1 2020-02-26 Jakub Jelinek <jakub@redhat.com>
3 PR tree-optimization/93820
4 * gimple-ssa-store-merging.c (check_no_overlap): Change RHS_CODE
5 argument to ALL_INTEGER_CST_P boolean.
6 (imm_store_chain_info::try_coalesce_bswap): Adjust caller.
7 (imm_store_chain_info::coalesce_immediate_stores): Likewise. Handle
8 adjacent INTEGER_CST store into merged_store->only_constants like
11 2020-02-25 Jakub Jelinek <jakub@redhat.com>
14 * config/sh/sh.c (expand_cbranchdi4): Fix comment typo, probablity
16 * cfghooks.c (verify_flow_info): Likewise.
17 * predict.c (combine_predictions_for_bb): Likewise.
18 * bb-reorder.c (connect_better_edge_p): Likewise. Fix comment typo,
19 sucessor -> successor.
20 (find_traces_1_round): Fix comment typo, destinarion -> destination.
21 * omp-expand.c (expand_oacc_for): Fix comment typo, sucessors ->
23 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Fix dump
24 message typo, sucessors -> successors.
26 2020-02-25 Martin Sebor <msebor@redhat.com>
28 * doc/extend.texi (attribute access): Correct an example.
30 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
32 * config/aarch64/aarch64-builtins.c (aarch64_scalar_builtin_types):
34 (aarch64_init_simd_builtin_scalar_types): Register simd_bf.
36 * config/aarch64/iterators.md (VALLDIF): Enable for V4BF and V8BF.
37 (VD): Enable for V4BF.
39 (VQ): Enable for V8BF.
42 (VDBL, Vdbl): Add V4BF.
43 (V_INT_EQUIV, v_int_equiv): Add V4BF and V8BF.
44 * config/aarch64/arm_neon.h (bfloat16x4x2_t): New typedef.
45 (bfloat16x8x2_t): Likewise.
46 (bfloat16x4x3_t): Likewise.
47 (bfloat16x8x3_t): Likewise.
48 (bfloat16x4x4_t): Likewise.
49 (bfloat16x8x4_t): Likewise.
51 (vld1_bf16, vld1_bf16_x2): New.
52 (vld1_bf16_x3, vld1_bf16_x4): New.
53 (vld1q_bf16, vld1q_bf16_x2): New.
54 (vld1q_bf16_x3, vld1q_bf16_x4): New.
55 (vld1_lane_bf16): New.
56 (vld1q_lane_bf16): New.
58 (vld1q_dup_bf16): New.
62 (vld2q_dup_bf16): New.
66 (vld3q_dup_bf16): New.
70 (vld4q_dup_bf16): New.
71 (vst1_bf16, vst1_bf16_x2): New.
72 (vst1_bf16_x3, vst1_bf16_x4): New.
73 (vst1q_bf16, vst1q_bf16_x2): New.
74 (vst1q_bf16_x3, vst1q_bf16_x4): New.
75 (vst1_lane_bf16): New.
76 (vst1q_lane_bf16): New.
84 2020-02-25 Mihail Ionescu <mihail.ionescu@arm.com>
86 * config/aarch64/iterators.md (VDQF_F16) Add V4BF and V8BF.
88 (VALLDI_F16): Likewise.
91 (vswap_width_name): Likewise.
92 (VSWAP_WIDTH): Likewise.
96 * config/aarch64/arm_neon.h (vset_lane_bf16, vsetq_lane_bf16): New.
97 (vget_lane_bf16, vgetq_lane_bf16): New.
99 (vdup_n_bf16, vdupq_n_bf16): New.
100 (vdup_lane_bf16, vdup_laneq_bf16): New.
101 (vdupq_lane_bf16, vdupq_laneq_bf16): New.
102 (vduph_lane_bf16, vduph_laneq_bf16): New.
103 (vreinterpret_bf16_u8, vreinterpretq_bf16_u8): New.
104 (vreinterpret_bf16_u16, vreinterpretq_bf16_u16): New.
105 (vreinterpret_bf16_u32, vreinterpretq_bf16_u32): New.
106 (vreinterpret_bf16_u64, vreinterpretq_bf16_u64): New.
107 (vreinterpret_bf16_s8, vreinterpretq_bf16_s8): New.
108 (vreinterpret_bf16_s16, vreinterpretq_bf16_s16): New.
109 (vreinterpret_bf16_s32, vreinterpretq_bf16_s32): New.
110 (vreinterpret_bf16_s64, vreinterpretq_bf16_s64): New.
111 (vreinterpret_bf16_p8, vreinterpretq_bf16_p8): New.
112 (vreinterpret_bf16_p16, vreinterpretq_bf16_p16): New.
113 (vreinterpret_bf16_p64, vreinterpretq_bf16_p64): New
114 (vreinterpret_bf16_f16, vreinterpretq_bf16_f16): New
115 (vreinterpret_bf16_f32, vreinterpretq_bf16_f32): New.
116 (vreinterpret_bf16_f64, vreinterpretq_bf16_f64): New.
117 (vreinterpretq_bf16_p128): New.
118 (vreinterpret_s8_bf16, vreinterpretq_s8_bf16): New.
119 (vreinterpret_s16_bf16, vreinterpretq_s16_bf16): New.
120 (vreinterpret_s32_bf16, vreinterpretq_s32_bf16): New.
121 (vreinterpret_s64_bf16, vreinterpretq_s64_bf16): New.
122 (vreinterpret_u8_bf16, vreinterpretq_u8_bf16): New.
123 (vreinterpret_u16_bf16, vreinterpretq_u16_bf16): New.
124 (vreinterpret_u32_bf16, vreinterpretq_u32_bf16): New.
125 (vreinterpret_u64_bf16, vreinterpretq_u64_bf16): New.
126 (vreinterpret_p8_bf16, vreinterpretq_p8_bf16): New.
127 (vreinterpret_p16_bf16, vreinterpretq_p16_bf16): New.
128 (vreinterpret_p64_bf16, vreinterpretq_p64_bf16): New.
129 (vreinterpret_f32_bf16, vreinterpretq_f32_bf16): New.
130 (vreinterpret_f64_bf16,vreinterpretq_f64_bf16): New.
131 (vreinterpret_f16_bf16,vreinterpretq_f16_bf16): New.
132 (vreinterpretq_p128_bf16): New.
134 2020-02-25 Dennis Zhang <dennis.zhang@arm.com>
136 * config/arm/arm_neon.h (vbfdot_f32, vbfdotq_f32): New
137 (vbfdot_lane_f32, vbfdotq_laneq_f32): New.
138 (vbfdot_laneq_f32, vbfdotq_lane_f32): New.
139 * config/arm/arm_neon_builtins.def (vbfdot): New entry.
140 (vbfdot_lanev4bf, vbfdot_lanev8bf): Likewise.
141 * config/arm/iterators.md (VSF2BF): New attribute.
142 * config/arm/neon.md (neon_vbfdot<VCVTF:mode>): New entry.
143 (neon_vbfdot_lanev4bf<VCVTF:mode>): Likewise.
144 (neon_vbfdot_lanev8bf<VCVTF:mode>): Likewise.
146 2020-02-25 Christophe Lyon <christophe.lyon@linaro.org>
148 * config/arm/arm.md (required_for_purecode): New attribute.
149 (enabled): Handle required_for_purecode.
150 * config/arm/thumb1.md (thumb1_movsi_insn): Add alternative to
151 work with -mpure-code.
153 2020-02-25 Jakub Jelinek <jakub@redhat.com>
155 PR rtl-optimization/93908
156 * combine.c (find_split_point): For store into ZERO_EXTRACT, and src
159 2019-02-25 Eric Botcazou <ebotcazou@adacore.com>
161 * dwarf2out.c (dwarf2out_size_function): Run in early-DWARF mode.
163 2020-02-25 Roman Zhuykov <zhroma@ispras.ru>
165 * doc/install.texi (--enable-checking): Adjust wording.
167 2020-02-25 Richard Biener <rguenther@suse.de>
169 PR tree-optimization/93868
170 * tree-vect-slp.c (slp_copy_subtree): New function.
171 (vect_attempt_slp_rearrange_stmts): Copy the SLP tree before
172 re-arranging stmts in it.
174 2020-02-25 Jakub Jelinek <jakub@redhat.com>
177 * passes.c (pass_manager::dump_passes): Create a cgraph node for the
178 dummy function and remove it at the end.
181 * config/lm32/lm32.c (lm32_setup_incoming_varargs): Fix comment typo
182 paramter -> parameter.
183 * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Likewise.
184 * ipa-prop.h (struct ipa_agg_replacement_value): Likewise.
186 2020-02-24 Roman Zhuykov <zhroma@ispras.ru>
188 * doc/install.texi (--enable-checking): Properly document current
190 (--enable-stage1-checking): Minor clarification about bootstrap.
192 2020-02-24 David Malcolm <dmalcolm@redhat.com>
195 * doc/invoke.texi (-Wnanalyzer-tainted-array-index): Note that
196 -fanalyzer-checker=taint is also required.
197 (-fanalyzer-checker=): Note that providing this option enables the
198 given checker, and doing so may be required for checkers that are
201 2020-02-24 David Malcolm <dmalcolm@redhat.com>
203 * doc/invoke.texi (-fanalyzer-verbosity=): "2" only shows
204 significant control flow events; add a "3" which shows all
205 control flow events; the old "3" becomes "4".
207 2020-02-24 Jakub Jelinek <jakub@redhat.com>
209 PR tree-optimization/93582
210 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Consider
211 pd.offset and pd.size to be counted in bits rather than bytes, add
212 support for maxsizei that is not a multiple of BITS_PER_UNIT and
213 handle bitfield stores and loads.
214 (vn_reference_lookup_3): Don't call ranges_known_overlap_p with
215 uncomparable quantities - bytes vs. bits. Allow push_partial_def
216 on offsets/sizes that aren't multiple of BITS_PER_UNIT and adjust
217 pd.offset/pd.size to be counted in bits rather than bytes.
218 Formatting fix. Rename shadowed len variable to buflen.
220 2020-02-24 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
221 Kugan Vivekandarajah <kugan.vivekanandarajah@linaro.org>
224 * gcc.c (putenv_COLLECT_AS_OPTIONS): New function.
225 (driver::main): Call putenv_COLLECT_AS_OPTIONS.
226 * opts-common.c (parse_options_from_collect_gcc_options): New function.
227 (prepend_xassembler_to_collect_as_options): Likewise.
228 * opts.h (parse_options_from_collect_gcc_options): Declare prototype.
229 (prepend_xassembler_to_collect_as_options): Likewise.
230 * lto-opts.c (lto_write_options): Stream assembler options
231 in COLLECT_AS_OPTIONS.
232 * lto-wrapper.c (xassembler_options_error): New static variable.
233 (get_options_from_collect_gcc_options): Move parsing options code to
234 parse_options_from_collect_gcc_options and call it.
235 (merge_and_complain): Validate -Xassembler options.
236 (append_compiler_options): Handle OPT_Xassembler.
237 (run_gcc): Append command line -Xassembler options to
239 * doc/invoke.texi: Add documentation about using Xassembler
242 2020-02-24 Kito Cheng <kito.cheng@sifive.com>
244 * config/riscv/riscv.c (riscv_emit_float_compare): Change the code gen
246 (riscv_rtx_costs): Update cost model for LTGT.
248 2020-02-23 Vladimir Makarov <vmakarov@redhat.com>
250 PR rtl-optimization/93564
251 * ira-color.c (struct update_cost_queue_elem): New member start.
252 (queue_update_cost, get_next_update_cost): Add new arg start.
253 (allocnos_conflict_p): New function.
254 (update_costs_from_allocno): Add new arg conflict_cost_update_p.
255 Add checking conflicts with allocnos_conflict_p.
256 (update_costs_from_prefs, restore_costs_from_copies): Adjust
257 update_costs_from_allocno calls.
258 (update_conflict_hard_regno_costs): Add checking conflicts with
259 allocnos_conflict_p. Adjust calls of queue_update_cost and
260 get_next_update_cost.
261 (assign_hard_reg): Adjust calls of queue_update_cost. Add
263 (bucket_allocno_compare_func): Restore previous version.
265 2020-02-21 John David Anglin <danglin@gcc.gnu.org>
267 * gcc/config/pa/pa.c (pa_function_value): Fix check for word and
268 double-word size when handling aggregate return values.
269 * gcc/config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Fix to indicate
270 that homogeneous SFmode and DFmode aggregates are passed and returned
271 in general registers.
273 2020-02-21 Jakub Jelinek <jakub@redhat.com>
276 * opts.c (print_filtered_help): Translate help before appending
277 messages to it rather than after that.
279 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
281 PR rtl-optimization/PR92989
282 * lra-lives.c (process_bb_lives): Restore the original order
283 of the bb liveness update. Call make_hard_regno_dead for each
284 register clobbered at the start of an EH receiver.
286 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
289 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
290 self-recursively generated.
292 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
295 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
298 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
300 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
301 Document new target supports option.
303 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
305 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
306 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
307 * config/arm/iterators.md (MATMUL): New iterator.
308 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
309 (mmla_sfx): New attribute.
310 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
311 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
312 (UNSPEC_MATMUL_US): New.
314 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
316 * config/arm/arm.md: Prevent scalar shifts from being used when big
319 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
320 Richard Biener <rguenther@suse.de>
322 PR tree-optimization/93586
323 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
324 after mismatched array refs; do not sure type size information to
325 recover from unmatched referneces with !flag_strict_aliasing_p.
327 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
329 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
330 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
331 (scatter_store<mode>): Rename to ...
332 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
333 (scatter<mode>_exec): Delete. Move contents ...
334 (mask_scatter_store<mode>): ... here, and rename that to ...
335 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
336 Remove mode conversion.
337 (mask_gather_load<mode>): Rename to ...
338 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
339 Remove mode conversion.
340 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
342 2020-02-21 Martin Jambor <mjambor@suse.cz>
344 PR tree-optimization/93845
345 * tree-sra.c (verify_sra_access_forest): Only test access size of
348 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
350 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
351 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
352 (addv64di3_exec): Likewise.
353 (subv64di3): Likewise.
354 (subv64di3_exec): Likewise.
355 (addv64di3_zext): Likewise.
356 (addv64di3_zext_exec): Likewise.
357 (addv64di3_zext_dup): Likewise.
358 (addv64di3_zext_dup_exec): Likewise.
359 (addv64di3_zext_dup2): Likewise.
360 (addv64di3_zext_dup2_exec): Likewise.
361 (addv64di3_sext_dup2): Likewise.
362 (addv64di3_sext_dup2_exec): Likewise.
363 (<expander>v64di3): Likewise.
364 (<expander>v64di3_exec): Likewise.
365 (*<reduc_op>_dpp_shr_v64di): Likewise.
366 (*plus_carry_dpp_shr_v64di): Likewise.
367 * config/gcn/gcn.md (adddi3): Likewise.
368 (addptrdi3): Likewise.
369 (<expander>di3): Likewise.
371 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
373 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
375 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
377 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
378 support. Use aarch64_emit_mult instead of emitting multiplication
379 instructions directly.
380 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
381 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
383 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
385 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
386 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
387 instead of emitting multiplication instructions directly.
388 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
389 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
390 (@aarch64_frecps<mode>): New expanders.
392 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
394 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
395 on and produce uint64_ts rather than ints.
396 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
397 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
399 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
401 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
402 an unused xmsk register when handling approximate rsqrt.
404 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
406 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
407 flag_finite_math_only condition.
409 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
412 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
413 to destination operand for shufps alternative.
414 (*vec_extractv2si_1): Ditto.
416 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
419 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
422 2020-02-20 Martin Liska <mliska@suse.cz>
425 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
427 2020-02-20 Martin Liska <mliska@suse.cz>
430 * common/config/avr/avr-common.c: Remote trailing "|".
432 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
434 * collect2.c (maybe_run_lto_and_relink): Fix typo in
437 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
439 PR tree-optimization/93767
440 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
441 access-size bias from the offset calculations for negative strides.
443 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
445 * collect2.c (c_file, o_file): Make const again.
446 (ldout,lderrout, dump_ld_file): Remove.
447 (tool_cleanup): Avoid calling not signal-safe functions.
448 (maybe_run_lto_and_relink): Avoid possible signal handler
449 access to unintialzed memory (lto_o_files).
450 (main): Avoid leaking temp files in $TMPDIR.
451 Initialize c_file/o_file with concat, which avoids exposing
452 uninitialized memory to signal handler, which calls unlink(!).
453 Avoid calling maybe_unlink when the main function returns,
454 since the atexit handler is already doing this.
455 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
457 2020-02-19 Martin Jambor <mjambor@suse.cz>
459 PR tree-optimization/93776
460 * tree-sra.c (create_access): Do not create zero size accesses.
461 (get_access_for_expr): Do not search for zero sized accesses.
463 2020-02-19 Martin Jambor <mjambor@suse.cz>
465 PR tree-optimization/93667
466 * tree-sra.c (scalarizable_type_p): Return false if record fields
467 do not follow wach other.
469 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
471 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
472 rather than fmv.x.s/fmv.s.x.
474 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
476 * config/aarch64/aarch64-simd-builtins.def
477 (intrinsic_vec_smult_lo_): New.
478 (intrinsic_vec_umult_lo_): Likewise.
479 (vec_widen_smult_hi_): Likewise.
480 (vec_widen_umult_hi_): Likewise.
481 * config/aarch64/aarch64-simd.md
482 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
483 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
484 (vmull_high_s16): Likewise.
485 (vmull_high_s32): Likewise.
486 (vmull_high_u8): Likewise.
487 (vmull_high_u16): Likewise.
488 (vmull_high_u32): Likewise.
489 (vmull_s8): Likewise.
490 (vmull_s16): Likewise.
491 (vmull_s32): Likewise.
492 (vmull_u8): Likewise.
493 (vmull_u16): Likewise.
494 (vmull_u32): Likewise.
496 2020-02-18 Martin Liska <mliska@suse.cz>
498 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
499 bootstrap by missing removal of invalid sanity check.
501 2020-02-18 Martin Liska <mliska@suse.cz>
504 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
505 Always compare LHS of gimple_assign.
507 2020-02-18 Martin Liska <mliska@suse.cz>
510 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
511 and return type of functions.
512 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
513 Drop MALLOC attribute for void functions.
514 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
515 malloc_state for a new VOID clone.
517 2020-02-18 Martin Liska <mliska@suse.cz>
520 * common.opt: Add -fprofile-reproducibility.
521 * doc/invoke.texi: Document it.
522 * value-prof.c (dump_histogram_value):
523 Document and support behavior for counters[0]
524 being a negative value.
525 (get_nth_most_common_value): Handle negative
526 counters[0] in respect to flag_profile_reproducible.
528 2020-02-18 Jakub Jelinek <jakub@redhat.com>
531 * cgraph.c (verify_speculative_call): Use speculative_id instead of
532 speculative_uid in messages. Remove trailing whitespace from error
533 message. Use num_speculative_call_targets instead of
534 num_speculative_targets in a message.
535 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
536 edge messages and stmt instead of cal_stmt in reference message.
538 PR tree-optimization/93780
539 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
540 before calling build_vector_type.
541 (execute_update_addresses_taken): Likewise.
544 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
545 typo, functoin -> function.
546 * tree.c (free_lang_data_in_decl): Fix comment typo,
547 functoin -> function.
548 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
550 2020-02-17 David Malcolm <dmalcolm@redhat.com>
552 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
554 (print_option_information): Don't call get_option_url if URLs
557 2020-02-17 Alexandre Oliva <oliva@adacore.com>
559 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
560 handling of register_common-less targets.
562 2020-02-17 Martin Liska <mliska@suse.cz>
565 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
567 2020-02-17 Martin Liska <mliska@suse.cz>
570 * config/rs6000/rs6000.c (rs6000_option_override_internal):
573 2020-02-17 Martin Liska <mliska@suse.cz>
576 * config/rx/elf.opt: Fix typo.
578 2020-02-17 Richard Biener <rguenther@suse.de>
581 * opts-global.c (print_ignored_options): Use inform and
584 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
587 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
589 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
592 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
593 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
595 2020-02-15 Jason Merrill <jason@redhat.com>
597 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
599 2020-02-15 Jakub Jelinek <jakub@redhat.com>
601 PR tree-optimization/93744
602 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
603 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
604 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
605 sure @2 in the first and @1 in the other patterns has no side-effects.
607 2020-02-15 David Malcolm <dmalcolm@redhat.com>
608 Bernd Edlinger <bernd.edlinger@hotmail.de>
612 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
613 * configure.ac (--with-diagnostics-urls): New configuration
614 option, based on --with-diagnostics-color.
615 (DIAGNOSTICS_URLS_DEFAULT): New define.
616 * config.h: Regenerate.
617 * configure: Regenerate.
618 * diagnostic.c (diagnostic_urls_init): Handle -1 for
619 DIAGNOSTICS_URLS_DEFAULT from configure-time
620 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
621 and TERM_URLS environment variable.
622 * diagnostic-url.h (diagnostic_url_format): New enum type.
623 (diagnostic_urls_enabled_p): rename to...
624 (determine_url_format): ... this, and change return type.
625 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
626 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
627 the linux console, and mingw.
628 (diagnostic_urls_enabled_p): rename to...
629 (determine_url_format): ... this, and adjust.
630 * pretty-print.h (pretty_printer::show_urls): rename to...
631 (pretty_printer::url_format): ... this, and change to enum.
632 * pretty-print.c (pretty_printer::pretty_printer,
633 pp_begin_url, pp_end_url, test_urls): Adjust.
634 * doc/install.texi (--with-diagnostics-urls): Document the new
635 configuration option.
636 (--with-diagnostics-color): Document the existing interaction
637 with GCC_COLORS better.
638 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
639 vindex reference. Update description of defaults based on the above.
640 (-fdiagnostics-color): Update description of how -fdiagnostics-color
641 interacts with GCC_COLORS.
643 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
646 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
647 conjunction with TARGET_GNU_TLS in early return.
649 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
651 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
652 the mode is not wider than UNITS_PER_WORD.
654 2020-02-14 Martin Jambor <mjambor@suse.cz>
656 PR tree-optimization/93516
657 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
658 access of the same type as the parent.
659 (propagate_subaccesses_from_lhs): Likewise.
661 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
664 * config/i386/avx512vbmi2intrin.h
665 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
666 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
667 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
668 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
669 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
670 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
671 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
672 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
673 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
674 of lacking a closing parenthesis.
675 * config/i386/avx512vbmi2vlintrin.h
676 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
677 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
678 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
679 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
680 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
681 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
682 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
683 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
684 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
685 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
686 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
687 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
688 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
689 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
690 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
691 _mm_shldi_epi32, _mm_mask_shldi_epi32,
692 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
693 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
695 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
698 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
699 the target function entry.
701 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
703 * common/config/arc/arc-common.c (arc_option_optimization_table):
704 Disable if-conversion step when optimized for size.
706 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
708 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
709 R12-R15 are always in ARCOMPACT16_REGS register class.
710 * config/arc/arc.opt (mq-class): Deprecate.
711 * config/arc/constraint.md ("q"): Remove dependency on mq-class
713 * doc/invoke.texi (mq-class): Update text.
714 * common/config/arc/arc-common.c (arc_option_optimization_table):
717 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
719 * config/arc/arc.c (arc_insn_cost): New function.
720 (TARGET_INSN_COST): Define.
721 * config/arc/arc.md (cost): New attribute.
722 (add_n): Use arc_nonmemory_operand.
723 (ashlsi3_insn): Likewise, also update constraints.
724 (ashrsi3_insn): Likewise.
726 (add_shift): Likewise.
727 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
729 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
731 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
733 (umulsidi_600): Likewise.
735 2020-02-13 Jakub Jelinek <jakub@redhat.com>
738 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
739 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
740 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
741 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
742 pass __A to the builtin followed by __W instead of __A followed by
744 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
745 _mm512_mask_popcnt_epi64): Likewise.
746 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
747 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
748 _mm256_mask_popcnt_epi64): Likewise.
750 PR tree-optimization/93582
751 * fold-const.h (shift_bytes_in_array_left,
752 shift_bytes_in_array_right): Declare.
753 * fold-const.c (shift_bytes_in_array_left,
754 shift_bytes_in_array_right): New function, moved from
755 gimple-ssa-store-merging.c, no longer static.
756 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
757 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
758 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
759 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
760 shift_bytes_in_array.
761 (verify_shift_bytes_in_array): Rename to ...
762 (verify_shift_bytes_in_array_left): ... this. Use
763 shift_bytes_in_array_left instead of shift_bytes_in_array.
764 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
765 instead of verify_shift_bytes_in_array.
766 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
767 / native_interpret_expr where the store covers all needed bits,
768 punt on PDP-endian, otherwise allow all involved offsets and sizes
769 not to be byte-aligned.
772 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
773 use const_0_to_255_operand predicate instead of immediate_operand.
774 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
775 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
776 vgf2p8affineinvqb_<mode><mask_name>,
777 vgf2p8affineqb_<mode><mask_name>): Drop mode from
778 const_0_to_255_operand predicated operands.
780 2020-02-12 Jeff Law <law@redhat.com>
782 * config/h8300/h8300.md (comparison shortening peepholes): Use
783 a mode iterator to merge the HImode and SImode peepholes.
785 2020-02-12 Jakub Jelinek <jakub@redhat.com>
788 * real.c (is_even): Make static. Function comment fix.
789 (is_halfway_below): Make static, don't assert R is not inf/nan,
790 instead return false for those. Small formatting fixes.
792 2020-02-12 Martin Sebor <msebor@redhat.com>
795 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
796 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
797 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
798 (strlen_check_and_optimize_call): Adjust callee name.
800 2020-02-12 Jeff Law <law@redhat.com>
802 * config/h8300/h8300.md (comparison shortening peepholes): Drop
803 (and (xor)) variant. Combine other two into single peephole.
805 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
807 PR rtl-optimization/93565
808 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
810 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
812 * config/aarch64/aarch64-simd.md
813 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
814 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
815 generating separate ADDV and zero_extend patterns.
816 * config/aarch64/iterators.md (VDQV_E): New iterator.
818 2020-02-12 Jeff Law <law@redhat.com>
820 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
821 expanders, splits, etc.
822 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
823 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
824 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
825 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
826 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
827 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
829 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
831 2020-02-12 Jakub Jelinek <jakub@redhat.com>
834 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
835 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
836 TARGET_AVX512DQ from condition.
837 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
838 instead of <mask_mode512bit_condition> in condition. If
839 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
841 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
844 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
847 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
849 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
851 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
852 where strlen is more legible.
853 (rs6000_builtin_vectorized_libmass): Ditto.
854 (rs6000_print_options_internal): Ditto.
856 2020-02-11 Martin Sebor <msebor@redhat.com>
858 PR tree-optimization/93683
859 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
861 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
863 * config/rs6000/predicates.md (cint34_operand): Rename the
864 -mprefixed-addr option to be -mprefixed.
865 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
866 the -mprefixed-addr option to be -mprefixed.
867 (OTHER_FUTURE_MASKS): Likewise.
868 (POWERPC_MASKS): Likewise.
869 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
870 the -mprefixed-addr option to be -mprefixed. Change error
871 messages to refer to -mprefixed.
872 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
874 (rs6000_legitimate_offset_address_p): Likewise.
875 (rs6000_mode_dependent_address): Likewise.
876 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
877 "-mprefixed" for target attributes and pragmas.
878 (address_to_insn_form): Rename the -mprefixed-addr option to be
880 (rs6000_adjust_insn_length): Likewise.
881 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
882 -mprefixed-addr option to be -mprefixed.
883 (ASM_OUTPUT_OPCODE): Likewise.
884 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
885 -mprefixed-addr option to be -mprefixed.
886 * config/rs6000/rs6000.opt (-mprefixed): Rename the
887 -mprefixed-addr option to be prefixed. Change the option from
888 being undocumented to being documented.
889 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
890 -mprefixed option. Update the -mpcrel documentation to mention
893 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
895 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
896 including FIRST_PSEUDO_REGISTER - 1.
897 * ira-color.c (print_hard_reg_set): Ditto.
899 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
901 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
902 (USTERNOP_QUALIFIERS): New define.
903 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
904 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
905 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
906 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
907 * config/arm/arm_neon.h (vusdot_s32): New.
908 (vusdot_lane_s32): New.
909 (vusdotq_lane_s32): New.
910 (vsudot_lane_s32): New.
911 (vsudotq_lane_s32): New.
912 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
913 * config/arm/iterators.md (DOTPROD_I8MM): New.
914 (sup, opsuffix): Add <us/su>.
915 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
916 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
918 2020-02-11 Richard Biener <rguenther@suse.de>
920 PR tree-optimization/93661
921 PR tree-optimization/93662
922 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
924 * tree-sra.c (get_access_for_expr): Likewise.
926 2020-02-10 Jakub Jelinek <jakub@redhat.com>
929 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
930 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
931 Change condition from TARGET_AVX2 to TARGET_AVX.
933 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
936 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
939 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
941 Try to generate zero-based comparisons.
942 * config/cris/cris.c (cris_reduce_compare): New function.
943 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
944 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
945 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
947 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
950 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
951 in Thumb state and also as a destination in Arm state. Add T16
954 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
956 * md.texi (Define Subst): Match closing paren in example.
958 2020-02-10 Jakub Jelinek <jakub@redhat.com>
962 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
963 arguments of strncmp.
965 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
968 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
969 but different source value.
970 (adjust_callers_for_value_intersection): New function.
971 (gather_edges_for_value): Adjust order of callers to let a
972 non-self-recursive caller be the first element.
973 (self_recursive_pass_through_p): Add a new parameter "simple", and
974 check generalized self-recursive pass-through jump function.
975 (self_recursive_agg_pass_through_p): Likewise.
976 (find_more_scalar_values_for_callers_subset): Compute value from
977 pass-through jump function for self-recursive.
978 (intersect_with_plats): Cleanup previous implementation code for value
979 itersection with self-recursive call edge.
980 (intersect_with_agg_replacements): Likewise.
981 (intersect_aggregates_with_edge): Deduce value from pass-through jump
982 function for self-recursive call edge. Cleanup previous implementation
983 code for value intersection with self-recursive call edge.
984 (decide_whether_version_node): Remove dead callers and adjust order
985 to let a non-self-recursive caller be the first element.
987 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
989 * recog.c: Move pass_split_before_sched2 code in front of
990 pass_split_before_regstack.
991 (pass_data_split_before_sched2): Rename pass to split3 from split4.
992 (pass_data_split_before_regstack): Rename pass to split4 from split3.
993 (rest_of_handle_split_before_sched2): Remove.
994 (pass_split_before_sched2::execute): Unconditionally call
996 (enable_split_before_sched2): New function.
997 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
998 (pass_split_before_regstack::gate): Ditto.
999 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
1000 Update name check for renamed split4 pass.
1001 * config/sh/sh.c (register_sh_passes): Update pass insertion
1002 point for renamed split4 pass.
1004 2020-02-09 Jakub Jelinek <jakub@redhat.com>
1006 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
1007 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
1008 copying them around between host and target.
1010 2020-02-08 Andrew Pinski <apinski@marvell.com>
1013 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
1014 STRICT_ALIGNMENT also.
1016 2020-02-08 Jim Wilson <jimw@sifive.com>
1019 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
1021 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
1022 Jakub Jelinek <jakub@redhat.com>
1025 * config/i386/i386.h (CALL_USED_REGISTERS): Make
1026 xmm16-xmm31 call-used even in 64-bit ms-abi.
1028 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
1030 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
1031 (simd_ummla, simd_usmmla): Likewise.
1032 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
1033 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
1034 (vusmmlaq_s32): New.
1036 2020-02-07 Richard Biener <rguenther@suse.de>
1039 * tree-inline.c (fold_marked_statements): Do a PRE walk,
1040 skipping unreachable regions.
1041 (optimize_inline_calls): Skip folding stmts when we didn't
1044 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
1047 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
1048 Don't return aggregates with only SFmode and DFmode in SSE
1050 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
1052 2020-02-07 Jakub Jelinek <jakub@redhat.com>
1055 * config/rs6000/rs6000-logue.c
1056 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
1057 if it fails, move rs into end_addr and retry. Add
1058 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
1059 the insn pattern doesn't describe well what exactly happens to
1063 * config/i386/predicates.md (avx_identity_operand): Remove.
1064 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
1065 (avx_<castmode><avxsizesuffix>_<castmode>,
1066 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
1067 a VEC_CONCAT of the operand and UNSPEC_CAST.
1068 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
1069 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
1073 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
1074 recog_data.insn if distance_non_agu_define changed it.
1076 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
1079 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
1080 we only had X-FORM (reg+reg) addressing for vectors. Also before
1081 ISA 3.0, we only had X-FORM addressing for scalars in the
1082 traditional Altivec registers.
1084 2020-02-06 <zhongyunde@huawei.com>
1085 Vladimir Makarov <vmakarov@redhat.com>
1087 PR rtl-optimization/93561
1088 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
1089 hard register range.
1091 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1093 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
1096 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
1098 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
1099 where the low and the high 32 bits are equal to each other specially,
1100 with an rldimi instruction.
1102 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1104 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
1106 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
1108 * config/arm/arm-tables.opt: Regenerate.
1110 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1113 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
1114 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
1115 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
1117 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
1119 PR rtl-optimization/87763
1120 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
1122 2020-02-06 Delia Burduv <delia.burduv@arm.com>
1124 * config/aarch64/aarch64-simd-builtins.def
1125 (bfmlaq): New built-in function.
1126 (bfmlalb): New built-in function.
1127 (bfmlalt): New built-in function.
1128 (bfmlalb_lane): New built-in function.
1129 (bfmlalt_lane): New built-in function.
1130 * config/aarch64/aarch64-simd.md
1131 (aarch64_bfmmlaqv4sf): New pattern.
1132 (aarch64_bfmlal<bt>v4sf): New pattern.
1133 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
1134 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
1135 (vbfmlalbq_f32): New intrinsic.
1136 (vbfmlaltq_f32): New intrinsic.
1137 (vbfmlalbq_lane_f32): New intrinsic.
1138 (vbfmlaltq_lane_f32): New intrinsic.
1139 (vbfmlalbq_laneq_f32): New intrinsic.
1140 (vbfmlaltq_laneq_f32): New intrinsic.
1141 * config/aarch64/iterators.md (BF_MLA): New int iterator.
1142 (bt): New int attribute.
1144 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
1146 * config/i386/i386.md (*pushtf): Emit "#" instead of
1147 calling gcc_unreachable in insn output.
1150 (*pushsf_rex64): Ditto for alternatives other than 1.
1151 (*pushsf): Ditto for alternatives other than 1.
1153 2020-02-06 Martin Liska <mliska@suse.cz>
1155 PR gcov-profile/91971
1156 PR gcov-profile/93466
1157 * coverage.c (coverage_init): Revert mangling of
1158 path into filename. It can lead to huge filename length.
1159 Creation of subfolders seem more natural.
1161 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1164 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
1165 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
1166 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
1168 2020-02-06 Jakub Jelinek <jakub@redhat.com>
1171 * config/i386/predicates.md (avx_identity_operand): New predicate.
1172 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
1173 define_insn_and_split.
1176 * omp-low.c (use_pointer_for_field): For nested constructs, also
1177 look for map clauses on target construct.
1178 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
1179 taskreg_nesting_level.
1182 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
1183 shared clause, call omp_notice_variable on outer context if any.
1185 2020-02-05 Jason Merrill <jason@redhat.com>
1188 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
1189 non-zero address even if weak and not yet defined.
1191 2020-02-05 Martin Sebor <msebor@redhat.com>
1193 PR tree-optimization/92765
1194 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
1195 * tree-ssa-strlen.c (compute_string_length): Remove.
1196 (determine_min_objsize): Remove.
1197 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
1198 Avoid using type size as the upper bound on string length.
1199 (handle_builtin_string_cmp): Add an argument. Adjust.
1200 (strlen_check_and_optimize_call): Pass additional argument to
1201 handle_builtin_string_cmp.
1203 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
1205 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
1206 (*pushdi2_rex64 peephole2): Unconditionally split after
1208 (*ashl<mode>3_doubleword): Ditto.
1209 (*<shift_insn><mode>3_doubleword): Ditto.
1211 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
1214 * config/rs6000/rs6000.c (get_vector_offset): Fix
1216 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
1218 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
1220 2020-02-05 David Malcolm <dmalcolm@redhat.com>
1223 (Special Functions for Debugging the Analyzer): Update description
1224 of __analyzer_dump_exploded_nodes.
1226 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1229 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
1230 include sets and not clobbers in the vzeroupper pattern.
1231 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
1232 the parallel has 17 (64-bit) or 9 (32-bit) elts.
1233 (*avx_vzeroupper_1): New define_insn_and_split.
1236 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
1237 don't run when !optimize.
1238 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
1241 2020-02-05 Richard Biener <rguenther@suse.de>
1244 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
1245 checks before matching calls.
1247 2020-02-05 Jakub Jelinek <jakub@redhat.com>
1249 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
1250 function comment typo.
1253 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
1254 simd_clone_create failed when i == 0, adjust clone->nargs by
1257 2020-02-05 Martin Liska <mliska@suse.cz>
1260 * doc/invoke.texi: Document that one should
1261 not combine ASLR and -fpch.
1263 2020-02-04 Richard Biener <rguenther@suse.de>
1265 PR tree-optimization/93538
1266 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
1268 2020-02-04 Richard Biener <rguenther@suse.de>
1270 PR tree-optimization/91123
1271 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1272 (vn_walk_cb_data::last_vuse): New member.
1273 (vn_walk_cb_data::saved_operands): Likewsie.
1274 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1275 (vn_walk_cb_data::push_partial_def): Use finish.
1276 (vn_reference_lookup_2): Update last_vuse and use finish if
1277 we've saved operands.
1278 (vn_reference_lookup_3): Use finish and update calls to
1279 push_partial_defs everywhere. When translating through
1280 memcpy or aggregate copies save off operands and alias-set.
1281 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1282 operation for redundant store removal.
1284 2020-02-04 Richard Biener <rguenther@suse.de>
1286 PR tree-optimization/92819
1287 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1288 generating more stmts than before.
1290 2020-02-04 Martin Liska <mliska@suse.cz>
1292 * config/arm/arm.c (arm_gen_far_branch): Move the function
1293 outside of selftests.
1295 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1297 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1298 function to adjust PC-relative vector addresses.
1299 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1300 handle vectors with PC-relative addresses.
1302 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1304 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1306 (hard_reg_and_mode_to_addr_mask): Delete.
1307 (rs6000_adjust_vec_address): If the original vector address
1308 was REG+REG or REG+OFFSET and the element is not zero, do the add
1309 of the elements in the original address before adding the offset
1310 for the vector element. Use address_to_insn_form to validate the
1311 address using the register being loaded, rather than guessing
1312 whether the address is a DS-FORM or DQ-FORM address.
1314 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1316 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1317 to calculate the offset in memory from the start of a vector of a
1318 particular element. Add code to keep the element number in
1319 bounds if the element number is variable.
1320 (rs6000_adjust_vec_address): Move calculation of offset of the
1321 vector element to get_vector_offset.
1322 (rs6000_split_vec_extract_var): Do not do the initial AND of
1323 element here, move the code to get_vector_offset.
1325 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1327 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1330 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1332 * config/rs6000/constraints.md: Improve documentation.
1334 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1337 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1338 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1340 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1342 * config.gcc: Remove "carrizo" support.
1343 * config/gcn/gcn-opts.h (processor_type): Likewise.
1344 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1345 * config/gcn/gcn.opt (gpu_type): Likewise.
1346 * config/gcn/t-omp-device: Likewise.
1348 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1351 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1352 * config/arm/arm.c (arm_gen_far_branch): New function
1354 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1356 2020-02-03 Julian Brown <julian@codesourcery.com>
1357 Tobias Burnus <tobias@codesourcery.com>
1359 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1361 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1364 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1365 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1368 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1370 PR rtl-optimization/91333
1371 * ira-color.c (struct allocno_color_data): Add member
1373 (init_allocno_threads): Set the member up.
1374 (bucket_allocno_compare_func): Add compare hard reg
1377 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1379 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1381 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1382 * config.in: Regenerated.
1383 * configure: Regenerated.
1384 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1385 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1386 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1388 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1390 * configure: Regenerate.
1392 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1394 PR rtl-optimization/91333
1395 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1396 reg preferences comparison up.
1398 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1400 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1401 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1402 aarch64-sve-builtins-base.h.
1403 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1404 aarch64-sve-builtins-base.cc.
1405 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1406 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1408 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1409 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1410 (svcvtnt): New functions.
1411 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1412 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1413 (svcvtnt): New functions.
1414 (svcvt): Add a form that converts f32 to bf16.
1415 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1416 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1418 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1419 Treat B as bfloat16_t.
1420 (ternary_bfloat_lane_base): New class.
1421 (ternary_bfloat_def): Likewise.
1422 (ternary_bfloat): New shape.
1423 (ternary_bfloat_lane_def): New class.
1424 (ternary_bfloat_lane): New shape.
1425 (ternary_bfloat_lanex2_def): New class.
1426 (ternary_bfloat_lanex2): New shape.
1427 (ternary_bfloat_opt_n_def): New class.
1428 (ternary_bfloat_opt_n): New shape.
1429 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1430 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1431 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1432 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1433 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1434 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1435 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1436 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1437 the pattern off the narrow mode instead of the wider one.
1438 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1439 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1440 (sve_fp_op): Handle them.
1441 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1442 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1444 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1446 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1447 * config/aarch64/aarch64-modes.def (BF): Move definition before
1448 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1449 (SVE_MODES): Handle BF modes.
1450 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1452 (aarch64_full_sve_mode): Likewise.
1453 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1455 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1456 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1457 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1458 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1460 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1462 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1463 (TYPES_all_data): Add bf16.
1464 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1465 (register_tuple_type): Increase buffer size.
1466 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1467 (bf16): New type suffix.
1468 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1469 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1470 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1471 Change type from all_data to all_arith.
1472 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1475 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1476 Matthew Malcomson <matthew.malcomson@arm.com>
1477 Richard Sandiford <richard.sandiford@arm.com>
1479 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1480 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1481 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1482 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1483 __ARM_FEATURE_MATMUL_FP64.
1484 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1485 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1486 be disabled at the same time.
1487 (f32mm): New extension.
1488 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1489 (AARCH64_FL_F64MM): Bump to the next bit up.
1490 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1491 (TARGET_SVE_F64MM): New macros.
1492 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1493 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1494 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1495 (UNSPEC_ZIP2Q): New unspeccs.
1496 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1497 (optab, sur, perm_insn): Handle the new unspecs.
1498 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1499 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1500 TARGET_SVE_F64MM instead of separate tests.
1501 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1502 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1503 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1504 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1505 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1506 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1507 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1508 (TYPES_s_signed): New macro.
1509 (TYPES_s_integer): Use it.
1510 (TYPES_d_float): New macro.
1511 (TYPES_d_data): Use it.
1512 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1513 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1514 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1515 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1516 (svmmla): New shape.
1517 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1518 template parameters.
1519 (ternary_resize2_lane_base): Likewise.
1520 (ternary_resize2_base): New class.
1521 (ternary_qq_lane_base): Likewise.
1522 (ternary_intq_uintq_lane_def): Likewise.
1523 (ternary_intq_uintq_lane): New shape.
1524 (ternary_intq_uintq_opt_n_def): New class
1525 (ternary_intq_uintq_opt_n): New shape.
1526 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1527 (ternary_uintq_intq_def): New class.
1528 (ternary_uintq_intq): New shape.
1529 (ternary_uintq_intq_lane_def): New class.
1530 (ternary_uintq_intq_lane): New shape.
1531 (ternary_uintq_intq_opt_n_def): New class.
1532 (ternary_uintq_intq_opt_n): New shape.
1533 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1534 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1535 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1536 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1538 (svdotprod_lane_impl): ...this new class.
1539 (svmmla_impl, svusdot_impl): New classes.
1540 (svdot_lane): Update to use svdotprod_lane_impl.
1541 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1542 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1544 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1545 function, with no types defined.
1546 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1547 AARCH64_FL_I8MM functions.
1548 (svmmla): New AARCH64_FL_F32MM function.
1549 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1550 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1551 AARCH64_FL_F64MM function.
1552 (REQUIRED_EXTENSIONS):
1554 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1556 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1559 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1561 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1562 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1563 (*movti_internal): Do not check for
1564 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1565 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1566 just after check for TARGET_AVX.
1567 (*movdf_internal): Ditto.
1568 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1569 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1570 * config/i386/sse.md (mov<mode>_internal): Only check
1571 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1572 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1573 (<sse>_andnot<mode>3<mask_name>): Move check for
1574 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1575 (<code><mode>3<mask_name>): Ditto.
1576 (*andnot<mode>3): Ditto.
1577 (*andnottf3): Ditto.
1578 (*<code><mode>3): Ditto.
1579 (*<code>tf3): Ditto.
1580 (*andnot<VI:mode>3): Remove
1581 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1582 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1583 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1584 (sse4_1_blendv<ssemodesuffix>): Ditto.
1585 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1586 Explain that tune applies to 128bit instructions only.
1588 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1590 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1591 to definition of hsa_kernel_description. Parse assembly to find SGPR
1592 and VGPR count of kernel and store in hsa_kernel_description.
1594 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1596 PR rtl-optimization/91838
1597 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1598 to truncate if allowed or reject combination.
1600 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1602 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1603 (find_inv_vars_cb): Likewise.
1605 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1607 * calls.c (special_function_p): Split out the check for DECL_NAME
1608 being non-NULL and fndecl being extern at file scope into a
1609 new maybe_special_function_p and call it. Drop check for fndecl
1610 being non-NULL that was after a usage of DECL_NAME (fndecl).
1611 * tree.h (maybe_special_function_p): New inline function.
1613 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1615 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1616 (mask_gather_load<mode>): ... here, and zero-initialize the
1618 (maskload<mode>di): Zero-initialize the destination.
1621 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1624 * doc/analyzer.texi (Limitations): Note that constraints on
1625 floating-point values are currently ignored.
1627 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1630 * symtab.c (symtab_node::noninterposable_alias): If localalias
1631 already exists, but is not usable, append numbers after it until
1632 a unique name is found. Formatting fix.
1635 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1638 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1640 * config/gcn/gcn.c (print_operand): Handle LTGT.
1641 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1643 2020-01-30 Richard Biener <rguenther@suse.de>
1645 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1646 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1648 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1650 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1651 without a DECL in .data.rel.ro.local.
1653 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1656 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1660 * config/i386/sse.md
1661 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1662 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1663 any_extend code iterator instead of always zero_extend.
1664 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1665 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1666 Use any_extend code iterator instead of always zero_extend.
1667 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1668 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1669 Use any_extend code iterator instead of always zero_extend.
1670 (*sse2_pmovmskb_ext): New define_insn.
1671 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1674 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1675 (*popcountsi2_zext_falsedep): New define_insn.
1677 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1679 * config.in: Regenerated.
1680 * configure: Regenerated.
1682 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1685 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1686 LLVM's assembler changed the default in version 9.
1688 2020-01-24 Jeff Law <law@redhat.com>
1690 PR tree-optimization/89689
1691 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1693 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1697 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1699 PR rtl-optimization/87763
1700 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1701 simplification to handle subregs as well as bare regs.
1702 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1704 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1707 * ira.c (ira): Revert use of simplified LRA algorithm.
1709 2020-01-29 Martin Jambor <mjambor@suse.cz>
1711 PR tree-optimization/92706
1712 * tree-sra.c (struct access): Fields first_link, last_link,
1713 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1714 next_rhs_queued and grp_rhs_queued respectively, new fields
1715 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1716 (struct assign_link): Field next renamed to next_rhs, new field
1717 next_lhs. Updated comment.
1718 (work_queue_head): Renamed to rhs_work_queue_head.
1719 (lhs_work_queue_head): New variable.
1720 (add_link_to_lhs): New function.
1721 (relink_to_new_repr): Also relink LHS lists.
1722 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1723 (add_access_to_lhs_work_queue): New function.
1724 (pop_access_from_work_queue): Renamed to
1725 pop_access_from_rhs_work_queue.
1726 (pop_access_from_lhs_work_queue): New function.
1727 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1729 (child_would_conflict_in_lacc): Renamed to
1730 child_would_conflict_in_acc. Adjusted parameter names.
1731 (create_artificial_child_access): New parameter set_grp_read, use it.
1732 (subtree_mark_written_and_enqueue): Renamed to
1733 subtree_mark_written_and_rhs_enqueue.
1734 (propagate_subaccesses_across_link): Renamed to
1735 propagate_subaccesses_from_rhs.
1736 (propagate_subaccesses_from_lhs): New function.
1737 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1740 2020-01-29 Martin Jambor <mjambor@suse.cz>
1742 PR tree-optimization/92706
1743 * tree-sra.c (struct access): Adjust comment of
1744 grp_total_scalarization.
1745 (find_access_in_subtree): Look for single children spanning an entire
1747 (scalarizable_type_p): Allow register accesses, adjust callers.
1748 (completely_scalarize): Remove function.
1749 (scalarize_elem): Likewise.
1750 (create_total_scalarization_access): Likewise.
1751 (sort_and_splice_var_accesses): Do not track total scalarization
1753 (analyze_access_subtree): New parameter totally, adjust to new meaning
1754 of grp_total_scalarization.
1755 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1756 (can_totally_scalarize_forest_p): New function.
1757 (create_total_scalarization_access): Likewise.
1758 (create_total_access_and_reshape): Likewise.
1759 (total_should_skip_creating_access): Likewise.
1760 (totally_scalarize_subtree): Likewise.
1761 (analyze_all_variable_accesses): Perform total scalarization after
1762 subaccess propagation using the new functions above.
1763 (initialize_constant_pool_replacements): Output initializers by
1764 traversing the access tree.
1766 2020-01-29 Martin Jambor <mjambor@suse.cz>
1768 * tree-sra.c (verify_sra_access_forest): New function.
1769 (verify_all_sra_access_forests): Likewise.
1770 (create_artificial_child_access): Set parent.
1771 (analyze_all_variable_accesses): Call the verifier.
1773 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1775 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1776 if called on indirect edge.
1777 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1778 speculative call if needed.
1780 2020-01-29 Richard Biener <rguenther@suse.de>
1782 PR tree-optimization/93428
1783 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1784 permutation when the load node is created.
1785 (vect_analyze_slp_instance): Re-use it here.
1787 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1789 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1791 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1793 PR rtl-optimization/93272
1794 * ira-lives.c (process_out_of_region_eh_regs): New function.
1795 (process_bb_node_lives): Call it.
1797 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1799 * coverage.c (read_counts_file): Make error message lowercase.
1801 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1803 * profile-count.c (profile_quality_display_names): Fix ordering.
1805 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1808 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1809 hash only when edge is first within the sequence.
1810 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1811 (symbol_table::create_edge): Do not set target_prob.
1812 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1814 (cgraph_edge::make_speculative): Drop target_prob parameter.
1815 (cgraph_edge::speculative_call_info): Remove.
1816 (cgraph_edge::first_speculative_call_target): New member function.
1817 (update_call_stmt_hash_for_removing_direct_edge): New function.
1818 (cgraph_edge::resolve_speculation): Rewrite to new API.
1819 (cgraph_edge::speculative_call_for_target): New member function.
1820 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1821 multiple speculation targets.
1822 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1824 (verify_speculative_call): Verify that targets form an interval.
1825 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1826 (cgraph_edge::first_speculative_call_target): New member function.
1827 (cgraph_edge::next_speculative_call_target): New member function.
1828 (cgraph_edge::speculative_call_target_ref): New member function.
1829 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1830 (cgraph_edge): Remove target_prob.
1831 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1832 Fix handling of speculative calls.
1833 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1834 * ipa-fnsummary.c (analyze_function_body): Likewise.
1835 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1836 * ipa-profile.c (dump_histogram): Fix formating.
1837 (ipa_profile_generate_summary): Watch for overflows.
1838 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1839 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1840 (update_indirect_edges_after_inlining): Update to new API.
1841 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1843 * profile-count.h: (profile_probability::adjusted): New.
1844 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1845 updating of profile.
1846 * value-prof.c (gimple_ic_transform): Rename to ...
1847 (dump_ic_profile): ... this one; update dumping.
1848 (stream_in_histogram_value): Fix formating.
1849 (gimple_value_profile_transformations): Update.
1851 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1854 * config/i386/i386.md (*movoi_internal_avx): Remove
1855 TARGET_SSE_TYPELESS_STORES check.
1856 (*movti_internal): Prefer TARGET_AVX over
1857 TARGET_SSE_TYPELESS_STORES.
1858 (*movtf_internal): Likewise.
1859 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1860 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1861 from TARGET_SSE_TYPELESS_STORES.
1863 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1865 * diagnostic-core.h (warning_at): Rename overload to...
1866 (warning_meta): ...this.
1867 (emit_diagnostic_valist): Delete decl of overload taking
1868 diagnostic_metadata.
1869 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1870 (warning_at): Rename overload taking diagnostic_metadata to...
1871 (warning_meta): ...this.
1873 2020-01-28 Richard Biener <rguenther@suse.de>
1875 PR tree-optimization/93439
1876 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1877 * tree-cfg.c (move_sese_region_to_fn): ... here.
1878 (verify_types_in_gimple_reference): Verify used cliques are
1881 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1884 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1885 argument of a pointer to struct gcc_options and pass it to
1886 parse_mtune_ctrl_str.
1887 (ix86_function_specific_restore): Pass opts to
1888 set_ix86_tune_features.
1889 (ix86_option_override_internal): Likewise.
1890 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1891 gcc_options and use it for x_ix86_tune_ctrl_string.
1893 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1895 PR rtl-optimization/87763
1896 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1897 simplification to handle subregs as well as bare regs.
1898 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1900 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1902 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1903 for reduction chains that (now) include a call.
1905 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1907 PR tree-optimization/92822
1908 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1909 out the don't-care elements of a vector whose significant elements
1910 are duplicates, make the don't-care elements duplicates too.
1912 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1914 PR tree-optimization/93434
1915 * tree-predcom.c (split_data_refs_to_components): Record which
1916 components have had aliasing loads removed. Prevent store-store
1917 commoning for all such components.
1919 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1922 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1923 -1 or is_vshift is true, use new_vector with number of elts npatterns
1924 rather than new_unary_operation.
1926 PR tree-optimization/93454
1927 * gimple-fold.c (fold_array_ctor_reference): Perform
1928 elt_size.to_uhwi () just once, instead of calling it in every
1929 iteration. Punt if that value is above size of the temporary
1930 buffer. Decrease third native_encode_expr argument when
1931 bufoff + elt_sz is above size of buf.
1933 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1935 * config/mips/mips.c (mips_declare_object_name)
1936 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1938 2020-01-27 Martin Liska <mliska@suse.cz>
1940 PR gcov-profile/93403
1941 * tree-profile.c (gimple_init_gcov_profiler): Generate
1942 both __gcov_indirect_call_profiler_v4 and
1943 __gcov_indirect_call_profiler_v4_atomic.
1945 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1948 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1950 (@aarch64_split_simd_mov<mode>): Use it.
1951 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1952 Leave the vec_extract patterns to handle 2-element vectors.
1953 (aarch64_simd_mov_from_<mode>high): Likewise.
1954 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1955 (vec_extractv2dfv1df): Likewise.
1957 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1959 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1960 jump conditions for *compare_condjump<GPI:mode>.
1962 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1965 * digraph.cc (test_edge::test_edge): Specify template for base
1968 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1970 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1972 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1974 * config/arc/arc-protos.h (gen_mlo): Remove.
1975 (gen_mhi): Likewise.
1976 * config/arc/arc.c (AUX_MULHI): Define.
1977 (arc_must_save_reister): Special handling for r58/59.
1978 (arc_compute_frame_size): Consider mlo/mhi registers.
1979 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1981 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1982 mlo/mhi name selection.
1983 (arc_restore_callee_saves): Don't early restore blink when ISR.
1984 (arc_expand_prologue): Add mlo/mhi saving.
1985 (arc_expand_epilogue): Add mlo/mhi restoring.
1988 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1989 numbering when MUL64 option is used.
1990 (DWARF2_FRAME_REG_OUT): Define.
1991 * config/arc/arc.md (arc600_stall): New pattern.
1992 (VUNSPEC_ARC_ARC600_STALL): Define.
1993 (mulsi64): Use correct mlo/mhi registers.
1994 (mulsi_600): Clean it up.
1995 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1997 (mhi_operand): Likewise.
1999 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
2000 Petro Karashchenko <petro.karashchenko@ring.com>
2002 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
2003 attributes if needed.
2004 (prepare_move_operands): Generate special unspec instruction for
2006 (arc_isuncached_mem_p): Propagate uncached attribute to each
2008 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
2009 (VUNSPEC_ARC_STDI): Likewise.
2010 (ALLI): New mode iterator.
2011 (mALLI): New mode attribute.
2012 (lddi): New instruction pattern.
2014 (stdidi_split): Split instruction for architectures which are not
2015 supporting ll64 option.
2016 (lddidi_split): Likewise.
2018 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2020 PR rtl-optimization/92989
2021 * lra-lives.c (process_bb_lives): Update the live-in set before
2022 processing additional clobbers.
2024 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2026 PR rtl-optimization/93170
2027 * cselib.c (cselib_invalidate_regno_val): New function, split out
2029 (cselib_invalidate_regno): ...here.
2030 (cselib_invalidated_by_call_p): New function.
2031 (cselib_process_insn): Iterate over all the hard-register entries in
2032 REG_VALUES and invalidate any that cross call-clobbered registers.
2034 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
2036 * dojump.c (split_comparison): Use HONOR_NANS rather than
2037 HONOR_SNANS when splitting LTGT.
2039 2020-01-27 Martin Liska <mliska@suse.cz>
2042 * opts.c (print_filtered_help): Exclude language-specific
2043 options from --help=common unless enabled in all FEs.
2045 2020-01-27 Martin Liska <mliska@suse.cz>
2047 * opts.c (print_help): Exclude params from
2048 all except --help=param.
2050 2020-01-27 Martin Liska <mliska@suse.cz>
2053 * config/i386/i386-features.c (make_resolver_func):
2054 Align the code with ppc64 target implementation.
2055 Do not generate a unique name for resolver function.
2057 2020-01-27 Richard Biener <rguenther@suse.de>
2059 PR tree-optimization/93397
2060 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
2061 converted reduction chain SLP graph adjustment.
2063 2020-01-26 Marek Polacek <polacek@redhat.com>
2066 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
2069 2020-01-26 Jason Merrill <jason@redhat.com>
2072 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
2075 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
2077 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
2078 (rx_setmem): Likewise.
2080 2020-01-26 Jakub Jelinek <jakub@redhat.com>
2083 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
2084 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
2085 drop <di> from constraint of last operand.
2088 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
2089 TARGET_AVX2 and V4DFmode not in the split condition, but in the
2090 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
2092 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
2095 * ipa-cp.c (get_info_about_necessary_edges): Remove value
2098 2020-01-24 Jeff Law <law@redhat.com>
2100 PR tree-optimization/92788
2101 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
2104 2020-01-24 Jakub Jelinek <jakub@redhat.com>
2107 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
2108 *avx_vperm_broadcast_<mode>,
2109 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
2110 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
2111 Move before avx2_perm<mode>/avx512f_perm<mode>.
2114 * simplify-rtx.c (simplify_const_unary_operation,
2115 simplify_const_binary_operation): Punt for mode precision above
2116 MAX_BITSIZE_MODE_ANY_INT.
2118 2020-01-24 Andrew Pinski <apinski@marvell.com>
2120 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
2123 2020-01-24 Jeff Law <law@redhat.com>
2126 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
2127 for REGs. Call output_operand_lossage to get more reasonable
2130 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
2132 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
2133 gcn_fp_compare_operator.
2134 (vec_cmpu<mode>di): Use gcn_compare_operator.
2135 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
2136 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
2137 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
2138 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
2139 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
2140 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
2141 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
2142 gcn_fp_compare_operator.
2143 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
2144 gcn_fp_compare_operator.
2145 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
2146 gcn_fp_compare_operator.
2147 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
2148 gcn_fp_compare_operator.
2150 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
2152 * doc/install.texi (Cross-Compiler-Specific Options): Document
2153 `--with-toolexeclibdir' option.
2155 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
2157 * target.def (flags_regnum): Also mention effect on delay slot filling.
2158 * doc/tm.texi: Regenerate.
2160 2020-01-23 Jeff Law <law@redhat.com>
2162 PR translation/90162
2163 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
2165 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
2168 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
2171 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2173 PR rtl-optimization/93402
2174 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
2177 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2179 * config.in: Regenerated.
2180 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
2181 for TARGET_LIBC_GNUSTACK.
2182 * configure: Regenerated.
2183 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
2184 found to be 2.31 or greater.
2186 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
2188 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
2190 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
2191 (mips_asm_file_end): New function. Delegate to
2192 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
2193 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2195 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2198 * config/i386/i386-modes.def (POImode): New mode.
2199 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
2200 * config/i386/i386.md (DPWI): New mode attribute.
2201 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
2203 (QPWI): ... this. Use POI instead of OI for TImode.
2204 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
2205 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
2208 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2211 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
2213 (speculation_tracker_rev): New pattern.
2214 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
2215 Use speculation_tracker_rev to track the inverse condition.
2217 2020-01-23 Richard Biener <rguenther@suse.de>
2219 PR tree-optimization/93381
2220 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
2221 alias-set of the def as argument and record the first one.
2222 (vn_walk_cb_data::first_set): New member.
2223 (vn_reference_lookup_3): Pass the alias-set of the current def
2224 to push_partial_def. Fix alias-set used in the aggregate copy
2226 (vn_reference_lookup): Consistently set *last_vuse_ptr.
2227 * real.c (clear_significand_below): Fix out-of-bound access.
2229 2020-01-23 Jakub Jelinek <jakub@redhat.com>
2232 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
2233 New define_insn patterns.
2235 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2237 * doc/sourcebuild.texi (check-function-bodies): Add an
2238 optional target/xfail selector.
2240 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
2242 PR rtl-optimization/93124
2243 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
2244 bare USE and CLOBBER insns.
2246 2020-01-22 Andrew Pinski <apinski@marvell.com>
2248 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
2250 2020-01-22 David Malcolm <dmalcolm@redhat.com>
2253 * gdbinit.in (break-on-saved-diagnostic): Update for move of
2254 diagnostic_manager into "ana" namespace.
2255 * selftest-run-tests.c (selftest::run_tests): Update for move of
2256 selftest::run_analyzer_selftests to
2257 ana::selftest::run_analyzer_selftests.
2259 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
2261 * cfgexpand.c (union_stack_vars): Update the size.
2263 2020-01-22 Richard Biener <rguenther@suse.de>
2265 PR tree-optimization/93381
2266 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
2267 throughout, handle all conversions the same.
2269 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2272 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2273 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2274 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2275 Call force_reg on high_in2 unconditionally.
2277 2020-01-22 Martin Liska <mliska@suse.cz>
2279 PR tree-optimization/92924
2280 * profile.c (compute_value_histograms): Divide
2283 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2286 * output.h (assemble_name_resolve): Declare.
2287 * varasm.c (assemble_name_resolve): New function.
2288 (assemble_name): Use it.
2289 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2291 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2293 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2294 update_web_docs_git instead of update_web_docs_svn.
2296 2020-01-21 Andrew Pinski <apinski@marvell.com>
2299 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2300 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2301 (*tlsgd_small_<mode>): Likewise.
2302 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2303 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2304 register. Convert that register back to dest using convert_mode.
2306 2020-01-21 Jim Wilson <jimw@sifive.com>
2308 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2311 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2312 Uros Bizjak <ubizjak@gmail.com>
2315 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2317 (legitimize_tls_address): Do GNU2 TLS address computation in
2318 ptr_mode and zero-extend result to Pmode.
2319 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2320 :P with :PTR and Pmode with ptr_mode.
2321 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2322 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2323 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2325 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2328 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2329 the last two operands are CONST_INT_P before using them as such.
2331 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2333 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2334 to get the integer element types.
2336 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2338 * config/aarch64/aarch64-sve-builtins.h
2339 (function_expander::convert_to_pmode): Declare.
2340 * config/aarch64/aarch64-sve-builtins.cc
2341 (function_expander::convert_to_pmode): New function.
2342 (function_expander::get_contiguous_base): Use it.
2343 (function_expander::prepare_gather_address_operands): Likewise.
2344 * config/aarch64/aarch64-sve-builtins-sve2.cc
2345 (svwhilerw_svwhilewr_impl::expand): Likewise.
2347 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2350 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2351 cfun->machine->label_is_assembled.
2352 (aarch64_print_patchable_function_entry): New.
2353 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2354 * config/aarch64/aarch64.h (struct machine_function): New field,
2357 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2360 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2363 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2366 * cgraph.c (cgraph_edge::resolve_speculation,
2367 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2368 call_stmt_site_hash.
2370 2020-01-21 Martin Liska <mliska@suse.cz>
2372 * config/rs6000/rs6000.c (common_mode_defined): Remove
2375 2020-01-21 Richard Biener <rguenther@suse.de>
2377 PR tree-optimization/92328
2378 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2379 type when value-numbering same-sized store by inserting a
2381 (eliminate_dom_walker::eliminate_stmt): When eliminating
2382 a redundant store handle bit-reinterpretation of the same value.
2384 2020-01-21 Andrew Pinski <apinski@marvel.com>
2387 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2389 (prepare_block_for_update): This. Use a worklist instead of
2392 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2394 * gcc/config/arm/arm.c (clear_operation_p):
2395 Initialise last_regno, skip first iteration
2396 based on the first_set value and use ints instead
2397 of the unnecessary HOST_WIDE_INTs.
2399 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2402 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2403 compare_mode other than SFmode or DFmode.
2405 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2408 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2409 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2410 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2412 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2414 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2416 2020-01-20 Andrew Pinski <apinski@marvell.com>
2419 * targhooks.c (default_print_patchable_function_entry): Use
2420 output_asm_insn to emit the nop instruction.
2422 2020-01-20 Fangrui Song <maskray@google.com>
2425 * targhooks.c (default_print_patchable_function_entry): Align to
2428 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2431 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2432 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2433 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2434 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2435 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2436 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2437 Remove the {q} suffix from lea.
2438 (*tls_dynamic_gnu2_call_64): Renamed to ...
2439 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2440 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2441 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2442 Pass Pmode to gen_tls_dynamic_gnu2_64.
2444 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2446 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2448 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2450 * config/aarch64/aarch64-sve-builtins-base.cc
2451 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2453 2020-01-20 Richard Biener <rguenther@suse.de>
2456 * dwarf2out.c (prune_unused_types): Unconditionally mark
2457 called function DIEs.
2459 2020-01-20 Martin Liska <mliska@suse.cz>
2461 PR tree-optimization/93199
2462 * tree-eh.c (struct leh_state): Add
2463 new field outer_non_cleanup.
2464 (cleanup_is_dead_in): Pass leh_state instead
2465 of eh_region. Add a checking that state->outer_non_cleanup
2466 points to outer non-clean up region.
2467 (lower_try_finally): Record outer_non_cleanup
2469 (lower_catch): Likewise.
2470 (lower_eh_filter): Likewise.
2471 (lower_eh_must_not_throw): Likewise.
2472 (lower_cleanup): Likewise.
2474 2020-01-20 Richard Biener <rguenther@suse.de>
2476 PR tree-optimization/93094
2477 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2478 (vect_transform_loop): Likewise.
2479 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2480 loop_vectorized_call to vect_transform_loop.
2481 * tree-vect-loop.c (vect_transform_loop): Pass down
2482 loop_vectorized_call to vect_loop_versioning.
2483 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2484 the earlier discovered loop_vectorized_call.
2486 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2488 * doc/contribute.texi: Update for SVN -> Git transition.
2489 * doc/install.texi: Likewise.
2491 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2494 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2495 speculative targets.
2496 (verify_speculative_call): New function
2497 (cgraph_node::verify_node): Use it.
2498 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2501 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2504 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2505 (cgraph_edge::make_direct): Remove all indirect targets.
2506 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2507 (cgraph_node::verify_node): Verify that only one call_stmt or
2508 lto_stmt_uid is set.
2509 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2511 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2512 (lto_output_ref): Simplify streaming of stmt.
2513 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2515 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2517 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2518 Mark parameter unused.
2520 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2522 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2524 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2526 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2528 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2530 * Makefile.in: Add coroutine-passes.o.
2531 * builtin-types.def (BT_CONST_SIZE): New.
2532 (BT_FN_BOOL_PTR): New.
2533 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2534 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2535 * coroutine-builtins.def: New file.
2536 * coroutine-passes.cc: New file.
2537 * function.h (struct GTY function): Add a bit to indicate that the
2538 function is a coroutine component.
2539 * internal-fn.c (expand_CO_FRAME): New.
2540 (expand_CO_YIELD): New.
2541 (expand_CO_SUSPN): New.
2542 (expand_CO_ACTOR): New.
2543 * internal-fn.def (CO_ACTOR): New.
2547 * passes.def: Add pass_coroutine_lower_builtins,
2548 pass_coroutine_early_expand_ifns.
2549 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2550 (make_pass_coroutine_early_expand_ifns): New.
2551 * doc/invoke.texi: Document the fcoroutines command line
2554 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2556 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2559 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2560 after checking the argument is a REG. Don't use REGNO (reg)
2561 again to set last_regno, reuse regno variable instead.
2563 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2565 * doc/analyzer.texi (Limitations): Add note about NaN.
2567 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2568 Sudakshina Das <sudi.das@arm.com>
2570 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2571 and valid immediate.
2572 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2573 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2574 * config/arm/constraints.md (Pg): New.
2575 * config/arm/predicates.md (long_shift_imm): New.
2576 (arm_reg_or_long_shift_imm): Likewise.
2577 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2578 (thumb2_lsll): Likewise.
2581 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2582 Sudakshina Das <sudi.das@arm.com>
2584 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2585 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2586 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2587 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2588 * config/arm/thumb2.md (thumb2_asrl): New.
2589 (thumb2_lsll): Likewise.
2591 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2593 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2596 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2598 * gdbinit.in (help-gcc-hooks): New command.
2599 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2600 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2603 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2605 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2606 correct target macro.
2608 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2610 * config/aarch64/aarch64-protos.h
2611 (aarch64_sve_ld1ro_operand_p): New.
2612 * config/aarch64/aarch64-sve-builtins-base.cc
2613 (class load_replicate): New.
2614 (class svld1ro_impl): New.
2615 (class svld1rq_impl): Change to inherit from load_replicate.
2616 (svld1ro): New sve intrinsic function base.
2617 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2618 New DEF_SVE_FUNCTION.
2619 * config/aarch64/aarch64-sve-builtins-base.h
2620 (svld1ro): New decl.
2621 * config/aarch64/aarch64-sve-builtins.cc
2622 (function_expander::add_mem_operand): Modify assert to allow
2624 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2626 * config/aarch64/aarch64.c
2627 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2628 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2629 (aarch64_sve_ld1ro_operand_p): New.
2630 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2631 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2632 * config/aarch64/predicates.md
2633 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2635 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2637 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2638 Introduce this ACLE specified predefined macro.
2639 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2640 (fp): Disabling this disables f64mm.
2641 (simd): Disabling this disables f64mm.
2642 (fp16): Disabling this disables f64mm.
2643 (sve): Disabling this disables f64mm.
2644 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2645 (AARCH64_ISA_F64MM): New.
2646 (TARGET_F64MM): New.
2647 * doc/invoke.texi (f64mm): Document new option.
2649 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2651 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2652 (neoversen1_tunings): Likewise.
2654 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2657 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2658 Add assert to ensure prolog has been emitted.
2659 (aarch64_split_atomic_op): Likewise.
2660 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2661 Use epilogue_completed rather than reload_completed.
2662 (aarch64_atomic_exchange<mode>): Likewise.
2663 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2664 (atomic_nand<mode>): Likewise.
2665 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2666 (atomic_fetch_nand<mode>): Likewise.
2667 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2668 (atomic_nand_fetch<mode>): Likewise.
2670 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2673 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2675 (REVERSE_CONDITION): Delete.
2676 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2677 (CCFP_CCFPE): Likewise.
2678 (e): New mode attribute.
2679 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2680 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2681 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2682 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2683 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2684 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2685 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2686 name of generator from gen_ccmpdi to gen_ccmpccdi.
2687 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2688 the previous comparison but aren't able to, use the new ccmp_rev
2691 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2693 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2694 than testing directly for INTEGER_CST.
2695 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2697 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2699 PR tree-optimization/93292
2700 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2701 get_vectype_for_scalar_type returns NULL.
2703 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2705 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2706 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2708 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2710 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2712 * params.opt: (max-predicted-iterations): Set bounds.
2713 * predict.c (real_almost_one, real_br_prob_base,
2714 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2715 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2716 probabilities; do not truncate to reg_br_prob_bases.
2717 (estimate_loops_at_level): Pass max_cyclic_prob.
2718 (estimate_loops): Compute max_cyclic_prob.
2719 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2721 * profile-count.c (profile_probability::to_sreal): New.
2722 * profile-count.h (class sreal): Move up in file.
2723 (profile_probability::to_sreal): Declare.
2725 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2728 (arm_invalid_conversion): New function for target hook.
2729 (arm_invalid_unary_op): New function for target hook.
2730 (arm_invalid_binary_op): New function for target hook.
2732 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2734 * config.gcc: Add arm_bf16.h.
2735 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2736 (arm_simd_builtin_std_type): Add BFmode.
2737 (arm_init_simd_builtin_types): Define element types for vector types.
2738 (arm_init_bf16_types): New function.
2739 (arm_init_builtins): Add arm_init_bf16_types function call.
2740 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2741 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2742 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2743 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2744 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2745 (arm_mangle_type): Add __bf16.
2746 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2747 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2748 arm_bf16_ptr_type_node.
2749 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2750 define_split between ARM registers.
2751 * config/arm/arm_bf16.h: New file.
2752 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2753 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2755 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2756 * config/arm/vfp.md: Add BFmode to movhf patterns.
2758 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2759 Andre Vieira <andre.simoesdiasvieira@arm.com>
2761 * config/arm/arm-cpus.in (mve, mve_float): New features.
2762 (dsp, mve, mve.fp): New options.
2763 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2764 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2765 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2767 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2768 Thomas Preud'homme <thomas.preudhomme@arm.com>
2770 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2772 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2773 error for using -mcmse when targeting Armv8.1-M Mainline.
2775 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2776 Thomas Preud'homme <thomas.preudhomme@arm.com>
2778 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2779 address in r4 when targeting Armv8.1-M Mainline.
2780 (nonsecure_call_value_internal): Likewise.
2781 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2782 a register match_operand again. Emit BLXNS when targeting
2784 (nonsecure_call_value_reg_thumb2): Likewise.
2786 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2787 Thomas Preud'homme <thomas.preudhomme@arm.com>
2789 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2790 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2791 variable as true when floating-point ABI is not hard. Replace
2792 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2793 Generate VLSTM and VLLDM instruction respectively before and
2794 after a function call to cmse_nonsecure_call function.
2795 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2796 (VUNSPEC_VLLDM): Likewise.
2797 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2798 (lazy_load_multiple_insn): Likewise.
2800 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2801 Thomas Preud'homme <thomas.preudhomme@arm.com>
2803 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2804 (arm_emit_vfp_multi_reg_pop): Likewise.
2805 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2806 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2807 restore callee-saved VFP registers.
2809 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2810 Thomas Preud'homme <thomas.preudhomme@arm.com>
2812 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2813 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2814 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2815 callee-saved GPRs as well as clear ip register before doing a nonsecure
2816 call then restore callee-saved GPRs after it when targeting
2818 (arm_reorg): Adapt to function rename.
2820 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2821 Thomas Preud'homme <thomas.preudhomme@arm.com>
2823 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2824 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2825 clear_vfp_multiple pattern based on a new vfp parameter.
2826 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2827 targeting Armv8.1-M Mainline.
2828 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2829 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2830 whether VFP registers are available before looking call_used_regs for a
2832 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2833 of prototype of clear_operation_p.
2834 (clear_vfp_multiple_operation): New predicate.
2835 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2836 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2838 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2839 Thomas Preud'homme <thomas.preudhomme@arm.com>
2841 * config/arm/arm-protos.h (clear_operation_p): Declare.
2842 * config/arm/arm.c (clear_operation_p): New function.
2843 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2844 targeting Armv8.1-M Mainline or successor.
2845 (output_return_instruction): Only output APSR register clearing if
2846 Armv8.1-M Mainline instructions not available.
2847 (thumb_exit): Likewise.
2848 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2849 * config/arm/thumb2.md (clear_apsr): New define_insn.
2850 (clear_multiple): Likewise.
2851 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2853 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2854 Thomas Preud'homme <thomas.preudhomme@arm.com>
2856 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2857 (use_return_insn): Also return false for Armv8.1-M Mainline.
2858 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2859 Mainline instructions are available.
2860 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2861 when targeting Armv8.1-M Mainline Security Extensions.
2862 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2863 Mainline entry function.
2864 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2865 targeting Armv8.1-M Mainline or successor.
2866 (arm_expand_epilogue): Fix indentation of caller-saved register
2867 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2869 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2870 (FP_SYSREGS): Likewise.
2871 (enum vfp_sysregs_encoding): Define enum.
2872 (fp_sysreg_names): Declare.
2873 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2874 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2875 (pop_fpsysreg_insn): Likewise.
2877 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2878 Thomas Preud'homme <thomas.preudhomme@arm.com>
2880 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2881 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2882 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2883 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2884 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2885 (ARMv8_1m_main): New feature group.
2886 (armv8.1-m.main): New architecture.
2887 * config/arm/arm-tables.opt: Regenerate.
2888 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2889 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2890 (arm_options_perform_arch_sanity_checks): Error out when targeting
2891 Armv8.1-M Mainline Security Extensions.
2892 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2894 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2896 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2897 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2898 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2899 aarch64_bfdot_laneq): New.
2900 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2901 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2902 vbfdotq_laneq_f32): New.
2903 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2904 VBFMLA_W, VBF): New.
2905 (isquadop): Add V4BF, V8BF.
2907 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2909 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2910 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2911 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2912 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2913 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2914 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2915 usdot_laneq, sudot_lane,sudot_laneq): New.
2916 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2917 (aarch64_<sur>dot_lane): New.
2918 * config/aarch64/arm_neon.h (vusdot_s32): New.
2920 (vusdot_lane_s32): New.
2921 (vsudot_lane_s32): New.
2922 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2923 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2925 2020-01-16 Martin Liska <mliska@suse.cz>
2927 * value-prof.c (dump_histogram_value): Fix
2928 obvious spacing issue.
2930 2020-01-16 Andrew Pinski <apinski@marvell.com>
2932 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2933 !storage_order_barrier_p.
2935 2020-01-16 Andrew Pinski <apinski@marvell.com>
2937 * sched-int.h (_dep): Add unused bit-field field for the padding.
2938 * sched-deps.c (init_dep_1): Init unused field.
2940 2020-01-16 Andrew Pinski <apinski@marvell.com>
2942 * optabs.h (create_expand_operand): Initialize target field also.
2944 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2946 PR tree-optimization/92429
2947 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2948 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2950 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2953 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2955 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2956 aarch64_sve_int_mode to each mode.
2958 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2960 * doc/analyzer.texi (Overview): Add note about
2961 -fdump-ipa-analyzer.
2963 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2965 PR tree-optimization/93231
2966 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2967 input_type is unsigned. Use tree_to_shwi for shift constant.
2968 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2969 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2970 inputs more efficiently.
2972 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2974 * config/i386/i386.md (*movsf_internal): Do not require
2975 SSE2 ISA for alternatives 14 and 15.
2977 2020-01-15 Richard Biener <rguenther@suse.de>
2980 * tree-eh.c (sink_clobbers): If we already visited the destination
2981 block do not defer insertion.
2982 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2983 the purpose of defered insertion.
2985 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2987 * BASE-VER: Bump to 10.0.1.
2989 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2991 PR tree-optimization/93247
2992 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2993 type of the stmt that we're going to vectorize.
2995 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2997 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2998 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
3001 2020-01-15 Martin Liska <mliska@suse.cz>
3003 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
3004 2 calls of streamer_read_hwi in a function call.
3006 2020-01-15 Richard Biener <rguenther@suse.de>
3008 * alias.c (record_alias_subset): Avoid redundant work when
3009 subset is already recorded.
3011 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3013 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
3014 the analyzer options provide CWE identifiers.
3016 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3018 * tree-diagnostic-path.cc (path_summary::event_range::print):
3019 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
3020 using get_pure_location.
3022 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3024 PR tree-optimization/93262
3025 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
3026 perform head trimming only if the last argument is constant,
3027 either all ones, or larger or equal to head trim, in the latter
3028 case decrease the last argument by head_trim.
3030 PR tree-optimization/93249
3031 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
3032 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
3033 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
3034 perform head trim unless we can prove there are no '\0' chars
3035 from the source among the first head_trim chars.
3037 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3039 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
3041 2020-01-15 Jakub Jelinek <jakub@redhat.com>
3044 * config/i386/sse.md
3045 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
3046 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
3047 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
3048 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
3049 just a single alternative instead of two, make operands 1 and 2
3052 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
3055 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
3058 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3060 * Makefile.in (lang_opt_files): Add analyzer.opt.
3061 (ANALYZER_OBJS): New.
3062 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
3063 tristate.o and ANALYZER_OBJS.
3064 (TEXI_GCCINT_FILES): Add analyzer.texi.
3065 * common.opt (-fanalyzer): New driver option.
3066 * config.in: Regenerate.
3067 * configure: Regenerate.
3068 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
3069 (gccdepdir): Also create depdir for "analyzer" subdir.
3070 * digraph.cc: New file.
3071 * digraph.h: New file.
3072 * doc/analyzer.texi: New file.
3073 * doc/gccint.texi ("Static Analyzer") New menu item.
3074 (analyzer.texi): Include it.
3075 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
3076 ("Warning Options"): Add static analysis warnings to the list.
3077 (-Wno-analyzer-double-fclose): New option.
3078 (-Wno-analyzer-double-free): New option.
3079 (-Wno-analyzer-exposure-through-output-file): New option.
3080 (-Wno-analyzer-file-leak): New option.
3081 (-Wno-analyzer-free-of-non-heap): New option.
3082 (-Wno-analyzer-malloc-leak): New option.
3083 (-Wno-analyzer-possible-null-argument): New option.
3084 (-Wno-analyzer-possible-null-dereference): New option.
3085 (-Wno-analyzer-null-argument): New option.
3086 (-Wno-analyzer-null-dereference): New option.
3087 (-Wno-analyzer-stale-setjmp-buffer): New option.
3088 (-Wno-analyzer-tainted-array-index): New option.
3089 (-Wno-analyzer-use-after-free): New option.
3090 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
3091 (-Wno-analyzer-use-of-uninitialized-value): New option.
3092 (-Wanalyzer-too-complex): New option.
3093 (-fanalyzer-call-summaries): New warning.
3094 (-fanalyzer-checker=): New warning.
3095 (-fanalyzer-fine-grained): New warning.
3096 (-fno-analyzer-state-merge): New warning.
3097 (-fno-analyzer-state-purge): New warning.
3098 (-fanalyzer-transitivity): New warning.
3099 (-fanalyzer-verbose-edges): New warning.
3100 (-fanalyzer-verbose-state-changes): New warning.
3101 (-fanalyzer-verbosity=): New warning.
3102 (-fdump-analyzer): New warning.
3103 (-fdump-analyzer-callgraph): New warning.
3104 (-fdump-analyzer-exploded-graph): New warning.
3105 (-fdump-analyzer-exploded-nodes): New warning.
3106 (-fdump-analyzer-exploded-nodes-2): New warning.
3107 (-fdump-analyzer-exploded-nodes-3): New warning.
3108 (-fdump-analyzer-supergraph): New warning.
3109 * doc/sourcebuild.texi (dg-require-dot): New.
3110 (dg-check-dot): New.
3111 * gdbinit.in (break-on-saved-diagnostic): New command.
3112 * graphviz.cc: New file.
3113 * graphviz.h: New file.
3114 * ordered-hash-map-tests.cc: New file.
3115 * ordered-hash-map.h: New file.
3116 * passes.def (pass_analyzer): Add before
3117 pass_ipa_whole_program_visibility.
3118 * selftest-run-tests.c (selftest::run_tests): Call
3119 selftest::ordered_hash_map_tests_cc_tests.
3120 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
3122 * shortest-paths.h: New file.
3123 * timevar.def (TV_ANALYZER): New timevar.
3124 (TV_ANALYZER_SUPERGRAPH): Likewise.
3125 (TV_ANALYZER_STATE_PURGE): Likewise.
3126 (TV_ANALYZER_PLAN): Likewise.
3127 (TV_ANALYZER_SCC): Likewise.
3128 (TV_ANALYZER_WORKLIST): Likewise.
3129 (TV_ANALYZER_DUMP): Likewise.
3130 (TV_ANALYZER_DIAGNOSTICS): Likewise.
3131 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
3132 * tree-pass.h (make_pass_analyzer): New decl.
3133 * tristate.cc: New file.
3134 * tristate.h: New file.
3136 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
3139 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
3140 alternatives 9 and 10.
3142 2020-01-14 David Malcolm <dmalcolm@redhat.com>
3144 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
3145 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
3146 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
3147 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
3148 (selftest::hash_map_tests_c_tests): Call it.
3149 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
3150 New static constant, using the value of = H::empty_zero_p.
3151 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
3152 from default_hash_traits <Value>.
3153 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
3155 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
3156 * hash-table.h (hash_table::alloc_entries): Guard the loop of
3157 calls to mark_empty with !Descriptor::empty_zero_p.
3158 (hash_table::empty_slow): Conditionalize the memset call with a
3159 check that Descriptor::empty_zero_p; otherwise, loop through the
3160 entries calling mark_empty on them.
3161 * hash-traits.h (int_hash::empty_zero_p): New static constant.
3162 (pointer_hash::empty_zero_p): Likewise.
3163 (pair_hash::empty_zero_p): Likewise.
3164 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
3166 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
3167 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
3168 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
3169 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
3170 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
3171 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
3172 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
3173 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
3175 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
3178 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
3180 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
3181 fix typo on return value.
3183 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
3186 * cgraph.c (symbol_table::create_edge): Init speculative_id and
3188 (cgraph_edge::make_speculative): Add param for setting speculative_id
3190 (cgraph_edge::speculative_call_info): Update comments and find reference
3191 by speculative_id for multiple indirect targets.
3192 (cgraph_edge::resolve_speculation): Decrease the speculations
3193 for indirect edge, drop it's speculative if not direct target
3194 left. Update comments.
3195 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
3196 (cgraph_node::dump): Print num_speculative_call_targets.
3197 (cgraph_node::verify_node): Don't report error if speculative
3198 edge not include statement.
3199 (cgraph_edge::num_speculative_call_targets_p): New function.
3200 * cgraph.h (int common_target_id): Remove.
3201 (int common_target_probability): Remove.
3202 (num_speculative_call_targets): New variable.
3203 (make_speculative): Add param for setting speculative_id.
3204 (cgraph_edge::num_speculative_call_targets_p): New declare.
3205 (target_prob): New variable.
3206 (speculative_id): New variable.
3207 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
3208 call summaries for multiple speculative call targets.
3209 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
3210 * ipa-profile.c (struct speculative_call_target): New struct.
3211 (class speculative_call_summary): New class.
3212 (class speculative_call_summaries): New class.
3213 (call_sums): New variable.
3214 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
3215 (ipa_profile_write_edge_summary): New function.
3216 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
3217 (ipa_profile_dump_all_summaries): New function.
3218 (ipa_profile_read_edge_summary): New function.
3219 (ipa_profile_read_summary_section): New function.
3220 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
3221 (ipa_profile): Generate num_speculative_call_targets from
3223 * ipa-ref.h (speculative_id): New variable.
3224 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
3225 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
3226 common_target_probability. Stream out speculative_id and
3227 num_speculative_call_targets.
3228 (input_edge): Likewise.
3229 * predict.c (dump_prediction): Remove edges count assert to be
3231 * symtab.c (symtab_node::create_reference): Init speculative_id.
3232 (symtab_node::clone_references): Clone speculative_id.
3233 (symtab_node::clone_referring): Clone speculative_id.
3234 (symtab_node::clone_reference): Clone speculative_id.
3235 (symtab_node::clear_stmts_in_references): Clear speculative_id.
3236 * tree-inline.c (copy_bb): Duplicate all the speculative edges
3237 if indirect call contains multiple speculative targets.
3238 * value-prof.h (check_ic_target): Remove.
3239 * value-prof.c (gimple_value_profile_transformations):
3240 Use void function gimple_ic_transform.
3241 * value-prof.c (gimple_ic_transform): Handle topn case.
3242 Fix comment typos. Change it to a void function.
3244 2020-01-13 Andrew Pinski <apinski@marvell.com>
3246 * config/aarch64/aarch64-cores.def (octeontx2): New define.
3247 (octeontx2t98): New define.
3248 (octeontx2t96): New define.
3249 (octeontx2t93): New define.
3250 (octeontx2f95): New define.
3251 (octeontx2f95n): New define.
3252 (octeontx2f95mm): New define.
3253 * config/aarch64/aarch64-tune.md: Regenerate.
3254 * doc/invoke.texi (-mcpu=): Document the new cpu types.
3256 2020-01-13 Jason Merrill <jason@redhat.com>
3258 PR c++/33799 - destroy return value if local cleanup throws.
3259 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
3261 2020-01-13 Martin Liska <mliska@suse.cz>
3263 * ipa-cp.c (get_max_overall_size): Use newly
3264 renamed param param_ipa_cp_unit_growth.
3265 * params.opt: Remove legacy param name.
3267 2020-01-13 Martin Sebor <msebor@redhat.com>
3269 PR tree-optimization/93213
3270 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
3271 stores to be eliminated.
3273 2020-01-13 Martin Liska <mliska@suse.cz>
3275 * opts.c (print_help): Do not print CL_PARAM
3276 and CL_WARNING for CL_OPTIMIZATION.
3278 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3281 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3282 depending on optimization settings.
3284 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3286 PR tree-optimization/90838
3287 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3288 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3289 argument rather than to initialize temporary for targets that
3290 don't use the mode argument at all. Initialize ctzval to avoid
3293 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3295 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3296 * tree-core.h: Document it.
3297 * gimplify.c (gimplify_omp_workshare): Set it.
3298 * omp-low.c (lower_omp_target): Use it.
3299 * tree-pretty-print.c (dump_omp_clause): Print it.
3301 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3302 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3304 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3306 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3307 * common.opt (fdiagnostics-path-format=): New option.
3308 (diagnostic_path_format): New enum.
3309 (fdiagnostics-show-path-depths): New option.
3310 * coretypes.h (diagnostic_event_id_t): New forward decl.
3311 * diagnostic-color.c (color_dict): Add "path".
3312 * diagnostic-event-id.h: New file.
3313 * diagnostic-format-json.cc (json_from_expanded_location): Make
3315 (json_end_diagnostic): Call context->make_json_for_path if it
3316 exists and the diagnostic has a path.
3317 (diagnostic_output_format_init): Clear context->print_path.
3318 * diagnostic-path.h: New file.
3319 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3320 when printing a run of events in a diagnostic_path so that they
3321 all get the same color.
3322 (layout::m_diagnostic_path_p): New field.
3323 (layout::layout): Initialize it.
3324 (layout::print_any_labels): Don't colorize the label text for an
3325 event in a diagnostic_path.
3326 (gcc_rich_location::add_location_if_nearby): Add
3327 "restrict_to_current_line_spans" and "label" params. Pass the
3328 former to layout.maybe_add_location_range; pass the latter
3329 when calling add_range.
3330 * diagnostic.c: Include "diagnostic-path.h".
3331 (diagnostic_initialize): Initialize context->path_format and
3332 context->show_path_depths.
3333 (diagnostic_show_any_path): New function.
3334 (diagnostic_path::interprocedural_p): New function.
3335 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3336 (simple_diagnostic_path::num_events): New function.
3337 (simple_diagnostic_path::get_event): New function.
3338 (simple_diagnostic_path::add_event): New function.
3339 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3340 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3341 (debug): New overload taking a diagnostic_path *.
3342 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3343 * diagnostic.h (enum diagnostic_path_format): New enum.
3344 (json::value): New forward decl.
3345 (diagnostic_context::path_format): New field.
3346 (diagnostic_context::show_path_depths): New field.
3347 (diagnostic_context::print_path): New callback field.
3348 (diagnostic_context::make_json_for_path): New callback field.
3349 (diagnostic_show_any_path): New decl.
3350 (json_from_expanded_location): New decl.
3351 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3352 (-fdiagnostics-show-path-depths): New option.
3353 (-fdiagnostics-color): Add "path" to description of default
3354 GCC_COLORS; describe it.
3355 (-fdiagnostics-format=json): Document how diagnostic paths are
3356 represented in the JSON output format.
3357 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3358 Add optional params "restrict_to_current_line_spans" and "label".
3359 * opts.c (common_handle_option): Handle
3360 OPT_fdiagnostics_path_format_ and
3361 OPT_fdiagnostics_show_path_depths.
3362 * pretty-print.c: Include "diagnostic-event-id.h".
3363 (pp_format): Implement "%@" format code for printing
3364 diagnostic_event_id_t *.
3365 (selftest::test_pp_format): Add tests for "%@".
3366 * selftest-run-tests.c (selftest::run_tests): Call
3367 selftest::tree_diagnostic_path_cc_tests.
3368 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3369 * toplev.c (general_init): Initialize global_dc->path_format and
3370 global_dc->show_path_depths.
3371 * tree-diagnostic-path.cc: New file.
3372 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3373 non-static. Drop "diagnostic" param in favor of storing the
3374 original value of "where" and re-using it.
3375 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3376 maybe_unwind_expanded_macro_loc.
3377 (tree_diagnostics_defaults): Initialize context->print_path and
3378 context->make_json_for_path.
3379 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3381 (default_tree_make_json_for_path): New decl.
3382 (maybe_unwind_expanded_macro_loc): New decl.
3384 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3386 PR tree-optimization/93210
3387 * fold-const.h (native_encode_initializer,
3388 can_native_interpret_type_p): Declare.
3389 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3391 (native_encode_initializer): New function, moved from dwarf2out.c.
3392 Adjust to native_encode_expr compatible arguments, including dry-run
3393 and partial extraction modes. Don't handle STRING_CST.
3394 (can_native_interpret_type_p): No longer static.
3395 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3396 offset / BITS_PER_UNIT fits into int and don't call it if
3397 can_native_interpret_type_p fails. If suboff is NULL and for
3398 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3399 native_encode_initializer.
3400 (fold_const_aggregate_ref_1): Formatting fix.
3401 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3402 (tree_add_const_value_attribute): Adjust caller.
3404 PR tree-optimization/90838
3405 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3406 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3407 CTZ_DEFINED_VALUE_AT_ZERO.
3409 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3412 * lra-constraints.c (match_reload): Permit input operands have the
3413 same mode as output while other input operands have a different
3416 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3418 PR tree-optimization/90838
3419 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3420 (check_ctz_string): Likewise.
3421 (optimize_count_trailing_zeroes): Likewise.
3422 (simplify_count_trailing_zeroes): Likewise.
3423 (pass_forwprop::execute): Try ctz simplification.
3424 * match.pd: Add matching for ctz idioms.
3426 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3428 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3430 (aarch64_invalid_unary_op): New function for target hook.
3431 (aarch64_invalid_binary_op): New function for target hook.
3433 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3435 * config.gcc: Add arm_bf16.h.
3436 * config/aarch64/aarch64-builtins.c
3437 (aarch64_simd_builtin_std_type): Add BFmode.
3438 (aarch64_init_simd_builtin_types): Define element types for vector
3440 (aarch64_init_bf16_types): New function.
3441 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3442 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3444 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3445 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3447 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3448 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3449 * config/aarch64/aarch64.c
3450 (aarch64_classify_vector_mode): Add support for BF types.
3451 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3452 (aarch64_vq_mode): Add support for BF types.
3453 (aarch64_simd_container_mode): Add support for BF types.
3454 (aarch64_mangle_type): Add support for BF scalar type.
3455 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3456 * config/aarch64/arm_bf16.h: New file.
3457 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3458 * config/aarch64/iterators.md: Add BF types to mode attributes.
3459 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3461 2020-01-10 Jason Merrill <jason@redhat.com>
3463 PR c++/93173 - incorrect tree sharing.
3464 * gimplify.c (copy_if_shared): No longer static.
3465 * gimplify.h: Declare it.
3467 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3469 * doc/invoke.texi (-msve-vector-bits=): Document that
3470 -msve-vector-bits=128 now generates VL-specific code for
3471 little-endian targets.
3472 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3473 build_vector_type_for_mode to construct the data vector types.
3474 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3475 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3476 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3477 for 128-bit vectors.
3479 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3481 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3484 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3486 * config/aarch64/aarch64-builtins.c
3487 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3488 rather than checking the number of elements and the element mode.
3490 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3492 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3493 get_related_vectype_for_scalar_type rather than build_vector_type
3494 to create the index type for a conditional reduction.
3496 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3498 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3499 for any type of gather or scatter, including strided accesses.
3501 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3503 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3506 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3508 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3510 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3511 parameter and its use to reset DR_OFFSET's.
3512 (vect_transform_loop): Remove orig_drs_init argument.
3513 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3514 member of dr_vec_info rather than the offset of the associated
3515 data_reference's innermost_loop_behavior.
3516 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3517 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3518 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3519 get_dr_vinfo_offset.
3520 (vectorizable_store): Likewise.
3521 (vectorizable_load): Likewise.
3523 2020-01-10 Richard Biener <rguenther@suse.de>
3525 * gimple-ssa-store-merging
3526 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3528 2020-01-10 Martin Liska <mliska@suse.cz>
3531 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3532 encapsulation that was there before r280040.
3534 2020-01-10 Richard Biener <rguenther@suse.de>
3537 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3538 sequences to avoid walking them again for secondary opportunities.
3539 (pass_lower_eh_dispatch::execute): Instead actually insert
3542 2020-01-10 Richard Biener <rguenther@suse.de>
3545 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3546 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3547 avoid quadraticness.
3549 2020-01-10 Martin Jambor <mjambor@suse.cz>
3551 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3552 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3553 to get param_ipa_sra_max_replacements.
3554 (param_splitting_across_edge): Pass the caller to
3555 pull_accesses_from_callee.
3557 2020-01-10 Martin Jambor <mjambor@suse.cz>
3559 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3560 * ipa-cp.c (max_new_size): Removed.
3561 (orig_overall_size): New variable.
3562 (get_max_overall_size): New function.
3563 (estimate_local_effects): Use it. Adjust dump.
3564 (decide_about_value): Likewise.
3565 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3566 orig_overall_size. Adjust dump.
3567 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3569 2020-01-10 Martin Jambor <mjambor@suse.cz>
3571 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3572 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3573 instead of param_ipa_max_agg_items.
3574 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3575 optimization info for the callee.
3577 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3579 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3580 markers if debug_inline_points is false.
3582 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3584 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3586 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3587 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3588 aarch64-sve-builtins-sve2.h.
3589 (aarch64-sve-builtins-sve2.o): New rule.
3590 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3591 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3592 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3593 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3594 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3595 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3597 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3598 instructions that are handled here.
3599 (@cond_asrd<mode>): Generalize to...
3600 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3601 (*cond_asrd<mode>_2): Generalize to...
3602 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3603 (*cond_asrd<mode>_z): Generalize to...
3604 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3605 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3606 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3607 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3608 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3610 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3611 (@aarch64_scatter_stnt<mode>): Likewise.
3612 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3613 (@aarch64_mul_lane_<mode>): Likewise.
3614 (@aarch64_sve_suqadd<mode>_const): Likewise.
3615 (*<sur>h<addsub><mode>): Generalize to...
3616 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3618 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3619 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3620 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3621 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3622 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3623 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3624 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3625 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3626 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3627 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3628 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3629 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3630 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3631 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3632 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3633 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3634 (@aarch64_sve2_xar<mode>): Likewise.
3635 (@aarch64_sve2_bcax<mode>): Likewise.
3636 (*aarch64_sve2_eor3<mode>): Rename to...
3637 (@aarch64_sve2_eor3<mode>): ...this.
3638 (@aarch64_sve2_bsl<mode>): New expander.
3639 (@aarch64_sve2_nbsl<mode>): Likewise.
3640 (@aarch64_sve2_bsl1n<mode>): Likewise.
3641 (@aarch64_sve2_bsl2n<mode>): Likewise.
3642 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3643 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3644 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3645 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3646 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3647 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3648 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3649 (<su>mull<bt><Vwide>): Generalize to...
3650 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3652 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3653 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3654 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3655 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3656 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3657 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3658 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3659 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3660 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3661 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3662 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3663 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3664 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3665 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3666 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3667 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3668 (<SHRNB:r>shrnb<mode>): Generalize to...
3669 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3671 (<SHRNT:r>shrnt<mode>): Generalize to...
3672 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3674 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3675 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3676 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3677 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3678 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3679 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3680 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3681 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3682 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3683 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3684 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3685 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3686 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3687 (@aarch64_sve2_cvtnt<mode>): Likewise.
3688 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3689 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3690 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3691 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3692 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3693 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3694 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3695 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3696 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3697 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3698 (@aarch64_sve2_pmul<mode>): Likewise.
3699 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3700 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3701 (@aarch64_sve2_tbl2<mode>): Likewise.
3702 (@aarch64_sve2_tbx<mode>): Likewise.
3703 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3704 (@aarch64_sve2_histcnt<mode>): Likewise.
3705 (@aarch64_sve2_histseg<mode>): Likewise.
3706 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3707 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3708 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3709 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3710 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3711 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3712 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3713 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3714 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3715 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3716 (SVE2_PMULL_PAIR_I): New mode iterators.
3717 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3718 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3719 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3720 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3721 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3722 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3723 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3724 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3725 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3726 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3727 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3728 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3729 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3730 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3731 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3732 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3733 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3734 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3735 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3736 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3737 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3738 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3739 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3740 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3741 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3742 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3743 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3744 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3745 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3746 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3747 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3749 (VNARROW, Ventype): New mode attributes.
3750 (Vewtype): Handle VNx2DI. Fix typo in comment.
3751 (VDOUBLE): New mode attribute.
3752 (sve_lane_con): Handle VNx8HI.
3753 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3754 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3755 (sve_int_op, sve_int_op_rev): Handle the above codes.
3756 (sve_pred_int_rhs2_operand): Likewise.
3757 (MULLBT, SHRNB, SHRNT): Delete.
3758 (SVE_INT_SHIFT_IMM): New int iterator.
3759 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3760 and UNSPEC_WHILEHS for TARGET_SVE2.
3761 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3762 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3763 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3764 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3765 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3766 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3767 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3768 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3769 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3770 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3771 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3772 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3773 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3774 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3775 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3776 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3777 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3778 (optab): Handle the new unspecs.
3779 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3781 (lr): Handle the new unspecs.
3783 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3784 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3785 (sve_int_qsub_op): New int attributes.
3786 (sve_fp_op, rot): Handle the new unspecs.
3787 * config/aarch64/aarch64-sve-builtins.h
3788 (function_resolver::require_matching_pointer_type): Declare.
3789 (function_resolver::resolve_unary): Add an optional boolean argument.
3790 (function_resolver::finish_opt_n_resolution): Add an optional
3791 type_suffix_index argument.
3792 (gimple_folder::redirect_call): Declare.
3793 (gimple_expander::prepare_gather_address_operands): Add an optional
3795 * config/aarch64/aarch64-sve-builtins.cc: Include
3796 aarch64-sve-builtins-sve2.h.
3797 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3798 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3799 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3800 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3801 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3802 (TYPES_s_unsigned): Likewise.
3803 (TYPES_s_integer): Use TYPES_s_unsigned.
3804 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3805 (TYPES_sd_integer): Use them.
3806 (TYPES_d_unsigned): New macro.
3807 (TYPES_d_integer): Use it.
3808 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3809 (TYPES_cvt_narrow): Likewise.
3810 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3811 (preds_mx): New variable.
3812 (function_builder::add_overloaded_function): Allow the new feature
3813 set to be more restrictive than the original one.
3814 (function_resolver::infer_pointer_type): Remove qualifiers from
3815 the pointer type before printing it.
3816 (function_resolver::require_matching_pointer_type): New function.
3817 (function_resolver::resolve_sv_displacement): Handle functions
3818 that don't support 32-bit vector indices or svint32_t vector offsets.
3819 (function_resolver::finish_opt_n_resolution): Take the inferred type
3820 as a separate argument.
3821 (function_resolver::resolve_unary): Optionally treat all forms in
3822 the same way as normal merging functions.
3823 (gimple_folder::redirect_call): New function.
3824 (function_expander::prepare_gather_address_operands): Add an argument
3825 that says whether scaled forms are available. If they aren't,
3826 handle scaling of vector indices and don't add the extension and
3828 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3829 fall back to using cond_* instead.
3830 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3831 Split out the member variables into...
3832 (rtx_code_function_base): ...this new base class.
3833 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3834 (unspec_based_function): Split out the member variables into...
3835 (unspec_based_function_base): ...this new base class.
3836 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3837 (unspec_based_function_exact_insn): New class.
3838 (unspec_based_add_function, unspec_based_add_lane_function)
3839 (unspec_based_lane_function, unspec_based_pred_function)
3840 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3841 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3842 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3844 (unspec_based_fused_function): New class.
3845 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3846 (unspec_based_fused_lane_function): New class.
3847 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3849 (CODE_FOR_MODE1): New macro.
3850 (fixed_insn_function): New class.
3851 (while_comparison): Likewise.
3852 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3853 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3854 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3855 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3856 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3857 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3858 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3859 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3860 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3861 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3862 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3863 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3864 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3865 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3866 Also add an initial argument for unary_convert_narrowt, regardless
3867 of the predication type.
3868 (build_32_64): Allow loads and stores to specify MODE_none.
3869 (build_sv_index64, build_sv_uint_offset): New functions.
3870 (long_type_suffix): New function.
3871 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3872 (binary_imm_long_base, load_gather_sv_base): Likewise.
3873 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3874 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3875 (unary_narrowb_base, unary_narrowt_base): Likewise.
3876 (binary_long_lane_def, binary_long_lane): New shape.
3877 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3878 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3879 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3880 (binary_to_uint_def, binary_to_uint): Likewise.
3881 (binary_wide_def, binary_wide): Likewise.
3882 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3883 (compare_def, compare): Likewise.
3884 (compare_ptr_def, compare_ptr): Likewise.
3885 (load_ext_gather_index_restricted_def,
3886 load_ext_gather_index_restricted): Likewise.
3887 (load_ext_gather_offset_restricted_def,
3888 load_ext_gather_offset_restricted): Likewise.
3889 (load_gather_sv_def): Inherit from load_gather_sv_base.
3890 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3891 (shift_left_imm_def, shift_left_imm): Likewise.
3892 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3893 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3894 (store_scatter_index_restricted_def,
3895 store_scatter_index_restricted): Likewise.
3896 (store_scatter_offset_restricted_def,
3897 store_scatter_offset_restricted): Likewise.
3898 (tbl_tuple_def, tbl_tuple): Likewise.
3899 (ternary_long_lane_def, ternary_long_lane): Likewise.
3900 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3901 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3902 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3903 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3904 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3905 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3906 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3907 (ternary_uint_def, ternary_uint): Likewise.
3908 (unary_convert): Fix typo in comment.
3909 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3910 (unary_long_def, unary_long): Likewise.
3911 (unary_narrowb_def, unary_narrowb): Likewise.
3912 (unary_narrowt_def, unary_narrowt): Likewise.
3913 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3914 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3915 (unary_to_int_def, unary_to_int): Likewise.
3916 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3917 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3918 (svasrd_impl): Delete.
3919 (svcadd_impl::expand): Handle integer operations too.
3920 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3921 new functions to derive the unspec numbers.
3922 (svmla_svmls_lane_impl): Replace with...
3923 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3924 integer operations too.
3925 (svwhile_impl): Rename to...
3926 (svwhilelx_impl): ...this and inherit from while_comparison.
3927 (svasrd): Use unspec_based_function.
3928 (svmla_lane): Use svmla_lane_impl.
3929 (svmls_lane): Use svmls_lane_impl.
3930 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3931 (svwhilele, svwhilelt): Use svwhilelx_impl.
3932 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3933 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3934 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3935 * config/aarch64/aarch64-sve-builtins.def: Include
3936 aarch64-sve-builtins-sve2.def.
3938 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3940 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3941 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3942 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3943 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3944 immediates as well as vector ones.
3945 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3946 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3947 (aarch64_sve_qsub_immediate): Update calls accordingly.
3949 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3951 * config/aarch64/aarch64-sve2.md: Add banner comments.
3952 (<su>mulh<r>s<mode>3): Move further up file.
3953 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3954 (*aarch64_sve2_sra<mode>): Move further down file.
3955 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3957 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3959 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3961 (while_optab_cmp): Handle them.
3962 * config/aarch64/aarch64-sve.md
3963 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3964 and add a "@" marker.
3965 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3966 instead of gen_aarch64_sve2_while_ptest.
3967 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3969 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3971 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3972 (UNSPEC_WHILELE): ...this.
3973 (UNSPEC_WHILE_LO): Rename to...
3974 (UNSPEC_WHILELO): ...this.
3975 (UNSPEC_WHILE_LS): Rename to...
3976 (UNSPEC_WHILELS): ...this.
3977 (UNSPEC_WHILE_LT): Rename to...
3978 (UNSPEC_WHILELT): ...this.
3979 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3980 (cmp_op, while_optab_cmp): Likewise.
3981 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3982 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3983 (svwhilelt): Likewise.
3985 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3987 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3988 (unary_to_uint): Define.
3989 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3990 (unary_count): Rename to...
3991 (unary_to_uint_def, unary_to_uint): ...this.
3992 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3994 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3996 * config/aarch64/aarch64-sve-builtins-functions.h
3997 (code_for_mode_function): New class.
3998 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3999 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
4000 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
4001 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
4002 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
4004 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4006 * config/aarch64/iterators.md (addsub): New code attribute.
4007 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
4009 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
4010 in the asm string and attributes. Fix indentation.
4011 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
4013 (@aarch64_sve_<optab><mode>): ...this.
4014 * config/aarch64/aarch64-sve-builtins.h
4015 (function_expander::expand_signed_unpred_op): Delete.
4016 * config/aarch64/aarch64-sve-builtins.cc
4017 (function_expander::expand_signed_unpred_op): Likewise.
4018 (function_expander::map_to_rtx_codes): If the optab isn't defined,
4019 try using code_for_aarch64_sve instead.
4020 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
4021 (svqsub_impl): Likewise.
4022 (svqadd, svqsub): Use rtx_code_function instead.
4024 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4026 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
4027 (HADDSUB, sur, addsub): Remove them.
4029 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4031 * tree-nrv.c (pass_return_slot::execute): Handle all internal
4032 functions the same way, rather than singling out those that
4033 aren't mapped directly to optabs.
4035 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
4037 * target.def (compatible_vector_types_p): New target hook.
4038 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
4039 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
4040 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
4041 * doc/tm.texi: Regenerate.
4042 * gimple-expr.c: Include target.h.
4043 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
4044 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
4046 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
4047 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
4048 Use the original predicate if it already has a suitable type.
4050 2020-01-09 Martin Jambor <mjambor@suse.cz>
4052 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
4053 resolve_speculation and redirect_call_stmt_to_callee static. Change
4054 return type of set_call_stmt to cgraph_edge *.
4055 * auto-profile.c (afdo_indirect_call): Adjust call to
4056 redirect_call_stmt_to_callee.
4057 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
4058 make the this pointer explicit, adjust self-recursive calls and the
4059 call top make_direct. Return the resulting edge.
4060 (cgraph_edge::remove): Make this pointer explicit.
4061 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
4062 (cgraph_edge::make_direct): Likewise, adjust call to
4063 resolve_speculation.
4064 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
4065 call to set_call_stmt.
4066 (cgraph_update_edges_for_call_stmt_node): Update call to
4067 set_call_stmt and remove.
4068 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
4069 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
4070 (cgraph_node::create_edge_including_clones): Moved "first" definition
4071 of edge to the block where it was used. Adjusted calls to
4073 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
4074 cgraph_edge::remove.
4075 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
4076 make_direct and redirect_call_stmt_to_callee.
4077 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
4078 resolve_speculation and make_direct.
4079 * ipa-inline-transform.c (inline_transform): Adjust call to
4080 redirect_call_stmt_to_callee.
4081 (check_speculations_1):: Adjust call to resolve_speculation.
4082 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
4083 resolve-speculation.
4084 (inline_small_functions): Adjust call to resolve_speculation.
4085 (ipa_inline): Likewise.
4086 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
4088 * ipa-visibility.c (function_and_variable_visibility): Make iteration
4089 safe with regards to edge removal, adjust calls to
4090 redirect_call_stmt_to_callee.
4091 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
4092 and redirect_call_stmt_to_callee.
4093 * multiple_target.c (create_dispatcher_calls): Adjust call to
4094 redirect_call_stmt_to_callee
4095 (redirect_to_specific_clone): Likewise.
4096 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
4097 Adjust calls to cgraph_edge::remove.
4098 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
4099 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
4100 (expand_call_inline): Adjust call to cgraph_edge::remove.
4102 2020-01-09 Martin Liska <mliska@suse.cz>
4104 * params.opt: Set Optimization for
4105 param_max_speculative_devirt_maydefs.
4107 2020-01-09 Martin Sebor <msebor@redhat.com>
4111 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
4113 2020-01-09 Martin Liska <mliska@suse.cz>
4115 * auto-profile.c (auto_profile): Use opt_for_fn
4117 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
4118 (propagate_vals_across_arith_jfunc): Likewise.
4119 (hint_time_bonus): Likewise.
4120 (incorporate_penalties): Likewise.
4121 (good_cloning_opportunity_p): Likewise.
4122 (perform_estimation_of_a_value): Likewise.
4123 (estimate_local_effects): Likewise.
4124 (ipcp_propagate_stage): Likewise.
4125 * ipa-fnsummary.c (decompose_param_expr): Likewise.
4126 (set_switch_stmt_execution_predicate): Likewise.
4127 (analyze_function_body): Likewise.
4128 * ipa-inline-analysis.c (offline_size): Likewise.
4129 * ipa-inline.c (early_inliner): Likewise.
4130 * ipa-prop.c (ipa_analyze_node): Likewise.
4131 (ipcp_transform_function): Likewise.
4132 * ipa-sra.c (process_scan_results): Likewise.
4133 (ipa_sra_summarize_function): Likewise.
4134 * params.opt: Rename ipcp-unit-growth to
4135 ipa-cp-unit-growth. Add Optimization for various
4136 IPA-related parameters.
4138 2020-01-09 Richard Biener <rguenther@suse.de>
4141 * gimplify.c (gimplify_expr): Deal with NOP definitions.
4143 2020-01-09 Richard Biener <rguenther@suse.de>
4145 PR tree-optimization/93040
4146 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
4148 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
4150 * common/config/avr/avr-common.c (avr_option_optimization_table)
4151 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
4153 2020-01-09 Martin Liska <mliska@suse.cz>
4155 * cgraphclones.c (symbol_table::materialize_all_clones):
4156 Use cgraph_node::dump_name.
4158 2020-01-09 Jakub Jelinek <jakub@redhat.com>
4161 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
4162 output_operand_lossage instead of gcc_unreachable.
4163 * doc/md.texi (riscv f constraint): Fix typo.
4166 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
4167 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
4168 CONST_SCALAR_INT_P instead of CONST_INT_P.
4169 (*subv<mode>4_1): Rename to ...
4170 (subv<mode>4_1): ... this.
4171 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4172 define_insn_and_split patterns.
4173 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4176 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4178 * vec.c (class selftest::count_dtor): New class.
4179 (selftest::test_auto_delete_vec): New test.
4180 (selftest::vec_c_tests): Call it.
4181 * vec.h (class auto_delete_vec): New class template.
4182 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
4184 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4186 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
4188 2020-01-08 Jim Wilson <jimw@sifive.com>
4190 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
4191 use of TLS_MODEL_LOCAL_EXEC when not pic.
4193 2020-01-08 David Malcolm <dmalcolm@redhat.com>
4195 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
4198 2020-01-08 Jakub Jelinek <jakub@redhat.com>
4201 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
4202 *stack_protect_set_3 peephole2): Also check that the second
4203 insns source is general_operand.
4206 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
4207 predicate for output operand instead of register_operand.
4208 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
4209 memory destination and non-memory operands[2].
4211 2020-01-08 Martin Liska <mliska@suse.cz>
4213 * cgraph.c (cgraph_node::dump): Use ::dump_name or
4214 ::dump_asm_name instead of (::name or ::asm_name).
4215 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
4216 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
4217 (analyze_functions): Likewise.
4218 (expand_all_functions): Likewise.
4219 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
4220 (propagate_bits_across_jump_function): Likewise.
4221 (dump_profile_updates): Likewise.
4222 (ipcp_store_bits_results): Likewise.
4223 (ipcp_store_vr_results): Likewise.
4224 * ipa-devirt.c (dump_targets): Likewise.
4225 * ipa-fnsummary.c (analyze_function_body): Likewise.
4226 * ipa-hsa.c (check_warn_node_versionable): Likewise.
4227 (process_hsa_functions): Likewise.
4228 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
4229 (set_alias_uids): Likewise.
4230 * ipa-inline-transform.c (save_inline_function_body): Likewise.
4231 * ipa-inline.c (recursive_inlining): Likewise.
4232 (inline_to_all_callers_1): Likewise.
4233 (ipa_inline): Likewise.
4234 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
4235 (ipa_propagate_frequency): Likewise.
4236 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
4237 (remove_described_reference): Likewise.
4238 * ipa-pure-const.c (worse_state): Likewise.
4239 (check_retval_uses): Likewise.
4240 (analyze_function): Likewise.
4241 (propagate_pure_const): Likewise.
4242 (propagate_nothrow): Likewise.
4243 (dump_malloc_lattice): Likewise.
4244 (propagate_malloc): Likewise.
4245 (pass_local_pure_const::execute): Likewise.
4246 * ipa-visibility.c (optimize_weakref): Likewise.
4247 (function_and_variable_visibility): Likewise.
4248 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
4249 (ipa_discover_variable_flags): Likewise.
4250 * lto-streamer-out.c (output_function): Likewise.
4251 (output_constructor): Likewise.
4252 * tree-inline.c (copy_bb): Likewise.
4253 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
4254 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
4256 2020-01-08 Richard Biener <rguenther@suse.de>
4259 * tree-eh.c (sink_clobbers): Update virtual operands for
4260 the first and last stmt only. Add a dry-run capability.
4261 (pass_lower_eh_dispatch::execute): Perform clobber sinking
4262 after CFG manipulations and in RPO order to catch all
4263 secondary opportunities reliably.
4265 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4268 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4270 2019-01-08 Richard Biener <rguenther@suse.de>
4273 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4274 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4275 virtual operand, also updating SSA use.
4276 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4277 Update stmt after resetting virtual operand.
4278 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4279 * gimple-iterator.c (gsi_remove): When not removing the stmt
4280 permanently do not delink immediate uses or mark the stmt modified.
4282 2020-01-08 Martin Liska <mliska@suse.cz>
4284 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4285 (ipa_call_context::estimate_size_and_time): Likewise.
4286 (inline_analyze_function): Likewise.
4288 2020-01-08 Martin Liska <mliska@suse.cz>
4290 * cgraph.c (cgraph_node::dump): Use systematically
4293 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4295 Add -nodevicespecs option for avr.
4298 * config/avr/avr.opt (-nodevicespecs): New driver option.
4299 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4300 "-specs=device-specs/..." if that option is not set.
4301 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4303 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4305 Implement 64-bit double functions for avr.
4308 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4309 --with-double-comparison.
4310 * doc/install.texi: Document them.
4311 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4312 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4313 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4314 * doc/invoke.texi (AVR Built-in Macros): Document them.
4315 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4316 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4317 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4319 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4322 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4323 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4324 when only building rm-profile multilibs.
4326 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4329 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4330 lattice for a value to check.
4331 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4332 finite propagation in self-recursive scc.
4334 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4336 * ipa-inline.c (caller_growth_limits): Restore the AND.
4338 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4340 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4341 (VEC_ALLREG_ALT): New iterator.
4342 (VEC_ALLREG_INT_MODE): New iterator.
4343 (VCMP_MODE): New iterator.
4344 (VCMP_MODE_INT): New iterator.
4345 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4346 (vec_cmp<u>v64qidi): New define_expand.
4347 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4348 (vec_cmpu<mode>di_exec): New define_expand.
4349 (vec_cmp<u>v64qidi_exec): New define_expand.
4350 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4351 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4352 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4353 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4354 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4355 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4356 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4357 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4358 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4359 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4361 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4362 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4364 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4366 * config/gcn/constraints.md (DA): Update description and match.
4368 (Db): New constraint.
4369 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4371 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4372 Implement 'Db' mixed immediate type.
4373 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4374 (addcv64si3_dup<exec_vcc>): Delete.
4375 (subcv64si3<exec_vcc>): Rework constraints.
4376 (addv64di3): Rework constraints.
4377 (addv64di3_exec): Rework constraints.
4378 (subv64di3): Rework constraints.
4379 (addv64di3_dup): Delete.
4380 (addv64di3_dup_exec): Delete.
4381 (addv64di3_zext): Rework constraints.
4382 (addv64di3_zext_exec): Rework constraints.
4383 (addv64di3_zext_dup): Rework constraints.
4384 (addv64di3_zext_dup_exec): Rework constraints.
4385 (addv64di3_zext_dup2): Rework constraints.
4386 (addv64di3_zext_dup2_exec): Rework constraints.
4387 (addv64di3_sext_dup2): Rework constraints.
4388 (addv64di3_sext_dup2_exec): Rework constraints.
4390 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4392 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4393 existing target checks.
4395 2020-01-07 Richard Biener <rguenther@suse.de>
4397 * doc/install.texi: Bump minimal supported MPC version.
4399 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4401 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4402 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4403 * langhooks.c: Include stor-layout.h.
4404 (lhd_simulate_enum_decl): New function.
4405 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4406 handle_arm_sve_h for the LTO frontend.
4407 (register_vector_type): Cope with null returns from pushdecl.
4409 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4411 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4412 (aarch64_sve::nvectors_if_data_type): Replace with...
4413 (aarch64_sve::builtin_type_p): ...this.
4414 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4415 (find_vector_type): Delete.
4416 (add_sve_type_attribute): New function.
4417 (lookup_sve_type_attribute): Likewise.
4418 (register_builtin_types): Add an "SVE type" attribute to each type.
4419 (register_tuple_type): Likewise.
4420 (svbool_type_p, nvectors_if_data_type): Delete.
4421 (mangle_builtin_type): Use lookup_sve_type_attribute.
4422 (builtin_type_p): Likewise. Add an overload that returns the
4423 number of constituent vector and predicate registers.
4424 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4425 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4426 instead of aarch64_sve_argument_p.
4427 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4428 (aarch64_pass_by_reference): Likewise.
4429 (aarch64_function_value_1): Likewise.
4430 (aarch64_return_in_memory): Likewise.
4431 (aarch64_layout_arg): Likewise.
4433 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4435 PR tree-optimization/93156
4436 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4437 least significant bit is always clear.
4439 PR tree-optimization/93118
4440 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4441 simplifier with two intermediate conversions.
4443 2020-01-07 Martin Liska <mliska@suse.cz>
4445 * params.opt: Add Optimization for various parameters.
4447 2020-01-07 Martin Liska <mliska@suse.cz>
4450 * doc/extend.texi: Explain cloning for target_clone
4453 2020-01-07 Martin Liska <mliska@suse.cz>
4455 PR tree-optimization/92860
4456 * common.opt: Make in Optimization option
4457 as it is affected by -O0, which is an Optimization
4459 * tree-inline.c (tree_inlinable_function_p):
4460 Use opt_for_fn for warn_inline.
4461 (expand_call_inline): Likewise.
4463 2020-01-07 Martin Liska <mliska@suse.cz>
4465 PR tree-optimization/92860
4466 * common.opt: Make flag_ree as optimization
4469 2020-01-07 Martin Liska <mliska@suse.cz>
4471 PR optimization/92860
4472 * params.opt: Mark param_min_crossjump_insns with Optimization
4475 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4477 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4478 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4480 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4482 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4483 helper function to return the valid addressing formats for a given
4484 hard register and mode.
4485 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4487 * config/rs6000/constraints.md (Q constraint): Update
4489 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4492 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4493 Use 'Q' for doing vector extract from memory.
4494 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4496 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4497 doing vector extract from memory.
4498 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4499 extract from memory.
4501 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4502 for the offset being 34-bits when -mcpu=future is used.
4504 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4506 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4507 instead of cmpib_comparison_operator in cmpib patterns.
4508 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4509 of cmpib_comparison_operator. Revise comment.
4511 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4513 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4514 in an IFN_DIV_POW2 node to be equal.
4516 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4518 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4519 (vect_check_scalar_mask): ...this.
4520 (vectorizable_store, vectorizable_load): Update call accordingly.
4521 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4522 argument in calls to conditional internal functions.
4524 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4526 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4527 '0' matching inputs.
4528 (subv64di3_exec): Likewise.
4530 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4532 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4533 * doc/md.texi (movstr): Likewise.
4535 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4537 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4540 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4542 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4544 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4545 to a temporary file and use move-if-change to update the real
4546 file where necessary.
4548 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4550 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4551 rather than Upa for CPY /M.
4553 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4555 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4558 2020-01-06 Martin Liska <mliska@suse.cz>
4560 PR tree-optimization/92860
4561 * params.opt: Mark param_max_combine_insns with Optimization
4564 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4567 * config/i386/i386.md (SWIDWI): New mode iterator.
4568 (DWI, dwi): Add TImode variants.
4569 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4570 <general_hilo_operand> instead of <general_operand>. Use
4571 CONST_SCALAR_INT_P instead of CONST_INT_P.
4572 (*addv<mode>4_1): Rename to ...
4573 (addv<mode>4_1): ... this.
4574 (QWI): New mode attribute.
4575 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4576 define_insn_and_split patterns.
4577 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4579 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4580 <general_hilo_operand> instead of <general_operand>.
4581 (*addcarry<mode>_1): New define_insn.
4582 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4584 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4586 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4587 Use "call" instead of "set".
4589 2020-01-03 Martin Jambor <mjambor@suse.cz>
4592 * ipa-cp.c (print_all_lattices): Skip functions without info.
4594 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4597 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4598 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4599 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4600 for 'e' simd clones.
4603 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4605 (mprefer-vector-width=): Add Save.
4606 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4607 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4608 (ix86_debug_options, ix86_function_specific_print): Adjust
4609 ix86_target_string callers.
4610 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4611 (ix86_valid_target_attribute_tree): Likewise.
4612 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4613 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4614 ix86_target_string caller.
4617 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4618 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4619 instead of gen_int_shift_amount + convert_modes.
4621 PR rtl-optimization/93088
4622 * loop-iv.c (find_single_def_src): Punt after looking through
4623 128 reg copies for regs with single definitions. Move definitions
4626 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4628 * config/arm/arm-c.c (arm_cpu_builtins): Define
4629 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4630 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4631 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4632 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4633 * config/arm/arm-tables.opt: Regenerated.
4634 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4635 arm_arch_i8mm and arm_arch_bf16 when enabled.
4636 * config/arm/arm.h (TARGET_I8MM): New macro.
4637 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4638 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4639 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4640 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4641 (v8_6_a_simd_variants): New.
4642 (v8_*_a_simd_variants): Add i8mm and bf16.
4643 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4645 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4648 * predict.c (compute_function_frequency): Don't call
4649 warn_function_cold on functions that already have cold attribute.
4651 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4654 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4655 COMDAT group function labels in .data.rel.ro.local section.
4656 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4659 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4660 comparison_operator in B and S integer comparisons. Likewise, use
4661 ordered_comparison_operator instead of cmpib_comparison_operator in
4663 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4665 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4667 Update copyright years.
4669 * gcc.c (process_command): Update copyright notice dates.
4670 * gcov-dump.c (print_version): Ditto.
4671 * gcov.c (print_version): Ditto.
4672 * gcov-tool.c (print_version): Ditto.
4673 * gengtype.c (create_file): Ditto.
4674 * doc/cpp.texi: Bump @copying's copyright year.
4675 * doc/cppinternals.texi: Ditto.
4676 * doc/gcc.texi: Ditto.
4677 * doc/gccint.texi: Ditto.
4678 * doc/gcov.texi: Ditto.
4679 * doc/install.texi: Ditto.
4680 * doc/invoke.texi: Ditto.
4682 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4684 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4687 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4689 PR tree-optimization/93098
4690 * match.pd (popcount): For shift amounts, use integer_onep
4691 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4692 tests. Make sure that precision is power of two larger than or equal
4693 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4694 instead of ULL suffixed constants. Formatting fixes.
4696 Copyright (C) 2020 Free Software Foundation, Inc.
4698 Copying and distribution of this file, with or without modification,
4699 are permitted in any medium without royalty provided the copyright
4700 notice and this notice are preserved.