edge::resolve_speculation,
2394 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2395 call_stmt_site_hash.
2397 2020-01-21 Martin Liska <mliska@suse.cz>
2399 * config/rs6000/rs6000.c (common_mode_defined): Remove
2402 2020-01-21 Richard Biener <rguenther@suse.de>
2404 PR tree-optimization/92328
2405 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2406 type when value-numbering same-sized store by inserting a
2408 (eliminate_dom_walker::eliminate_stmt): When eliminating
2409 a redundant store handle bit-reinterpretation of the same value.
2411 2020-01-21 Andrew Pinski <apinski@marvel.com>
2414 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2416 (prepare_block_for_update): This. Use a worklist instead of
2419 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2421 * gcc/config/arm/arm.c (clear_operation_p):
2422 Initialise last_regno, skip first iteration
2423 based on the first_set value and use ints instead
2424 of the unnecessary HOST_WIDE_INTs.
2426 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2429 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2430 compare_mode other than SFmode or DFmode.
2432 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2435 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2436 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2437 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2439 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2441 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2443 2020-01-20 Andrew Pinski <apinski@marvell.com>
2446 * targhooks.c (default_print_patchable_function_entry): Use
2447 output_asm_insn to emit the nop instruction.
2449 2020-01-20 Fangrui Song <maskray@google.com>
2452 * targhooks.c (default_print_patchable_function_entry): Align to
2455 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2458 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2459 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2460 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2461 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2462 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2463 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2464 Remove the {q} suffix from lea.
2465 (*tls_dynamic_gnu2_call_64): Renamed to ...
2466 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2467 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2468 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2469 Pass Pmode to gen_tls_dynamic_gnu2_64.
2471 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2473 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2475 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2477 * config/aarch64/aarch64-sve-builtins-base.cc
2478 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2480 2020-01-20 Richard Biener <rguenther@suse.de>
2483 * dwarf2out.c (prune_unused_types): Unconditionally mark
2484 called function DIEs.
2486 2020-01-20 Martin Liska <mliska@suse.cz>
2488 PR tree-optimization/93199
2489 * tree-eh.c (struct leh_state): Add
2490 new field outer_non_cleanup.
2491 (cleanup_is_dead_in): Pass leh_state instead
2492 of eh_region. Add a checking that state->outer_non_cleanup
2493 points to outer non-clean up region.
2494 (lower_try_finally): Record outer_non_cleanup
2496 (lower_catch): Likewise.
2497 (lower_eh_filter): Likewise.
2498 (lower_eh_must_not_throw): Likewise.
2499 (lower_cleanup): Likewise.
2501 2020-01-20 Richard Biener <rguenther@suse.de>
2503 PR tree-optimization/93094
2504 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2505 (vect_transform_loop): Likewise.
2506 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2507 loop_vectorized_call to vect_transform_loop.
2508 * tree-vect-loop.c (vect_transform_loop): Pass down
2509 loop_vectorized_call to vect_loop_versioning.
2510 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2511 the earlier discovered loop_vectorized_call.
2513 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2515 * doc/contribute.texi: Update for SVN -> Git transition.
2516 * doc/install.texi: Likewise.
2518 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2521 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2522 speculative targets.
2523 (verify_speculative_call): New function
2524 (cgraph_node::verify_node): Use it.
2525 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2528 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2531 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2532 (cgraph_edge::make_direct): Remove all indirect targets.
2533 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2534 (cgraph_node::verify_node): Verify that only one call_stmt or
2535 lto_stmt_uid is set.
2536 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2538 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2539 (lto_output_ref): Simplify streaming of stmt.
2540 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2542 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2544 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2545 Mark parameter unused.
2547 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2549 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2551 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2553 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2555 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2557 * Makefile.in: Add coroutine-passes.o.
2558 * builtin-types.def (BT_CONST_SIZE): New.
2559 (BT_FN_BOOL_PTR): New.
2560 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2561 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2562 * coroutine-builtins.def: New file.
2563 * coroutine-passes.cc: New file.
2564 * function.h (struct GTY function): Add a bit to indicate that the
2565 function is a coroutine component.
2566 * internal-fn.c (expand_CO_FRAME): New.
2567 (expand_CO_YIELD): New.
2568 (expand_CO_SUSPN): New.
2569 (expand_CO_ACTOR): New.
2570 * internal-fn.def (CO_ACTOR): New.
2574 * passes.def: Add pass_coroutine_lower_builtins,
2575 pass_coroutine_early_expand_ifns.
2576 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2577 (make_pass_coroutine_early_expand_ifns): New.
2578 * doc/invoke.texi: Document the fcoroutines command line
2581 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2583 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2586 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2587 after checking the argument is a REG. Don't use REGNO (reg)
2588 again to set last_regno, reuse regno variable instead.
2590 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2592 * doc/analyzer.texi (Limitations): Add note about NaN.
2594 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2595 Sudakshina Das <sudi.das@arm.com>
2597 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2598 and valid immediate.
2599 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2600 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2601 * config/arm/constraints.md (Pg): New.
2602 * config/arm/predicates.md (long_shift_imm): New.
2603 (arm_reg_or_long_shift_imm): Likewise.
2604 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2605 (thumb2_lsll): Likewise.
2608 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2609 Sudakshina Das <sudi.das@arm.com>
2611 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2612 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2613 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2614 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2615 * config/arm/thumb2.md (thumb2_asrl): New.
2616 (thumb2_lsll): Likewise.
2618 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2620 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2623 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2625 * gdbinit.in (help-gcc-hooks): New command.
2626 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2627 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2630 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2632 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2633 correct target macro.
2635 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2637 * config/aarch64/aarch64-protos.h
2638 (aarch64_sve_ld1ro_operand_p): New.
2639 * config/aarch64/aarch64-sve-builtins-base.cc
2640 (class load_replicate): New.
2641 (class svld1ro_impl): New.
2642 (class svld1rq_impl): Change to inherit from load_replicate.
2643 (svld1ro): New sve intrinsic function base.
2644 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2645 New DEF_SVE_FUNCTION.
2646 * config/aarch64/aarch64-sve-builtins-base.h
2647 (svld1ro): New decl.
2648 * config/aarch64/aarch64-sve-builtins.cc
2649 (function_expander::add_mem_operand): Modify assert to allow
2651 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2653 * config/aarch64/aarch64.c
2654 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2655 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2656 (aarch64_sve_ld1ro_operand_p): New.
2657 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2658 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2659 * config/aarch64/predicates.md
2660 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2662 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2664 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2665 Introduce this ACLE specified predefined macro.
2666 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2667 (fp): Disabling this disables f64mm.
2668 (simd): Disabling this disables f64mm.
2669 (fp16): Disabling this disables f64mm.
2670 (sve): Disabling this disables f64mm.
2671 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2672 (AARCH64_ISA_F64MM): New.
2673 (TARGET_F64MM): New.
2674 * doc/invoke.texi (f64mm): Document new option.
2676 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2678 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2679 (neoversen1_tunings): Likewise.
2681 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2684 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2685 Add assert to ensure prolog has been emitted.
2686 (aarch64_split_atomic_op): Likewise.
2687 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2688 Use epilogue_completed rather than reload_completed.
2689 (aarch64_atomic_exchange<mode>): Likewise.
2690 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2691 (atomic_nand<mode>): Likewise.
2692 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2693 (atomic_fetch_nand<mode>): Likewise.
2694 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2695 (atomic_nand_fetch<mode>): Likewise.
2697 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2700 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2702 (REVERSE_CONDITION): Delete.
2703 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2704 (CCFP_CCFPE): Likewise.
2705 (e): New mode attribute.
2706 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2707 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2708 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2709 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2710 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2711 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2712 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2713 name of generator from gen_ccmpdi to gen_ccmpccdi.
2714 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2715 the previous comparison but aren't able to, use the new ccmp_rev
2718 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2720 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2721 than testing directly for INTEGER_CST.
2722 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2724 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2726 PR tree-optimization/93292
2727 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2728 get_vectype_for_scalar_type returns NULL.
2730 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2732 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2733 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2735 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2737 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2739 * params.opt: (max-predicted-iterations): Set bounds.
2740 * predict.c (real_almost_one, real_br_prob_base,
2741 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2742 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2743 probabilities; do not truncate to reg_br_prob_bases.
2744 (estimate_loops_at_level): Pass max_cyclic_prob.
2745 (estimate_loops): Compute max_cyclic_prob.
2746 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2748 * profile-count.c (profile_probability::to_sreal): New.
2749 * profile-count.h (class sreal): Move up in file.
2750 (profile_probability::to_sreal): Declare.
2752 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2755 (arm_invalid_conversion): New function for target hook.
2756 (arm_invalid_unary_op): New function for target hook.
2757 (arm_invalid_binary_op): New function for target hook.
2759 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2761 * config.gcc: Add arm_bf16.h.
2762 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2763 (arm_simd_builtin_std_type): Add BFmode.
2764 (arm_init_simd_builtin_types): Define element types for vector types.
2765 (arm_init_bf16_types): New function.
2766 (arm_init_builtins): Add arm_init_bf16_types function call.
2767 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2768 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2769 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2770 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2771 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2772 (arm_mangle_type): Add __bf16.
2773 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2774 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2775 arm_bf16_ptr_type_node.
2776 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2777 define_split between ARM registers.
2778 * config/arm/arm_bf16.h: New file.
2779 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2780 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2782 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2783 * config/arm/vfp.md: Add BFmode to movhf patterns.
2785 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2786 Andre Vieira <andre.simoesdiasvieira@arm.com>
2788 * config/arm/arm-cpus.in (mve, mve_float): New features.
2789 (dsp, mve, mve.fp): New options.
2790 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2791 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2792 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2794 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2795 Thomas Preud'homme <thomas.preudhomme@arm.com>
2797 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2799 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2800 error for using -mcmse when targeting Armv8.1-M Mainline.
2802 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2803 Thomas Preud'homme <thomas.preudhomme@arm.com>
2805 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2806 address in r4 when targeting Armv8.1-M Mainline.
2807 (nonsecure_call_value_internal): Likewise.
2808 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2809 a register match_operand again. Emit BLXNS when targeting
2811 (nonsecure_call_value_reg_thumb2): Likewise.
2813 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2814 Thomas Preud'homme <thomas.preudhomme@arm.com>
2816 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2817 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2818 variable as true when floating-point ABI is not hard. Replace
2819 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2820 Generate VLSTM and VLLDM instruction respectively before and
2821 after a function call to cmse_nonsecure_call function.
2822 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2823 (VUNSPEC_VLLDM): Likewise.
2824 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2825 (lazy_load_multiple_insn): Likewise.
2827 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2828 Thomas Preud'homme <thomas.preudhomme@arm.com>
2830 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2831 (arm_emit_vfp_multi_reg_pop): Likewise.
2832 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2833 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2834 restore callee-saved VFP registers.
2836 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2837 Thomas Preud'homme <thomas.preudhomme@arm.com>
2839 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2840 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2841 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2842 callee-saved GPRs as well as clear ip register before doing a nonsecure
2843 call then restore callee-saved GPRs after it when targeting
2845 (arm_reorg): Adapt to function rename.
2847 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2848 Thomas Preud'homme <thomas.preudhomme@arm.com>
2850 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2851 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2852 clear_vfp_multiple pattern based on a new vfp parameter.
2853 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2854 targeting Armv8.1-M Mainline.
2855 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2856 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2857 whether VFP registers are available before looking call_used_regs for a
2859 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2860 of prototype of clear_operation_p.
2861 (clear_vfp_multiple_operation): New predicate.
2862 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2863 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2865 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2866 Thomas Preud'homme <thomas.preudhomme@arm.com>
2868 * config/arm/arm-protos.h (clear_operation_p): Declare.
2869 * config/arm/arm.c (clear_operation_p): New function.
2870 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2871 targeting Armv8.1-M Mainline or successor.
2872 (output_return_instruction): Only output APSR register clearing if
2873 Armv8.1-M Mainline instructions not available.
2874 (thumb_exit): Likewise.
2875 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2876 * config/arm/thumb2.md (clear_apsr): New define_insn.
2877 (clear_multiple): Likewise.
2878 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2880 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2881 Thomas Preud'homme <thomas.preudhomme@arm.com>
2883 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2884 (use_return_insn): Also return false for Armv8.1-M Mainline.
2885 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2886 Mainline instructions are available.
2887 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2888 when targeting Armv8.1-M Mainline Security Extensions.
2889 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2890 Mainline entry function.
2891 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2892 targeting Armv8.1-M Mainline or successor.
2893 (arm_expand_epilogue): Fix indentation of caller-saved register
2894 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2896 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2897 (FP_SYSREGS): Likewise.
2898 (enum vfp_sysregs_encoding): Define enum.
2899 (fp_sysreg_names): Declare.
2900 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2901 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2902 (pop_fpsysreg_insn): Likewise.
2904 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2905 Thomas Preud'homme <thomas.preudhomme@arm.com>
2907 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2908 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2909 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2910 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2911 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2912 (ARMv8_1m_main): New feature group.
2913 (armv8.1-m.main): New architecture.
2914 * config/arm/arm-tables.opt: Regenerate.
2915 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2916 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2917 (arm_options_perform_arch_sanity_checks): Error out when targeting
2918 Armv8.1-M Mainline Security Extensions.
2919 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2921 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2923 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2924 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2925 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2926 aarch64_bfdot_laneq): New.
2927 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2928 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2929 vbfdotq_laneq_f32): New.
2930 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2931 VBFMLA_W, VBF): New.
2932 (isquadop): Add V4BF, V8BF.
2934 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2936 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2937 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2938 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2939 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2940 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2941 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2942 usdot_laneq, sudot_lane,sudot_laneq): New.
2943 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2944 (aarch64_<sur>dot_lane): New.
2945 * config/aarch64/arm_neon.h (vusdot_s32): New.
2947 (vusdot_lane_s32): New.
2948 (vsudot_lane_s32): New.
2949 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2950 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2952 2020-01-16 Martin Liska <mliska@suse.cz>
2954 * value-prof.c (dump_histogram_value): Fix
2955 obvious spacing issue.
2957 2020-01-16 Andrew Pinski <apinski@marvell.com>
2959 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2960 !storage_order_barrier_p.
2962 2020-01-16 Andrew Pinski <apinski@marvell.com>
2964 * sched-int.h (_dep): Add unused bit-field field for the padding.
2965 * sched-deps.c (init_dep_1): Init unused field.
2967 2020-01-16 Andrew Pinski <apinski@marvell.com>
2969 * optabs.h (create_expand_operand): Initialize target field also.
2971 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2973 PR tree-optimization/92429
2974 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2975 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2977 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2980 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2982 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2983 aarch64_sve_int_mode to each mode.
2985 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2987 * doc/analyzer.texi (Overview): Add note about
2988 -fdump-ipa-analyzer.
2990 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2992 PR tree-optimization/93231
2993 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2994 input_type is unsigned. Use tree_to_shwi for shift constant.
2995 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2996 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2997 inputs more efficiently.
2999 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
3001 * config/i386/i386.md (*movsf_internal): Do not require
3002 SSE2 ISA for alternatives 14 and 15.
3004 2020-01-15 Richard Biener <rguenther@suse.de>
3007 * tree-eh.c (sink_clobbers): If we already visited the destination
3008 block do not defer insertion.
3009 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
3010 the purpose of defered insertion.
3012 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3014 * BASE-VER: Bump to 10.0.1.
3016 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3018 PR tree-optimization/93247
3019 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
3020 type of the stmt that we're going to vectorize.
3022 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
3024 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
3025 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3028 2020-01-15 Martin Liska <mliska@suse.cz>
3030 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3031 2 calls of streamer_read_hwi in a function call.
3033 2020-01-15 Richard Biener <rguenther@suse.de>
3035 * alias.c (record_alias_subset): Avoid redundant work when
3036 subset is already recorded.
3038 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3040 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3041 the analyzer options provide CWE identifiers.
3043 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3045 * tree-diagnostic-path.cc (path_summary::event_range::print):
3046 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3047 using get_pure_location.
3049 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3051 PR tree-optimization/93262
3052 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3053 perform head trimming only if the last argument is constant,
3054 either all ones, or larger or equal to head trim, in the latter
3055 case decrease the last argument by head_trim.
3057 PR tree-optimization/93249
3058 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3059 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3060 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3061 perform head trim unless we can prove there are no '\0' chars
3062 from the source among the first head_trim chars.
3064 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3066 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3068 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3071 * config/i386/sse.md
3072 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3073 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3074 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3075 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3076 just a single alternative instead of two, make operands 1 and 2
3079 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3082 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3085 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3087 * Makefile.in (lang_opt_files): Add analyzer.opt.
3088 (ANALYZER_OBJS): New.
3089 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3090 tristate.o and ANALYZER_OBJS.
3091 (TEXI_GCCINT_FILES): Add analyzer.texi.
3092 * common.opt (-fanalyzer): New driver option.
3093 * config.in: Regenerate.
3094 * configure: Regenerate.
3095 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3096 (gccdepdir): Also create depdir for "analyzer" subdir.
3097 * digraph.cc: New file.
3098 * digraph.h: New file.
3099 * doc/analyzer.texi: New file.
3100 * doc/gccint.texi ("Static Analyzer") New menu item.
3101 (analyzer.texi): Include it.
3102 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3103 ("Warning Options"): Add static analysis warnings to the list.
3104 (-Wno-analyzer-double-fclose): New option.
3105 (-Wno-analyzer-double-free): New option.
3106 (-Wno-analyzer-exposure-through-output-file): New option.
3107 (-Wno-analyzer-file-leak): New option.
3108 (-Wno-analyzer-free-of-non-heap): New option.
3109 (-Wno-analyzer-malloc-leak): New option.
3110 (-Wno-analyzer-possible-null-argument): New option.
3111 (-Wno-analyzer-possible-null-dereference): New option.
3112 (-Wno-analyzer-null-argument): New option.
3113 (-Wno-analyzer-null-dereference): New option.
3114 (-Wno-analyzer-stale-setjmp-buffer): New option.
3115 (-Wno-analyzer-tainted-array-index): New option.
3116 (-Wno-analyzer-use-after-free): New option.
3117 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3118 (-Wno-analyzer-use-of-uninitialized-value): New option.
3119 (-Wanalyzer-too-complex): New option.
3120 (-fanalyzer-call-summaries): New warning.
3121 (-fanalyzer-checker=): New warning.
3122 (-fanalyzer-fine-grained): New warning.
3123 (-fno-analyzer-state-merge): New warning.
3124 (-fno-analyzer-state-purge): New warning.
3125 (-fanalyzer-transitivity): New warning.
3126 (-fanalyzer-verbose-edges): New warning.
3127 (-fanalyzer-verbose-state-changes): New warning.
3128 (-fanalyzer-verbosity=): New warning.
3129 (-fdump-analyzer): New warning.
3130 (-fdump-analyzer-callgraph): New warning.
3131 (-fdump-analyzer-exploded-graph): New warning.
3132 (-fdump-analyzer-exploded-nodes): New warning.
3133 (-fdump-analyzer-exploded-nodes-2): New warning.
3134 (-fdump-analyzer-exploded-nodes-3): New warning.
3135 (-fdump-analyzer-supergraph): New warning.
3136 * doc/sourcebuild.texi (dg-require-dot): New.
3137 (dg-check-dot): New.
3138 * gdbinit.in (break-on-saved-diagnostic): New command.
3139 * graphviz.cc: New file.
3140 * graphviz.h: New file.
3141 * ordered-hash-map-tests.cc: New file.
3142 * ordered-hash-map.h: New file.
3143 * passes.def (pass_analyzer): Add before
3144 pass_ipa_whole_program_visibility.
3145 * selftest-run-tests.c (selftest::run_tests): Call
3146 selftest::ordered_hash_map_tests_cc_tests.
3147 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3149 * shortest-paths.h: New file.
3150 * timevar.def (TV_ANALYZER): New timevar.
3151 (TV_ANALYZER_SUPERGRAPH): Likewise.
3152 (TV_ANALYZER_STATE_PURGE): Likewise.
3153 (TV_ANALYZER_PLAN): Likewise.
3154 (TV_ANALYZER_SCC): Likewise.
3155 (TV_ANALYZER_WORKLIST): Likewise.
3156 (TV_ANALYZER_DUMP): Likewise.
3157 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3158 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3159 * tree-pass.h (make_pass_analyzer): New decl.
3160 * tristate.cc: New file.
3161 * tristate.h: New file.
3163 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3166 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3167 alternatives 9 and 10.
3169 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3171 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3172 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3173 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3174 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3175 (selftest::hash_map_tests_c_tests): Call it.
3176 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3177 New static constant, using the value of = H::empty_zero_p.
3178 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3179 from default_hash_traits <Value>.
3180 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3182 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3183 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3184 calls to mark_empty with !Descriptor::empty_zero_p.
3185 (hash_table::empty_slow): Conditionalize the memset call with a
3186 check that Descriptor::empty_zero_p; otherwise, loop through the
3187 entries calling mark_empty on them.
3188 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3189 (pointer_hash::empty_zero_p): Likewise.
3190 (pair_hash::empty_zero_p): Likewise.
3191 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3193 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3194 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3195 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3196 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3197 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3198 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3199 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3200 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3202 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3205 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3207 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3208 fix typo on return value.
3210 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3213 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3215 (cgraph_edge::make_speculative): Add param for setting speculative_id
3217 (cgraph_edge::speculative_call_info): Update comments and find reference
3218 by speculative_id for multiple indirect targets.
3219 (cgraph_edge::resolve_speculation): Decrease the speculations
3220 for indirect edge, drop it's speculative if not direct target
3221 left. Update comments.
3222 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3223 (cgraph_node::dump): Print num_speculative_call_targets.
3224 (cgraph_node::verify_node): Don't report error if speculative
3225 edge not include statement.
3226 (cgraph_edge::num_speculative_call_targets_p): New function.
3227 * cgraph.h (int common_target_id): Remove.
3228 (int common_target_probability): Remove.
3229 (num_speculative_call_targets): New variable.
3230 (make_speculative): Add param for setting speculative_id.
3231 (cgraph_edge::num_speculative_call_targets_p): New declare.
3232 (target_prob): New variable.
3233 (speculative_id): New variable.
3234 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3235 call summaries for multiple speculative call targets.
3236 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3237 * ipa-profile.c (struct speculative_call_target): New struct.
3238 (class speculative_call_summary): New class.
3239 (class speculative_call_summaries): New class.
3240 (call_sums): New variable.
3241 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3242 (ipa_profile_write_edge_summary): New function.
3243 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3244 (ipa_profile_dump_all_summaries): New function.
3245 (ipa_profile_read_edge_summary): New function.
3246 (ipa_profile_read_summary_section): New function.
3247 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3248 (ipa_profile): Generate num_speculative_call_targets from
3250 * ipa-ref.h (speculative_id): New variable.
3251 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3252 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3253 common_target_probability. Stream out speculative_id and
3254 num_speculative_call_targets.
3255 (input_edge): Likewise.
3256 * predict.c (dump_prediction): Remove edges count assert to be
3258 * symtab.c (symtab_node::create_reference): Init speculative_id.
3259 (symtab_node::clone_references): Clone speculative_id.
3260 (symtab_node::clone_referring): Clone speculative_id.
3261 (symtab_node::clone_reference): Clone speculative_id.
3262 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3263 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3264 if indirect call contains multiple speculative targets.
3265 * value-prof.h (check_ic_target): Remove.
3266 * value-prof.c (gimple_value_profile_transformations):
3267 Use void function gimple_ic_transform.
3268 * value-prof.c (gimple_ic_transform): Handle topn case.
3269 Fix comment typos. Change it to a void function.
3271 2020-01-13 Andrew Pinski <apinski@marvell.com>
3273 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3274 (octeontx2t98): New define.
3275 (octeontx2t96): New define.
3276 (octeontx2t93): New define.
3277 (octeontx2f95): New define.
3278 (octeontx2f95n): New define.
3279 (octeontx2f95mm): New define.
3280 * config/aarch64/aarch64-tune.md: Regenerate.
3281 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3283 2020-01-13 Jason Merrill <jason@redhat.com>
3285 PR c++/33799 - destroy return value if local cleanup throws.
3286 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3288 2020-01-13 Martin Liska <mliska@suse.cz>
3290 * ipa-cp.c (get_max_overall_size): Use newly
3291 renamed param param_ipa_cp_unit_growth.
3292 * params.opt: Remove legacy param name.
3294 2020-01-13 Martin Sebor <msebor@redhat.com>
3296 PR tree-optimization/93213
3297 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3298 stores to be eliminated.
3300 2020-01-13 Martin Liska <mliska@suse.cz>
3302 * opts.c (print_help): Do not print CL_PARAM
3303 and CL_WARNING for CL_OPTIMIZATION.
3305 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3308 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3309 depending on optimization settings.
3311 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3313 PR tree-optimization/90838
3314 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3315 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3316 argument rather than to initialize temporary for targets that
3317 don't use the mode argument at all. Initialize ctzval to avoid
3320 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3322 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3323 * tree-core.h: Document it.
3324 * gimplify.c (gimplify_omp_workshare): Set it.
3325 * omp-low.c (lower_omp_target): Use it.
3326 * tree-pretty-print.c (dump_omp_clause): Print it.
3328 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3329 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3331 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3333 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3334 * common.opt (fdiagnostics-path-format=): New option.
3335 (diagnostic_path_format): New enum.
3336 (fdiagnostics-show-path-depths): New option.
3337 * coretypes.h (diagnostic_event_id_t): New forward decl.
3338 * diagnostic-color.c (color_dict): Add "path".
3339 * diagnostic-event-id.h: New file.
3340 * diagnostic-format-json.cc (json_from_expanded_location): Make
3342 (json_end_diagnostic): Call context->make_json_for_path if it
3343 exists and the diagnostic has a path.
3344 (diagnostic_output_format_init): Clear context->print_path.
3345 * diagnostic-path.h: New file.
3346 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3347 when printing a run of events in a diagnostic_path so that they
3348 all get the same color.
3349 (layout::m_diagnostic_path_p): New field.
3350 (layout::layout): Initialize it.
3351 (layout::print_any_labels): Don't colorize the label text for an
3352 event in a diagnostic_path.
3353 (gcc_rich_location::add_location_if_nearby): Add
3354 "restrict_to_current_line_spans" and "label" params. Pass the
3355 former to layout.maybe_add_location_range; pass the latter
3356 when calling add_range.
3357 * diagnostic.c: Include "diagnostic-path.h".
3358 (diagnostic_initialize): Initialize context->path_format and
3359 context->show_path_depths.
3360 (diagnostic_show_any_path): New function.
3361 (diagnostic_path::interprocedural_p): New function.
3362 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3363 (simple_diagnostic_path::num_events): New function.
3364 (simple_diagnostic_path::get_event): New function.
3365 (simple_diagnostic_path::add_event): New function.
3366 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3367 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3368 (debug): New overload taking a diagnostic_path *.
3369 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3370 * diagnostic.h (enum diagnostic_path_format): New enum.
3371 (json::value): New forward decl.
3372 (diagnostic_context::path_format): New field.
3373 (diagnostic_context::show_path_depths): New field.
3374 (diagnostic_context::print_path): New callback field.
3375 (diagnostic_context::make_json_for_path): New callback field.
3376 (diagnostic_show_any_path): New decl.
3377 (json_from_expanded_location): New decl.
3378 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3379 (-fdiagnostics-show-path-depths): New option.
3380 (-fdiagnostics-color): Add "path" to description of default
3381 GCC_COLORS; describe it.
3382 (-fdiagnostics-format=json): Document how diagnostic paths are
3383 represented in the JSON output format.
3384 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3385 Add optional params "restrict_to_current_line_spans" and "label".
3386 * opts.c (common_handle_option): Handle
3387 OPT_fdiagnostics_path_format_ and
3388 OPT_fdiagnostics_show_path_depths.
3389 * pretty-print.c: Include "diagnostic-event-id.h".
3390 (pp_format): Implement "%@" format code for printing
3391 diagnostic_event_id_t *.
3392 (selftest::test_pp_format): Add tests for "%@".
3393 * selftest-run-tests.c (selftest::run_tests): Call
3394 selftest::tree_diagnostic_path_cc_tests.
3395 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3396 * toplev.c (general_init): Initialize global_dc->path_format and
3397 global_dc->show_path_depths.
3398 * tree-diagnostic-path.cc: New file.
3399 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3400 non-static. Drop "diagnostic" param in favor of storing the
3401 original value of "where" and re-using it.
3402 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3403 maybe_unwind_expanded_macro_loc.
3404 (tree_diagnostics_defaults): Initialize context->print_path and
3405 context->make_json_for_path.
3406 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3408 (default_tree_make_json_for_path): New decl.
3409 (maybe_unwind_expanded_macro_loc): New decl.
3411 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3413 PR tree-optimization/93210
3414 * fold-const.h (native_encode_initializer,
3415 can_native_interpret_type_p): Declare.
3416 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3418 (native_encode_initializer): New function, moved from dwarf2out.c.
3419 Adjust to native_encode_expr compatible arguments, including dry-run
3420 and partial extraction modes. Don't handle STRING_CST.
3421 (can_native_interpret_type_p): No longer static.
3422 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3423 offset / BITS_PER_UNIT fits into int and don't call it if
3424 can_native_interpret_type_p fails. If suboff is NULL and for
3425 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3426 native_encode_initializer.
3427 (fold_const_aggregate_ref_1): Formatting fix.
3428 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3429 (tree_add_const_value_attribute): Adjust caller.
3431 PR tree-optimization/90838
3432 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3433 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3434 CTZ_DEFINED_VALUE_AT_ZERO.
3436 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3439 * lra-constraints.c (match_reload): Permit input operands have the
3440 same mode as output while other input operands have a different
3443 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3445 PR tree-optimization/90838
3446 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3447 (check_ctz_string): Likewise.
3448 (optimize_count_trailing_zeroes): Likewise.
3449 (simplify_count_trailing_zeroes): Likewise.
3450 (pass_forwprop::execute): Try ctz simplification.
3451 * match.pd: Add matching for ctz idioms.
3453 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3455 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3457 (aarch64_invalid_unary_op): New function for target hook.
3458 (aarch64_invalid_binary_op): New function for target hook.
3460 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3462 * config.gcc: Add arm_bf16.h.
3463 * config/aarch64/aarch64-builtins.c
3464 (aarch64_simd_builtin_std_type): Add BFmode.
3465 (aarch64_init_simd_builtin_types): Define element types for vector
3467 (aarch64_init_bf16_types): New function.
3468 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3469 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3471 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3472 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3474 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3475 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3476 * config/aarch64/aarch64.c
3477 (aarch64_classify_vector_mode): Add support for BF types.
3478 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3479 (aarch64_vq_mode): Add support for BF types.
3480 (aarch64_simd_container_mode): Add support for BF types.
3481 (aarch64_mangle_type): Add support for BF scalar type.
3482 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3483 * config/aarch64/arm_bf16.h: New file.
3484 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3485 * config/aarch64/iterators.md: Add BF types to mode attributes.
3486 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3488 2020-01-10 Jason Merrill <jason@redhat.com>
3490 PR c++/93173 - incorrect tree sharing.
3491 * gimplify.c (copy_if_shared): No longer static.
3492 * gimplify.h: Declare it.
3494 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3496 * doc/invoke.texi (-msve-vector-bits=): Document that
3497 -msve-vector-bits=128 now generates VL-specific code for
3498 little-endian targets.
3499 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3500 build_vector_type_for_mode to construct the data vector types.
3501 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3502 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3503 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3504 for 128-bit vectors.
3506 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3508 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3511 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3513 * config/aarch64/aarch64-builtins.c
3514 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3515 rather than checking the number of elements and the element mode.
3517 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3519 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3520 get_related_vectype_for_scalar_type rather than build_vector_type
3521 to create the index type for a conditional reduction.
3523 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3525 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3526 for any type of gather or scatter, including strided accesses.
3528 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3530 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3533 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3535 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3537 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3538 parameter and its use to reset DR_OFFSET's.
3539 (vect_transform_loop): Remove orig_drs_init argument.
3540 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3541 member of dr_vec_info rather than the offset of the associated
3542 data_reference's innermost_loop_behavior.
3543 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3544 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3545 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3546 get_dr_vinfo_offset.
3547 (vectorizable_store): Likewise.
3548 (vectorizable_load): Likewise.
3550 2020-01-10 Richard Biener <rguenther@suse.de>
3552 * gimple-ssa-store-merging
3553 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3555 2020-01-10 Martin Liska <mliska@suse.cz>
3558 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3559 encapsulation that was there before r280040.
3561 2020-01-10 Richard Biener <rguenther@suse.de>
3564 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3565 sequences to avoid walking them again for secondary opportunities.
3566 (pass_lower_eh_dispatch::execute): Instead actually insert
3569 2020-01-10 Richard Biener <rguenther@suse.de>
3572 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3573 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3574 avoid quadraticness.
3576 2020-01-10 Martin Jambor <mjambor@suse.cz>
3578 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3579 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3580 to get param_ipa_sra_max_replacements.
3581 (param_splitting_across_edge): Pass the caller to
3582 pull_accesses_from_callee.
3584 2020-01-10 Martin Jambor <mjambor@suse.cz>
3586 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3587 * ipa-cp.c (max_new_size): Removed.
3588 (orig_overall_size): New variable.
3589 (get_max_overall_size): New function.
3590 (estimate_local_effects): Use it. Adjust dump.
3591 (decide_about_value): Likewise.
3592 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3593 orig_overall_size. Adjust dump.
3594 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3596 2020-01-10 Martin Jambor <mjambor@suse.cz>
3598 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3599 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3600 instead of param_ipa_max_agg_items.
3601 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3602 optimization info for the callee.
3604 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3606 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3607 markers if debug_inline_points is false.
3609 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3611 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3613 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3614 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3615 aarch64-sve-builtins-sve2.h.
3616 (aarch64-sve-builtins-sve2.o): New rule.
3617 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3618 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3619 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3620 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3621 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3622 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3624 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3625 instructions that are handled here.
3626 (@cond_asrd<mode>): Generalize to...
3627 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3628 (*cond_asrd<mode>_2): Generalize to...
3629 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3630 (*cond_asrd<mode>_z): Generalize to...
3631 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3632 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3633 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3634 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3635 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3637 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3638 (@aarch64_scatter_stnt<mode>): Likewise.
3639 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3640 (@aarch64_mul_lane_<mode>): Likewise.
3641 (@aarch64_sve_suqadd<mode>_const): Likewise.
3642 (*<sur>h<addsub><mode>): Generalize to...
3643 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3645 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3646 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3647 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3648 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3649 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3650 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3651 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3652 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3653 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3654 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3655 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3656 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3657 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3658 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3659 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3660 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3661 (@aarch64_sve2_xar<mode>): Likewise.
3662 (@aarch64_sve2_bcax<mode>): Likewise.
3663 (*aarch64_sve2_eor3<mode>): Rename to...
3664 (@aarch64_sve2_eor3<mode>): ...this.
3665 (@aarch64_sve2_bsl<mode>): New expander.
3666 (@aarch64_sve2_nbsl<mode>): Likewise.
3667 (@aarch64_sve2_bsl1n<mode>): Likewise.
3668 (@aarch64_sve2_bsl2n<mode>): Likewise.
3669 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3670 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3671 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3672 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3673 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3674 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3675 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3676 (<su>mull<bt><Vwide>): Generalize to...
3677 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3679 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3680 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3681 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3682 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3683 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3684 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3685 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3686 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3687 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3688 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3689 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3690 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3691 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3692 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3693 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3694 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3695 (<SHRNB:r>shrnb<mode>): Generalize to...
3696 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3698 (<SHRNT:r>shrnt<mode>): Generalize to...
3699 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3701 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3702 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3703 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3704 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3705 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3706 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3707 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3708 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3709 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3710 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3711 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3712 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3713 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3714 (@aarch64_sve2_cvtnt<mode>): Likewise.
3715 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3716 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3717 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3718 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3719 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3720 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3721 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3722 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3723 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3724 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3725 (@aarch64_sve2_pmul<mode>): Likewise.
3726 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3727 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3728 (@aarch64_sve2_tbl2<mode>): Likewise.
3729 (@aarch64_sve2_tbx<mode>): Likewise.
3730 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3731 (@aarch64_sve2_histcnt<mode>): Likewise.
3732 (@aarch64_sve2_histseg<mode>): Likewise.
3733 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3734 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3735 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3736 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3737 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3738 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3739 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3740 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3741 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3742 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3743 (SVE2_PMULL_PAIR_I): New mode iterators.
3744 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3745 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3746 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3747 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3748 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3749 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3750 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3751 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3752 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3753 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3754 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3755 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3756 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3757 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3758 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3759 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3760 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3761 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3762 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3763 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3764 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3765 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3766 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3767 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3768 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3769 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3770 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3771 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3772 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3773 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3774 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3776 (VNARROW, Ventype): New mode attributes.
3777 (Vewtype): Handle VNx2DI. Fix typo in comment.
3778 (VDOUBLE): New mode attribute.
3779 (sve_lane_con): Handle VNx8HI.
3780 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3781 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3782 (sve_int_op, sve_int_op_rev): Handle the above codes.
3783 (sve_pred_int_rhs2_operand): Likewise.
3784 (MULLBT, SHRNB, SHRNT): Delete.
3785 (SVE_INT_SHIFT_IMM): New int iterator.
3786 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3787 and UNSPEC_WHILEHS for TARGET_SVE2.
3788 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3789 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3790 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3791 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3792 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3793 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3794 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3795 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3796 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3797 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3798 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3799 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3800 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3801 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3802 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3803 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3804 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3805 (optab): Handle the new unspecs.
3806 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3808 (lr): Handle the new unspecs.
3810 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3811 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3812 (sve_int_qsub_op): New int attributes.
3813 (sve_fp_op, rot): Handle the new unspecs.
3814 * config/aarch64/aarch64-sve-builtins.h
3815 (function_resolver::require_matching_pointer_type): Declare.
3816 (function_resolver::resolve_unary): Add an optional boolean argument.
3817 (function_resolver::finish_opt_n_resolution): Add an optional
3818 type_suffix_index argument.
3819 (gimple_folder::redirect_call): Declare.
3820 (gimple_expander::prepare_gather_address_operands): Add an optional
3822 * config/aarch64/aarch64-sve-builtins.cc: Include
3823 aarch64-sve-builtins-sve2.h.
3824 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3825 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3826 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3827 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3828 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3829 (TYPES_s_unsigned): Likewise.
3830 (TYPES_s_integer): Use TYPES_s_unsigned.
3831 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3832 (TYPES_sd_integer): Use them.
3833 (TYPES_d_unsigned): New macro.
3834 (TYPES_d_integer): Use it.
3835 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3836 (TYPES_cvt_narrow): Likewise.
3837 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3838 (preds_mx): New variable.
3839 (function_builder::add_overloaded_function): Allow the new feature
3840 set to be more restrictive than the original one.
3841 (function_resolver::infer_pointer_type): Remove qualifiers from
3842 the pointer type before printing it.
3843 (function_resolver::require_matching_pointer_type): New function.
3844 (function_resolver::resolve_sv_displacement): Handle functions
3845 that don't support 32-bit vector indices or svint32_t vector offsets.
3846 (function_resolver::finish_opt_n_resolution): Take the inferred type
3847 as a separate argument.
3848 (function_resolver::resolve_unary): Optionally treat all forms in
3849 the same way as normal merging functions.
3850 (gimple_folder::redirect_call): New function.
3851 (function_expander::prepare_gather_address_operands): Add an argument
3852 that says whether scaled forms are available. If they aren't,
3853 handle scaling of vector indices and don't add the extension and
3855 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3856 fall back to using cond_* instead.
3857 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3858 Split out the member variables into...
3859 (rtx_code_function_base): ...this new base class.
3860 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3861 (unspec_based_function): Split out the member variables into...
3862 (unspec_based_function_base): ...this new base class.
3863 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3864 (unspec_based_function_exact_insn): New class.
3865 (unspec_based_add_function, unspec_based_add_lane_function)
3866 (unspec_based_lane_function, unspec_based_pred_function)
3867 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3868 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3869 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3871 (unspec_based_fused_function): New class.
3872 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3873 (unspec_based_fused_lane_function): New class.
3874 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3876 (CODE_FOR_MODE1): New macro.
3877 (fixed_insn_function): New class.
3878 (while_comparison): Likewise.
3879 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3880 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3881 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3882 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3883 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3884 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3885 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3886 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3887 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3888 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3889 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3890 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3891 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3892 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3893 Also add an initial argument for unary_convert_narrowt, regardless
3894 of the predication type.
3895 (build_32_64): Allow loads and stores to specify MODE_none.
3896 (build_sv_index64, build_sv_uint_offset): New functions.
3897 (long_type_suffix): New function.
3898 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3899 (binary_imm_long_base, load_gather_sv_base): Likewise.
3900 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3901 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3902 (unary_narrowb_base, unary_narrowt_base): Likewise.
3903 (binary_long_lane_def, binary_long_lane): New shape.
3904 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3905 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3906 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3907 (binary_to_uint_def, binary_to_uint): Likewise.
3908 (binary_wide_def, binary_wide): Likewise.
3909 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3910 (compare_def, compare): Likewise.
3911 (compare_ptr_def, compare_ptr): Likewise.
3912 (load_ext_gather_index_restricted_def,
3913 load_ext_gather_index_restricted): Likewise.
3914 (load_ext_gather_offset_restricted_def,
3915 load_ext_gather_offset_restricted): Likewise.
3916 (load_gather_sv_def): Inherit from load_gather_sv_base.
3917 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3918 (shift_left_imm_def, shift_left_imm): Likewise.
3919 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3920 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3921 (store_scatter_index_restricted_def,
3922 store_scatter_index_restricted): Likewise.
3923 (store_scatter_offset_restricted_def,
3924 store_scatter_offset_restricted): Likewise.
3925 (tbl_tuple_def, tbl_tuple): Likewise.
3926 (ternary_long_lane_def, ternary_long_lane): Likewise.
3927 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3928 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3929 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3930 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3931 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3932 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3933 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3934 (ternary_uint_def, ternary_uint): Likewise.
3935 (unary_convert): Fix typo in comment.
3936 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3937 (unary_long_def, unary_long): Likewise.
3938 (unary_narrowb_def, unary_narrowb): Likewise.
3939 (unary_narrowt_def, unary_narrowt): Likewise.
3940 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3941 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3942 (unary_to_int_def, unary_to_int): Likewise.
3943 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3944 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3945 (svasrd_impl): Delete.
3946 (svcadd_impl::expand): Handle integer operations too.
3947 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3948 new functions to derive the unspec numbers.
3949 (svmla_svmls_lane_impl): Replace with...
3950 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3951 integer operations too.
3952 (svwhile_impl): Rename to...
3953 (svwhilelx_impl): ...this and inherit from while_comparison.
3954 (svasrd): Use unspec_based_function.
3955 (svmla_lane): Use svmla_lane_impl.
3956 (svmls_lane): Use svmls_lane_impl.
3957 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3958 (svwhilele, svwhilelt): Use svwhilelx_impl.
3959 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3960 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3961 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3962 * config/aarch64/aarch64-sve-builtins.def: Include
3963 aarch64-sve-builtins-sve2.def.
3965 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3967 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3968 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3969 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3970 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3971 immediates as well as vector ones.
3972 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3973 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3974 (aarch64_sve_qsub_immediate): Update calls accordingly.
3976 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3978 * config/aarch64/aarch64-sve2.md: Add banner comments.
3979 (<su>mulh<r>s<mode>3): Move further up file.
3980 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3981 (*aarch64_sve2_sra<mode>): Move further down file.
3982 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3984 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3986 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3988 (while_optab_cmp): Handle them.
3989 * config/aarch64/aarch64-sve.md
3990 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3991 and add a "@" marker.
3992 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3993 instead of gen_aarch64_sve2_while_ptest.
3994 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3996 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3998 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3999 (UNSPEC_WHILELE): ...this.
4000 (UNSPEC_WHILE_LO): Rename to...
4001 (UNSPEC_WHILELO): ...this.
4002 (UNSPEC_WHILE_LS): Rename to...
4003 (UNSPEC_WHILELS): ...this.
4004 (UNSPEC_WHILE_LT): Rename to...
4005 (UNSPEC_WHILELT): ...this.
4006 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
4007 (cmp_op, while_optab_cmp): Likewise.
4008 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
4009 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
4010 (svwhilelt): Likewise.
4012 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4014 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
4015 (unary_to_uint): Define.
4016 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
4017 (unary_count): Rename to...
4018 (unary_to_uint_def, unary_to_uint): ...this.
4019 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
4021 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4023 * config/aarch64/aarch64-sve-builtins-functions.h
4024 (code_for_mode_function): New class.
4025 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
4026 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4027 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4028 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4029 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4031 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4033 * config/aarch64/iterators.md (addsub): New code attribute.
4034 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4036 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4037 in the asm string and attributes. Fix indentation.
4038 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4040 (@aarch64_sve_<optab><mode>): ...this.
4041 * config/aarch64/aarch64-sve-builtins.h
4042 (function_expander::expand_signed_unpred_op): Delete.
4043 * config/aarch64/aarch64-sve-builtins.cc
4044 (function_expander::expand_signed_unpred_op): Likewise.
4045 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4046 try using code_for_aarch64_sve instead.
4047 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4048 (svqsub_impl): Likewise.
4049 (svqadd, svqsub): Use rtx_code_function instead.
4051 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4053 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4054 (HADDSUB, sur, addsub): Remove them.
4056 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4058 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4059 functions the same way, rather than singling out those that
4060 aren't mapped directly to optabs.
4062 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4064 * target.def (compatible_vector_types_p): New target hook.
4065 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4066 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4067 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4068 * doc/tm.texi: Regenerate.
4069 * gimple-expr.c: Include target.h.
4070 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4071 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4073 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4074 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4075 Use the original predicate if it already has a suitable type.
4077 2020-01-09 Martin Jambor <mjambor@suse.cz>
4079 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4080 resolve_speculation and redirect_call_stmt_to_callee static. Change
4081 return type of set_call_stmt to cgraph_edge *.
4082 * auto-profile.c (afdo_indirect_call): Adjust call to
4083 redirect_call_stmt_to_callee.
4084 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4085 make the this pointer explicit, adjust self-recursive calls and the
4086 call top make_direct. Return the resulting edge.
4087 (cgraph_edge::remove): Make this pointer explicit.
4088 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4089 (cgraph_edge::make_direct): Likewise, adjust call to
4090 resolve_speculation.
4091 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4092 call to set_call_stmt.
4093 (cgraph_update_edges_for_call_stmt_node): Update call to
4094 set_call_stmt and remove.
4095 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4096 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4097 (cgraph_node::create_edge_including_clones): Moved "first" definition
4098 of edge to the block where it was used. Adjusted calls to
4100 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4101 cgraph_edge::remove.
4102 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4103 make_direct and redirect_call_stmt_to_callee.
4104 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4105 resolve_speculation and make_direct.
4106 * ipa-inline-transform.c (inline_transform): Adjust call to
4107 redirect_call_stmt_to_callee.
4108 (check_speculations_1):: Adjust call to resolve_speculation.
4109 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4110 resolve-speculation.
4111 (inline_small_functions): Adjust call to resolve_speculation.
4112 (ipa_inline): Likewise.
4113 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4115 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4116 safe with regards to edge removal, adjust calls to
4117 redirect_call_stmt_to_callee.
4118 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4119 and redirect_call_stmt_to_callee.
4120 * multiple_target.c (create_dispatcher_calls): Adjust call to
4121 redirect_call_stmt_to_callee
4122 (redirect_to_specific_clone): Likewise.
4123 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4124 Adjust calls to cgraph_edge::remove.
4125 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4126 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4127 (expand_call_inline): Adjust call to cgraph_edge::remove.
4129 2020-01-09 Martin Liska <mliska@suse.cz>
4131 * params.opt: Set Optimization for
4132 param_max_speculative_devirt_maydefs.
4134 2020-01-09 Martin Sebor <msebor@redhat.com>
4138 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4140 2020-01-09 Martin Liska <mliska@suse.cz>
4142 * auto-profile.c (auto_profile): Use opt_for_fn
4144 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4145 (propagate_vals_across_arith_jfunc): Likewise.
4146 (hint_time_bonus): Likewise.
4147 (incorporate_penalties): Likewise.
4148 (good_cloning_opportunity_p): Likewise.
4149 (perform_estimation_of_a_value): Likewise.
4150 (estimate_local_effects): Likewise.
4151 (ipcp_propagate_stage): Likewise.
4152 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4153 (set_switch_stmt_execution_predicate): Likewise.
4154 (analyze_function_body): Likewise.
4155 * ipa-inline-analysis.c (offline_size): Likewise.
4156 * ipa-inline.c (early_inliner): Likewise.
4157 * ipa-prop.c (ipa_analyze_node): Likewise.
4158 (ipcp_transform_function): Likewise.
4159 * ipa-sra.c (process_scan_results): Likewise.
4160 (ipa_sra_summarize_function): Likewise.
4161 * params.opt: Rename ipcp-unit-growth to
4162 ipa-cp-unit-growth. Add Optimization for various
4163 IPA-related parameters.
4165 2020-01-09 Richard Biener <rguenther@suse.de>
4168 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4170 2020-01-09 Richard Biener <rguenther@suse.de>
4172 PR tree-optimization/93040
4173 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4175 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4177 * common/config/avr/avr-common.c (avr_option_optimization_table)
4178 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4180 2020-01-09 Martin Liska <mliska@suse.cz>
4182 * cgraphclones.c (symbol_table::materialize_all_clones):
4183 Use cgraph_node::dump_name.
4185 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4188 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4189 output_operand_lossage instead of gcc_unreachable.
4190 * doc/md.texi (riscv f constraint): Fix typo.
4193 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4194 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4195 CONST_SCALAR_INT_P instead of CONST_INT_P.
4196 (*subv<mode>4_1): Rename to ...
4197 (subv<mode>4_1): ... this.
4198 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4199 define_insn_and_split patterns.
4200 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4203 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4205 * vec.c (class selftest::count_dtor): New class.
4206 (selftest::test_auto_delete_vec): New test.
4207 (selftest::vec_c_tests): Call it.
4208 * vec.h (class auto_delete_vec): New class template.
4209 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4211 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4213 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4215 2020-01-08 Jim Wilson <jimw@sifive.com>
4217 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4218 use of TLS_MODEL_LOCAL_EXEC when not pic.
4220 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4222 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4225 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4228 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4229 *stack_protect_set_3 peephole2): Also check that the second
4230 insns source is general_operand.
4233 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4234 predicate for output operand instead of register_operand.
4235 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4236 memory destination and non-memory operands[2].
4238 2020-01-08 Martin Liska <mliska@suse.cz>
4240 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4241 ::dump_asm_name instead of (::name or ::asm_name).
4242 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4243 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4244 (analyze_functions): Likewise.
4245 (expand_all_functions): Likewise.
4246 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4247 (propagate_bits_across_jump_function): Likewise.
4248 (dump_profile_updates): Likewise.
4249 (ipcp_store_bits_results): Likewise.
4250 (ipcp_store_vr_results): Likewise.
4251 * ipa-devirt.c (dump_targets): Likewise.
4252 * ipa-fnsummary.c (analyze_function_body): Likewise.
4253 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4254 (process_hsa_functions): Likewise.
4255 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4256 (set_alias_uids): Likewise.
4257 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4258 * ipa-inline.c (recursive_inlining): Likewise.
4259 (inline_to_all_callers_1): Likewise.
4260 (ipa_inline): Likewise.
4261 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4262 (ipa_propagate_frequency): Likewise.
4263 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4264 (remove_described_reference): Likewise.
4265 * ipa-pure-const.c (worse_state): Likewise.
4266 (check_retval_uses): Likewise.
4267 (analyze_function): Likewise.
4268 (propagate_pure_const): Likewise.
4269 (propagate_nothrow): Likewise.
4270 (dump_malloc_lattice): Likewise.
4271 (propagate_malloc): Likewise.
4272 (pass_local_pure_const::execute): Likewise.
4273 * ipa-visibility.c (optimize_weakref): Likewise.
4274 (function_and_variable_visibility): Likewise.
4275 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4276 (ipa_discover_variable_flags): Likewise.
4277 * lto-streamer-out.c (output_function): Likewise.
4278 (output_constructor): Likewise.
4279 * tree-inline.c (copy_bb): Likewise.
4280 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4281 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4283 2020-01-08 Richard Biener <rguenther@suse.de>
4286 * tree-eh.c (sink_clobbers): Update virtual operands for
4287 the first and last stmt only. Add a dry-run capability.
4288 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4289 after CFG manipulations and in RPO order to catch all
4290 secondary opportunities reliably.
4292 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4295 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4297 2019-01-08 Richard Biener <rguenther@suse.de>
4300 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4301 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4302 virtual operand, also updating SSA use.
4303 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4304 Update stmt after resetting virtual operand.
4305 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4306 * gimple-iterator.c (gsi_remove): When not removing the stmt
4307 permanently do not delink immediate uses or mark the stmt modified.
4309 2020-01-08 Martin Liska <mliska@suse.cz>
4311 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4312 (ipa_call_context::estimate_size_and_time): Likewise.
4313 (inline_analyze_function): Likewise.
4315 2020-01-08 Martin Liska <mliska@suse.cz>
4317 * cgraph.c (cgraph_node::dump): Use systematically
4320 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4322 Add -nodevicespecs option for avr.
4325 * config/avr/avr.opt (-nodevicespecs): New driver option.
4326 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4327 "-specs=device-specs/..." if that option is not set.
4328 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4330 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4332 Implement 64-bit double functions for avr.
4335 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4336 --with-double-comparison.
4337 * doc/install.texi: Document them.
4338 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4339 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4340 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4341 * doc/invoke.texi (AVR Built-in Macros): Document them.
4342 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4343 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4344 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4346 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4349 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4350 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4351 when only building rm-profile multilibs.
4353 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4356 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4357 lattice for a value to check.
4358 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4359 finite propagation in self-recursive scc.
4361 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4363 * ipa-inline.c (caller_growth_limits): Restore the AND.
4365 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4367 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4368 (VEC_ALLREG_ALT): New iterator.
4369 (VEC_ALLREG_INT_MODE): New iterator.
4370 (VCMP_MODE): New iterator.
4371 (VCMP_MODE_INT): New iterator.
4372 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4373 (vec_cmp<u>v64qidi): New define_expand.
4374 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4375 (vec_cmpu<mode>di_exec): New define_expand.
4376 (vec_cmp<u>v64qidi_exec): New define_expand.
4377 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4378 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4379 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4380 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4381 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4382 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4383 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4384 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4385 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4386 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4388 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4389 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4391 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4393 * config/gcn/constraints.md (DA): Update description and match.
4395 (Db): New constraint.
4396 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4398 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4399 Implement 'Db' mixed immediate type.
4400 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4401 (addcv64si3_dup<exec_vcc>): Delete.
4402 (subcv64si3<exec_vcc>): Rework constraints.
4403 (addv64di3): Rework constraints.
4404 (addv64di3_exec): Rework constraints.
4405 (subv64di3): Rework constraints.
4406 (addv64di3_dup): Delete.
4407 (addv64di3_dup_exec): Delete.
4408 (addv64di3_zext): Rework constraints.
4409 (addv64di3_zext_exec): Rework constraints.
4410 (addv64di3_zext_dup): Rework constraints.
4411 (addv64di3_zext_dup_exec): Rework constraints.
4412 (addv64di3_zext_dup2): Rework constraints.
4413 (addv64di3_zext_dup2_exec): Rework constraints.
4414 (addv64di3_sext_dup2): Rework constraints.
4415 (addv64di3_sext_dup2_exec): Rework constraints.
4417 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4419 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4420 existing target checks.
4422 2020-01-07 Richard Biener <rguenther@suse.de>
4424 * doc/install.texi: Bump minimal supported MPC version.
4426 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4428 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4429 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4430 * langhooks.c: Include stor-layout.h.
4431 (lhd_simulate_enum_decl): New function.
4432 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4433 handle_arm_sve_h for the LTO frontend.
4434 (register_vector_type): Cope with null returns from pushdecl.
4436 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4438 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4439 (aarch64_sve::nvectors_if_data_type): Replace with...
4440 (aarch64_sve::builtin_type_p): ...this.
4441 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4442 (find_vector_type): Delete.
4443 (add_sve_type_attribute): New function.
4444 (lookup_sve_type_attribute): Likewise.
4445 (register_builtin_types): Add an "SVE type" attribute to each type.
4446 (register_tuple_type): Likewise.
4447 (svbool_type_p, nvectors_if_data_type): Delete.
4448 (mangle_builtin_type): Use lookup_sve_type_attribute.
4449 (builtin_type_p): Likewise. Add an overload that returns the
4450 number of constituent vector and predicate registers.
4451 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4452 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4453 instead of aarch64_sve_argument_p.
4454 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4455 (aarch64_pass_by_reference): Likewise.
4456 (aarch64_function_value_1): Likewise.
4457 (aarch64_return_in_memory): Likewise.
4458 (aarch64_layout_arg): Likewise.
4460 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4462 PR tree-optimization/93156
4463 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4464 least significant bit is always clear.
4466 PR tree-optimization/93118
4467 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4468 simplifier with two intermediate conversions.
4470 2020-01-07 Martin Liska <mliska@suse.cz>
4472 * params.opt: Add Optimization for various parameters.
4474 2020-01-07 Martin Liska <mliska@suse.cz>
4477 * doc/extend.texi: Explain cloning for target_clone
4480 2020-01-07 Martin Liska <mliska@suse.cz>
4482 PR tree-optimization/92860
4483 * common.opt: Make in Optimization option
4484 as it is affected by -O0, which is an Optimization
4486 * tree-inline.c (tree_inlinable_function_p):
4487 Use opt_for_fn for warn_inline.
4488 (expand_call_inline): Likewise.
4490 2020-01-07 Martin Liska <mliska@suse.cz>
4492 PR tree-optimization/92860
4493 * common.opt: Make flag_ree as optimization
4496 2020-01-07 Martin Liska <mliska@suse.cz>
4498 PR optimization/92860
4499 * params.opt: Mark param_min_crossjump_insns with Optimization
4502 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4504 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4505 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4507 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4509 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4510 helper function to return the valid addressing formats for a given
4511 hard register and mode.
4512 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4514 * config/rs6000/constraints.md (Q constraint): Update
4516 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4519 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4520 Use 'Q' for doing vector extract from memory.
4521 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4523 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4524 doing vector extract from memory.
4525 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4526 extract from memory.
4528 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4529 for the offset being 34-bits when -mcpu=future is used.
4531 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4533 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4534 instead of cmpib_comparison_operator in cmpib patterns.
4535 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4536 of cmpib_comparison_operator. Revise comment.
4538 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4540 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4541 in an IFN_DIV_POW2 node to be equal.
4543 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4545 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4546 (vect_check_scalar_mask): ...this.
4547 (vectorizable_store, vectorizable_load): Update call accordingly.
4548 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4549 argument in calls to conditional internal functions.
4551 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4553 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4554 '0' matching inputs.
4555 (subv64di3_exec): Likewise.
4557 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4559 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4560 * doc/md.texi (movstr): Likewise.
4562 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4564 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4567 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4569 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4571 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4572 to a temporary file and use move-if-change to update the real
4573 file where necessary.
4575 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4577 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4578 rather than Upa for CPY /M.
4580 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4582 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4585 2020-01-06 Martin Liska <mliska@suse.cz>
4587 PR tree-optimization/92860
4588 * params.opt: Mark param_max_combine_insns with Optimization
4591 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4594 * config/i386/i386.md (SWIDWI): New mode iterator.
4595 (DWI, dwi): Add TImode variants.
4596 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4597 <general_hilo_operand> instead of <general_operand>. Use
4598 CONST_SCALAR_INT_P instead of CONST_INT_P.
4599 (*addv<mode>4_1): Rename to ...
4600 (addv<mode>4_1): ... this.
4601 (QWI): New mode attribute.
4602 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4603 define_insn_and_split patterns.
4604 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4606 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4607 <general_hilo_operand> instead of <general_operand>.
4608 (*addcarry<mode>_1): New define_insn.
4609 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4611 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4613 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4614 Use "call" instead of "set".
4616 2020-01-03 Martin Jambor <mjambor@suse.cz>
4619 * ipa-cp.c (print_all_lattices): Skip functions without info.
4621 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4624 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4625 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4626 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4627 for 'e' simd clones.
4630 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4632 (mprefer-vector-width=): Add Save.
4633 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4634 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4635 (ix86_debug_options, ix86_function_specific_print): Adjust
4636 ix86_target_string callers.
4637 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4638 (ix86_valid_target_attribute_tree): Likewise.
4639 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4640 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4641 ix86_target_string caller.
4644 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4645 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4646 instead of gen_int_shift_amount + convert_modes.
4648 PR rtl-optimization/93088
4649 * loop-iv.c (find_single_def_src): Punt after looking through
4650 128 reg copies for regs with single definitions. Move definitions
4653 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4655 * config/arm/arm-c.c (arm_cpu_builtins): Define
4656 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4657 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4658 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4659 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4660 * config/arm/arm-tables.opt: Regenerated.
4661 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4662 arm_arch_i8mm and arm_arch_bf16 when enabled.
4663 * config/arm/arm.h (TARGET_I8MM): New macro.
4664 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4665 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4666 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4667 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4668 (v8_6_a_simd_variants): New.
4669 (v8_*_a_simd_variants): Add i8mm and bf16.
4670 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4672 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4675 * predict.c (compute_function_frequency): Don't call
4676 warn_function_cold on functions that already have cold attribute.
4678 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4681 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4682 COMDAT group function labels in .data.rel.ro.local section.
4683 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4686 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4687 comparison_operator in B and S integer comparisons. Likewise, use
4688 ordered_comparison_operator instead of cmpib_comparison_operator in
4690 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4692 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4694 Update copyright years.
4696 * gcc.c (process_command): Update copyright notice dates.
4697 * gcov-dump.c (print_version): Ditto.
4698 * gcov.c (print_version): Ditto.
4699 * gcov-tool.c (print_version): Ditto.
4700 * gengtype.c (create_file): Ditto.
4701 * doc/cpp.texi: Bump @copying's copyright year.
4702 * doc/cppinternals.texi: Ditto.
4703 * doc/gcc.texi: Ditto.
4704 * doc/gccint.texi: Ditto.
4705 * doc/gcov.texi: Ditto.
4706 * doc/install.texi: Ditto.
4707 * doc/invoke.texi: Ditto.
4709 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4711 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4714 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4716 PR tree-optimization/93098
4717 * match.pd (popcount): For shift amounts, use integer_onep
4718 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4719 tests. Make sure that precision is power of two larger than or equal
4720 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4721 instead of ULL suffixed constants. Formatting fixes.
4723 Copyright (C) 2020 Free Software Foundation, Inc.
4725 Copying and distribution of this file, with or without modification,
4726 are permitted in any medium without royalty provided the copyright
4727 notice and this notice are preserved.