1 2020-02-21 Jakub Jelinek <jakub@redhat.com>
4 * opts.c (print_filtered_help): Translate help before appending
5 messages to it rather than after that.
7 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
9 PR rtl-optimization/PR92989
10 * lra-lives.c (process_bb_lives): Restore the original order
11 of the bb liveness update. Call make_hard_regno_dead for each
12 register clobbered at the start of an EH receiver.
14 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
17 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
18 self-recursively generated.
20 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
23 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
26 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
28 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
29 Document new target supports option.
31 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
33 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
34 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
35 * config/arm/iterators.md (MATMUL): New iterator.
36 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
37 (mmla_sfx): New attribute.
38 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
39 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
40 (UNSPEC_MATMUL_US): New.
42 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
44 * config/arm/arm.md: Prevent scalar shifts from being used when big
47 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
48 Richard Biener <rguenther@suse.de>
50 PR tree-optimization/93586
51 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
52 after mismatched array refs; do not sure type size information to
53 recover from unmatched referneces with !flag_strict_aliasing_p.
55 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
57 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
58 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
59 (scatter_store<mode>): Rename to ...
60 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
61 (scatter<mode>_exec): Delete. Move contents ...
62 (mask_scatter_store<mode>): ... here, and rename that to ...
63 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
64 Remove mode conversion.
65 (mask_gather_load<mode>): Rename to ...
66 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
67 Remove mode conversion.
68 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
70 2020-02-21 Martin Jambor <mjambor@suse.cz>
72 PR tree-optimization/93845
73 * tree-sra.c (verify_sra_access_forest): Only test access size of
76 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
78 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
79 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
80 (addv64di3_exec): Likewise.
81 (subv64di3): Likewise.
82 (subv64di3_exec): Likewise.
83 (addv64di3_zext): Likewise.
84 (addv64di3_zext_exec): Likewise.
85 (addv64di3_zext_dup): Likewise.
86 (addv64di3_zext_dup_exec): Likewise.
87 (addv64di3_zext_dup2): Likewise.
88 (addv64di3_zext_dup2_exec): Likewise.
89 (addv64di3_sext_dup2): Likewise.
90 (addv64di3_sext_dup2_exec): Likewise.
91 (<expander>v64di3): Likewise.
92 (<expander>v64di3_exec): Likewise.
93 (*<reduc_op>_dpp_shr_v64di): Likewise.
94 (*plus_carry_dpp_shr_v64di): Likewise.
95 * config/gcn/gcn.md (adddi3): Likewise.
96 (addptrdi3): Likewise.
97 (<expander>di3): Likewise.
99 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
101 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
103 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
105 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
106 support. Use aarch64_emit_mult instead of emitting multiplication
107 instructions directly.
108 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
109 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
111 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
113 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
114 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
115 instead of emitting multiplication instructions directly.
116 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
117 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
118 (@aarch64_frecps<mode>): New expanders.
120 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
122 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
123 on and produce uint64_ts rather than ints.
124 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
125 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
127 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
129 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
130 an unused xmsk register when handling approximate rsqrt.
132 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
134 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
135 flag_finite_math_only condition.
137 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
140 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
141 to destination operand for shufps alternative.
142 (*vec_extractv2si_1): Ditto.
144 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
147 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
150 2020-02-20 Martin Liska <mliska@suse.cz>
153 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
155 2020-02-20 Martin Liska <mliska@suse.cz>
158 * common/config/avr/avr-common.c: Remote trailing "|".
160 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
162 * collect2.c (maybe_run_lto_and_relink): Fix typo in
165 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
167 PR tree-optimization/93767
168 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
169 access-size bias from the offset calculations for negative strides.
171 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
173 * collect2.c (c_file, o_file): Make const again.
174 (ldout,lderrout, dump_ld_file): Remove.
175 (tool_cleanup): Avoid calling not signal-safe functions.
176 (maybe_run_lto_and_relink): Avoid possible signal handler
177 access to unintialzed memory (lto_o_files).
178 (main): Avoid leaking temp files in $TMPDIR.
179 Initialize c_file/o_file with concat, which avoids exposing
180 uninitialized memory to signal handler, which calls unlink(!).
181 Avoid calling maybe_unlink when the main function returns,
182 since the atexit handler is already doing this.
183 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
185 2020-02-19 Martin Jambor <mjambor@suse.cz>
187 PR tree-optimization/93776
188 * tree-sra.c (create_access): Do not create zero size accesses.
189 (get_access_for_expr): Do not search for zero sized accesses.
191 2020-02-19 Martin Jambor <mjambor@suse.cz>
193 PR tree-optimization/93667
194 * tree-sra.c (scalarizable_type_p): Return false if record fields
195 do not follow wach other.
197 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
199 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
200 rather than fmv.x.s/fmv.s.x.
202 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
204 * config/aarch64/aarch64-simd-builtins.def
205 (intrinsic_vec_smult_lo_): New.
206 (intrinsic_vec_umult_lo_): Likewise.
207 (vec_widen_smult_hi_): Likewise.
208 (vec_widen_umult_hi_): Likewise.
209 * config/aarch64/aarch64-simd.md
210 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
211 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
212 (vmull_high_s16): Likewise.
213 (vmull_high_s32): Likewise.
214 (vmull_high_u8): Likewise.
215 (vmull_high_u16): Likewise.
216 (vmull_high_u32): Likewise.
217 (vmull_s8): Likewise.
218 (vmull_s16): Likewise.
219 (vmull_s32): Likewise.
220 (vmull_u8): Likewise.
221 (vmull_u16): Likewise.
222 (vmull_u32): Likewise.
224 2020-02-18 Martin Liska <mliska@suse.cz>
226 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
227 bootstrap by missing removal of invalid sanity check.
229 2020-02-18 Martin Liska <mliska@suse.cz>
232 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
233 Always compare LHS of gimple_assign.
235 2020-02-18 Martin Liska <mliska@suse.cz>
238 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
239 and return type of functions.
240 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
241 Drop MALLOC attribute for void functions.
242 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
243 malloc_state for a new VOID clone.
245 2020-02-18 Martin Liska <mliska@suse.cz>
248 * common.opt: Add -fprofile-reproducibility.
249 * doc/invoke.texi: Document it.
250 * value-prof.c (dump_histogram_value):
251 Document and support behavior for counters[0]
252 being a negative value.
253 (get_nth_most_common_value): Handle negative
254 counters[0] in respect to flag_profile_reproducible.
256 2020-02-18 Jakub Jelinek <jakub@redhat.com>
259 * cgraph.c (verify_speculative_call): Use speculative_id instead of
260 speculative_uid in messages. Remove trailing whitespace from error
261 message. Use num_speculative_call_targets instead of
262 num_speculative_targets in a message.
263 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
264 edge messages and stmt instead of cal_stmt in reference message.
266 PR tree-optimization/93780
267 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
268 before calling build_vector_type.
269 (execute_update_addresses_taken): Likewise.
272 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
273 typo, functoin -> function.
274 * tree.c (free_lang_data_in_decl): Fix comment typo,
275 functoin -> function.
276 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
278 2020-02-17 David Malcolm <dmalcolm@redhat.com>
280 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
282 (print_option_information): Don't call get_option_url if URLs
285 2020-02-17 Alexandre Oliva <oliva@adacore.com>
287 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
288 handling of register_common-less targets.
290 2020-02-17 Martin Liska <mliska@suse.cz>
293 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
295 2020-02-17 Martin Liska <mliska@suse.cz>
298 * config/rs6000/rs6000.c (rs6000_option_override_internal):
301 2020-02-17 Martin Liska <mliska@suse.cz>
304 * config/rx/elf.opt: Fix typo.
306 2020-02-17 Richard Biener <rguenther@suse.de>
309 * opts-global.c (print_ignored_options): Use inform and
312 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
315 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
317 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
320 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
321 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
323 2020-02-15 Jason Merrill <jason@redhat.com>
325 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
327 2020-02-15 Jakub Jelinek <jakub@redhat.com>
329 PR tree-optimization/93744
330 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
331 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
332 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
333 sure @2 in the first and @1 in the other patterns has no side-effects.
335 2020-02-15 David Malcolm <dmalcolm@redhat.com>
336 Bernd Edlinger <bernd.edlinger@hotmail.de>
340 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
341 * configure.ac (--with-diagnostics-urls): New configuration
342 option, based on --with-diagnostics-color.
343 (DIAGNOSTICS_URLS_DEFAULT): New define.
344 * config.h: Regenerate.
345 * configure: Regenerate.
346 * diagnostic.c (diagnostic_urls_init): Handle -1 for
347 DIAGNOSTICS_URLS_DEFAULT from configure-time
348 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
349 and TERM_URLS environment variable.
350 * diagnostic-url.h (diagnostic_url_format): New enum type.
351 (diagnostic_urls_enabled_p): rename to...
352 (determine_url_format): ... this, and change return type.
353 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
354 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
355 the linux console, and mingw.
356 (diagnostic_urls_enabled_p): rename to...
357 (determine_url_format): ... this, and adjust.
358 * pretty-print.h (pretty_printer::show_urls): rename to...
359 (pretty_printer::url_format): ... this, and change to enum.
360 * pretty-print.c (pretty_printer::pretty_printer,
361 pp_begin_url, pp_end_url, test_urls): Adjust.
362 * doc/install.texi (--with-diagnostics-urls): Document the new
363 configuration option.
364 (--with-diagnostics-color): Document the existing interaction
365 with GCC_COLORS better.
366 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
367 vindex reference. Update description of defaults based on the above.
368 (-fdiagnostics-color): Update description of how -fdiagnostics-color
369 interacts with GCC_COLORS.
371 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
374 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
375 conjunction with TARGET_GNU_TLS in early return.
377 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
379 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
380 the mode is not wider than UNITS_PER_WORD.
382 2020-02-14 Martin Jambor <mjambor@suse.cz>
384 PR tree-optimization/93516
385 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
386 access of the same type as the parent.
387 (propagate_subaccesses_from_lhs): Likewise.
389 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
392 * config/i386/avx512vbmi2intrin.h
393 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
394 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
395 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
396 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
397 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
398 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
399 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
400 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
401 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
402 of lacking a closing parenthesis.
403 * config/i386/avx512vbmi2vlintrin.h
404 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
405 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
406 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
407 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
408 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
409 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
410 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
411 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
412 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
413 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
414 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
415 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
416 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
417 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
418 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
419 _mm_shldi_epi32, _mm_mask_shldi_epi32,
420 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
421 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
423 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
426 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
427 the target function entry.
429 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
431 * common/config/arc/arc-common.c (arc_option_optimization_table):
432 Disable if-conversion step when optimized for size.
434 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
436 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
437 R12-R15 are always in ARCOMPACT16_REGS register class.
438 * config/arc/arc.opt (mq-class): Deprecate.
439 * config/arc/constraint.md ("q"): Remove dependency on mq-class
441 * doc/invoke.texi (mq-class): Update text.
442 * common/config/arc/arc-common.c (arc_option_optimization_table):
445 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
447 * config/arc/arc.c (arc_insn_cost): New function.
448 (TARGET_INSN_COST): Define.
449 * config/arc/arc.md (cost): New attribute.
450 (add_n): Use arc_nonmemory_operand.
451 (ashlsi3_insn): Likewise, also update constraints.
452 (ashrsi3_insn): Likewise.
454 (add_shift): Likewise.
455 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
457 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
459 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
461 (umulsidi_600): Likewise.
463 2020-02-13 Jakub Jelinek <jakub@redhat.com>
466 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
467 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
468 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
469 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
470 pass __A to the builtin followed by __W instead of __A followed by
472 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
473 _mm512_mask_popcnt_epi64): Likewise.
474 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
475 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
476 _mm256_mask_popcnt_epi64): Likewise.
478 PR tree-optimization/93582
479 * fold-const.h (shift_bytes_in_array_left,
480 shift_bytes_in_array_right): Declare.
481 * fold-const.c (shift_bytes_in_array_left,
482 shift_bytes_in_array_right): New function, moved from
483 gimple-ssa-store-merging.c, no longer static.
484 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
485 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
486 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
487 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
488 shift_bytes_in_array.
489 (verify_shift_bytes_in_array): Rename to ...
490 (verify_shift_bytes_in_array_left): ... this. Use
491 shift_bytes_in_array_left instead of shift_bytes_in_array.
492 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
493 instead of verify_shift_bytes_in_array.
494 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
495 / native_interpret_expr where the store covers all needed bits,
496 punt on PDP-endian, otherwise allow all involved offsets and sizes
497 not to be byte-aligned.
500 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
501 use const_0_to_255_operand predicate instead of immediate_operand.
502 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
503 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
504 vgf2p8affineinvqb_<mode><mask_name>,
505 vgf2p8affineqb_<mode><mask_name>): Drop mode from
506 const_0_to_255_operand predicated operands.
508 2020-02-12 Jeff Law <law@redhat.com>
510 * config/h8300/h8300.md (comparison shortening peepholes): Use
511 a mode iterator to merge the HImode and SImode peepholes.
513 2020-02-12 Jakub Jelinek <jakub@redhat.com>
516 * real.c (is_even): Make static. Function comment fix.
517 (is_halfway_below): Make static, don't assert R is not inf/nan,
518 instead return false for those. Small formatting fixes.
520 2020-02-12 Martin Sebor <msebor@redhat.com>
523 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
524 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
525 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
526 (strlen_check_and_optimize_call): Adjust callee name.
528 2020-02-12 Jeff Law <law@redhat.com>
530 * config/h8300/h8300.md (comparison shortening peepholes): Drop
531 (and (xor)) variant. Combine other two into single peephole.
533 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
535 PR rtl-optimization/93565
536 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
538 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
540 * config/aarch64/aarch64-simd.md
541 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
542 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
543 generating separate ADDV and zero_extend patterns.
544 * config/aarch64/iterators.md (VDQV_E): New iterator.
546 2020-02-12 Jeff Law <law@redhat.com>
548 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
549 expanders, splits, etc.
550 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
551 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
552 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
553 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
554 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
555 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
557 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
559 2020-02-12 Jakub Jelinek <jakub@redhat.com>
562 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
563 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
564 TARGET_AVX512DQ from condition.
565 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
566 instead of <mask_mode512bit_condition> in condition. If
567 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
569 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
572 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
575 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
577 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
579 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
580 where strlen is more legible.
581 (rs6000_builtin_vectorized_libmass): Ditto.
582 (rs6000_print_options_internal): Ditto.
584 2020-02-11 Martin Sebor <msebor@redhat.com>
586 PR tree-optimization/93683
587 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
589 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
591 * config/rs6000/predicates.md (cint34_operand): Rename the
592 -mprefixed-addr option to be -mprefixed.
593 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
594 the -mprefixed-addr option to be -mprefixed.
595 (OTHER_FUTURE_MASKS): Likewise.
596 (POWERPC_MASKS): Likewise.
597 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
598 the -mprefixed-addr option to be -mprefixed. Change error
599 messages to refer to -mprefixed.
600 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
602 (rs6000_legitimate_offset_address_p): Likewise.
603 (rs6000_mode_dependent_address): Likewise.
604 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
605 "-mprefixed" for target attributes and pragmas.
606 (address_to_insn_form): Rename the -mprefixed-addr option to be
608 (rs6000_adjust_insn_length): Likewise.
609 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
610 -mprefixed-addr option to be -mprefixed.
611 (ASM_OUTPUT_OPCODE): Likewise.
612 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
613 -mprefixed-addr option to be -mprefixed.
614 * config/rs6000/rs6000.opt (-mprefixed): Rename the
615 -mprefixed-addr option to be prefixed. Change the option from
616 being undocumented to being documented.
617 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
618 -mprefixed option. Update the -mpcrel documentation to mention
621 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
623 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
624 including FIRST_PSEUDO_REGISTER - 1.
625 * ira-color.c (print_hard_reg_set): Ditto.
627 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
629 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
630 (USTERNOP_QUALIFIERS): New define.
631 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
632 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
633 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
634 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
635 * config/arm/arm_neon.h (vusdot_s32): New.
636 (vusdot_lane_s32): New.
637 (vusdotq_lane_s32): New.
638 (vsudot_lane_s32): New.
639 (vsudotq_lane_s32): New.
640 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
641 * config/arm/iterators.md (DOTPROD_I8MM): New.
642 (sup, opsuffix): Add <us/su>.
643 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
644 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
646 2020-02-11 Richard Biener <rguenther@suse.de>
648 PR tree-optimization/93661
649 PR tree-optimization/93662
650 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
652 * tree-sra.c (get_access_for_expr): Likewise.
654 2020-02-10 Jakub Jelinek <jakub@redhat.com>
657 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
658 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
659 Change condition from TARGET_AVX2 to TARGET_AVX.
661 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
664 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
667 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
669 Try to generate zero-based comparisons.
670 * config/cris/cris.c (cris_reduce_compare): New function.
671 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
672 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
673 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
675 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
678 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
679 in Thumb state and also as a destination in Arm state. Add T16
682 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
684 * md.texi (Define Subst): Match closing paren in example.
686 2020-02-10 Jakub Jelinek <jakub@redhat.com>
690 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
691 arguments of strncmp.
693 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
696 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
697 but different source value.
698 (adjust_callers_for_value_intersection): New function.
699 (gather_edges_for_value): Adjust order of callers to let a
700 non-self-recursive caller be the first element.
701 (self_recursive_pass_through_p): Add a new parameter "simple", and
702 check generalized self-recursive pass-through jump function.
703 (self_recursive_agg_pass_through_p): Likewise.
704 (find_more_scalar_values_for_callers_subset): Compute value from
705 pass-through jump function for self-recursive.
706 (intersect_with_plats): Cleanup previous implementation code for value
707 itersection with self-recursive call edge.
708 (intersect_with_agg_replacements): Likewise.
709 (intersect_aggregates_with_edge): Deduce value from pass-through jump
710 function for self-recursive call edge. Cleanup previous implementation
711 code for value intersection with self-recursive call edge.
712 (decide_whether_version_node): Remove dead callers and adjust order
713 to let a non-self-recursive caller be the first element.
715 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
717 * recog.c: Move pass_split_before_sched2 code in front of
718 pass_split_before_regstack.
719 (pass_data_split_before_sched2): Rename pass to split3 from split4.
720 (pass_data_split_before_regstack): Rename pass to split4 from split3.
721 (rest_of_handle_split_before_sched2): Remove.
722 (pass_split_before_sched2::execute): Unconditionally call
724 (enable_split_before_sched2): New function.
725 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
726 (pass_split_before_regstack::gate): Ditto.
727 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
728 Update name check for renamed split4 pass.
729 * config/sh/sh.c (register_sh_passes): Update pass insertion
730 point for renamed split4 pass.
732 2020-02-09 Jakub Jelinek <jakub@redhat.com>
734 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
735 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
736 copying them around between host and target.
738 2020-02-08 Andrew Pinski <apinski@marvell.com>
741 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
742 STRICT_ALIGNMENT also.
744 2020-02-08 Jim Wilson <jimw@sifive.com>
747 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
749 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
750 Jakub Jelinek <jakub@redhat.com>
753 * config/i386/i386.h (CALL_USED_REGISTERS): Make
754 xmm16-xmm31 call-used even in 64-bit ms-abi.
756 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
758 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
759 (simd_ummla, simd_usmmla): Likewise.
760 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
761 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
764 2020-02-07 Richard Biener <rguenther@suse.de>
767 * tree-inline.c (fold_marked_statements): Do a PRE walk,
768 skipping unreachable regions.
769 (optimize_inline_calls): Skip folding stmts when we didn't
772 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
775 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
776 Don't return aggregates with only SFmode and DFmode in SSE
778 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
780 2020-02-07 Jakub Jelinek <jakub@redhat.com>
783 * config/rs6000/rs6000-logue.c
784 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
785 if it fails, move rs into end_addr and retry. Add
786 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
787 the insn pattern doesn't describe well what exactly happens to
791 * config/i386/predicates.md (avx_identity_operand): Remove.
792 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
793 (avx_<castmode><avxsizesuffix>_<castmode>,
794 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
795 a VEC_CONCAT of the operand and UNSPEC_CAST.
796 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
797 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
801 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
802 recog_data.insn if distance_non_agu_define changed it.
804 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
807 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
808 we only had X-FORM (reg+reg) addressing for vectors. Also before
809 ISA 3.0, we only had X-FORM addressing for scalars in the
810 traditional Altivec registers.
812 2020-02-06 <zhongyunde@huawei.com>
813 Vladimir Makarov <vmakarov@redhat.com>
815 PR rtl-optimization/93561
816 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
819 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
821 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
824 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
826 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
827 where the low and the high 32 bits are equal to each other specially,
828 with an rldimi instruction.
830 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
832 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
834 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
836 * config/arm/arm-tables.opt: Regenerate.
838 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
841 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
842 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
843 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
845 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
847 PR rtl-optimization/87763
848 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
850 2020-02-06 Delia Burduv <delia.burduv@arm.com>
852 * config/aarch64/aarch64-simd-builtins.def
853 (bfmlaq): New built-in function.
854 (bfmlalb): New built-in function.
855 (bfmlalt): New built-in function.
856 (bfmlalb_lane): New built-in function.
857 (bfmlalt_lane): New built-in function.
858 * config/aarch64/aarch64-simd.md
859 (aarch64_bfmmlaqv4sf): New pattern.
860 (aarch64_bfmlal<bt>v4sf): New pattern.
861 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
862 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
863 (vbfmlalbq_f32): New intrinsic.
864 (vbfmlaltq_f32): New intrinsic.
865 (vbfmlalbq_lane_f32): New intrinsic.
866 (vbfmlaltq_lane_f32): New intrinsic.
867 (vbfmlalbq_laneq_f32): New intrinsic.
868 (vbfmlaltq_laneq_f32): New intrinsic.
869 * config/aarch64/iterators.md (BF_MLA): New int iterator.
870 (bt): New int attribute.
872 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
874 * config/i386/i386.md (*pushtf): Emit "#" instead of
875 calling gcc_unreachable in insn output.
878 (*pushsf_rex64): Ditto for alternatives other than 1.
879 (*pushsf): Ditto for alternatives other than 1.
881 2020-02-06 Martin Liska <mliska@suse.cz>
883 PR gcov-profile/91971
884 PR gcov-profile/93466
885 * coverage.c (coverage_init): Revert mangling of
886 path into filename. It can lead to huge filename length.
887 Creation of subfolders seem more natural.
889 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
892 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
893 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
894 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
896 2020-02-06 Jakub Jelinek <jakub@redhat.com>
899 * config/i386/predicates.md (avx_identity_operand): New predicate.
900 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
901 define_insn_and_split.
904 * omp-low.c (use_pointer_for_field): For nested constructs, also
905 look for map clauses on target construct.
906 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
907 taskreg_nesting_level.
910 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
911 shared clause, call omp_notice_variable on outer context if any.
913 2020-02-05 Jason Merrill <jason@redhat.com>
916 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
917 non-zero address even if weak and not yet defined.
919 2020-02-05 Martin Sebor <msebor@redhat.com>
921 PR tree-optimization/92765
922 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
923 * tree-ssa-strlen.c (compute_string_length): Remove.
924 (determine_min_objsize): Remove.
925 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
926 Avoid using type size as the upper bound on string length.
927 (handle_builtin_string_cmp): Add an argument. Adjust.
928 (strlen_check_and_optimize_call): Pass additional argument to
929 handle_builtin_string_cmp.
931 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
933 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
934 (*pushdi2_rex64 peephole2): Unconditionally split after
936 (*ashl<mode>3_doubleword): Ditto.
937 (*<shift_insn><mode>3_doubleword): Ditto.
939 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
942 * config/rs6000/rs6000.c (get_vector_offset): Fix
944 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
946 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
948 2020-02-05 David Malcolm <dmalcolm@redhat.com>
951 (Special Functions for Debugging the Analyzer): Update description
952 of __analyzer_dump_exploded_nodes.
954 2020-02-05 Jakub Jelinek <jakub@redhat.com>
957 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
958 include sets and not clobbers in the vzeroupper pattern.
959 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
960 the parallel has 17 (64-bit) or 9 (32-bit) elts.
961 (*avx_vzeroupper_1): New define_insn_and_split.
964 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
965 don't run when !optimize.
966 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
969 2020-02-05 Richard Biener <rguenther@suse.de>
972 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
973 checks before matching calls.
975 2020-02-05 Jakub Jelinek <jakub@redhat.com>
977 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
978 function comment typo.
981 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
982 simd_clone_create failed when i == 0, adjust clone->nargs by
985 2020-02-05 Martin Liska <mliska@suse.cz>
988 * doc/invoke.texi: Document that one should
989 not combine ASLR and -fpch.
991 2020-02-04 Richard Biener <rguenther@suse.de>
993 PR tree-optimization/93538
994 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
996 2020-02-04 Richard Biener <rguenther@suse.de>
998 PR tree-optimization/91123
999 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
1000 (vn_walk_cb_data::last_vuse): New member.
1001 (vn_walk_cb_data::saved_operands): Likewsie.
1002 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
1003 (vn_walk_cb_data::push_partial_def): Use finish.
1004 (vn_reference_lookup_2): Update last_vuse and use finish if
1005 we've saved operands.
1006 (vn_reference_lookup_3): Use finish and update calls to
1007 push_partial_defs everywhere. When translating through
1008 memcpy or aggregate copies save off operands and alias-set.
1009 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
1010 operation for redundant store removal.
1012 2020-02-04 Richard Biener <rguenther@suse.de>
1014 PR tree-optimization/92819
1015 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1016 generating more stmts than before.
1018 2020-02-04 Martin Liska <mliska@suse.cz>
1020 * config/arm/arm.c (arm_gen_far_branch): Move the function
1021 outside of selftests.
1023 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1025 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1026 function to adjust PC-relative vector addresses.
1027 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1028 handle vectors with PC-relative addresses.
1030 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1032 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1034 (hard_reg_and_mode_to_addr_mask): Delete.
1035 (rs6000_adjust_vec_address): If the original vector address
1036 was REG+REG or REG+OFFSET and the element is not zero, do the add
1037 of the elements in the original address before adding the offset
1038 for the vector element. Use address_to_insn_form to validate the
1039 address using the register being loaded, rather than guessing
1040 whether the address is a DS-FORM or DQ-FORM address.
1042 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1044 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1045 to calculate the offset in memory from the start of a vector of a
1046 particular element. Add code to keep the element number in
1047 bounds if the element number is variable.
1048 (rs6000_adjust_vec_address): Move calculation of offset of the
1049 vector element to get_vector_offset.
1050 (rs6000_split_vec_extract_var): Do not do the initial AND of
1051 element here, move the code to get_vector_offset.
1053 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1055 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1058 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1060 * config/rs6000/constraints.md: Improve documentation.
1062 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1065 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1066 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1068 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1070 * config.gcc: Remove "carrizo" support.
1071 * config/gcn/gcn-opts.h (processor_type): Likewise.
1072 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1073 * config/gcn/gcn.opt (gpu_type): Likewise.
1074 * config/gcn/t-omp-device: Likewise.
1076 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1079 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1080 * config/arm/arm.c (arm_gen_far_branch): New function
1082 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1084 2020-02-03 Julian Brown <julian@codesourcery.com>
1085 Tobias Burnus <tobias@codesourcery.com>
1087 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1089 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1092 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1093 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1096 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1098 PR rtl-optimization/91333
1099 * ira-color.c (struct allocno_color_data): Add member
1101 (init_allocno_threads): Set the member up.
1102 (bucket_allocno_compare_func): Add compare hard reg
1105 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1107 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1109 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1110 * config.in: Regenerated.
1111 * configure: Regenerated.
1112 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1113 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1114 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1116 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1118 * configure: Regenerate.
1120 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1122 PR rtl-optimization/91333
1123 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1124 reg preferences comparison up.
1126 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1128 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1129 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1130 aarch64-sve-builtins-base.h.
1131 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1132 aarch64-sve-builtins-base.cc.
1133 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1134 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1136 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1137 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1138 (svcvtnt): New functions.
1139 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1140 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1141 (svcvtnt): New functions.
1142 (svcvt): Add a form that converts f32 to bf16.
1143 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1144 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1146 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1147 Treat B as bfloat16_t.
1148 (ternary_bfloat_lane_base): New class.
1149 (ternary_bfloat_def): Likewise.
1150 (ternary_bfloat): New shape.
1151 (ternary_bfloat_lane_def): New class.
1152 (ternary_bfloat_lane): New shape.
1153 (ternary_bfloat_lanex2_def): New class.
1154 (ternary_bfloat_lanex2): New shape.
1155 (ternary_bfloat_opt_n_def): New class.
1156 (ternary_bfloat_opt_n): New shape.
1157 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1158 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1159 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1160 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1161 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1162 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1163 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1164 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1165 the pattern off the narrow mode instead of the wider one.
1166 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1167 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1168 (sve_fp_op): Handle them.
1169 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1170 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1172 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1174 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1175 * config/aarch64/aarch64-modes.def (BF): Move definition before
1176 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1177 (SVE_MODES): Handle BF modes.
1178 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1180 (aarch64_full_sve_mode): Likewise.
1181 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1183 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1184 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1185 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1186 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1188 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1190 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1191 (TYPES_all_data): Add bf16.
1192 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1193 (register_tuple_type): Increase buffer size.
1194 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1195 (bf16): New type suffix.
1196 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1197 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1198 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1199 Change type from all_data to all_arith.
1200 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1203 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1204 Matthew Malcomson <matthew.malcomson@arm.com>
1205 Richard Sandiford <richard.sandiford@arm.com>
1207 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1208 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1209 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1210 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1211 __ARM_FEATURE_MATMUL_FP64.
1212 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1213 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1214 be disabled at the same time.
1215 (f32mm): New extension.
1216 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1217 (AARCH64_FL_F64MM): Bump to the next bit up.
1218 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1219 (TARGET_SVE_F64MM): New macros.
1220 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1221 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1222 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1223 (UNSPEC_ZIP2Q): New unspeccs.
1224 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1225 (optab, sur, perm_insn): Handle the new unspecs.
1226 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1227 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1228 TARGET_SVE_F64MM instead of separate tests.
1229 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1230 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1231 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1232 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1233 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1234 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1235 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1236 (TYPES_s_signed): New macro.
1237 (TYPES_s_integer): Use it.
1238 (TYPES_d_float): New macro.
1239 (TYPES_d_data): Use it.
1240 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1241 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1242 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1243 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1244 (svmmla): New shape.
1245 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1246 template parameters.
1247 (ternary_resize2_lane_base): Likewise.
1248 (ternary_resize2_base): New class.
1249 (ternary_qq_lane_base): Likewise.
1250 (ternary_intq_uintq_lane_def): Likewise.
1251 (ternary_intq_uintq_lane): New shape.
1252 (ternary_intq_uintq_opt_n_def): New class
1253 (ternary_intq_uintq_opt_n): New shape.
1254 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1255 (ternary_uintq_intq_def): New class.
1256 (ternary_uintq_intq): New shape.
1257 (ternary_uintq_intq_lane_def): New class.
1258 (ternary_uintq_intq_lane): New shape.
1259 (ternary_uintq_intq_opt_n_def): New class.
1260 (ternary_uintq_intq_opt_n): New shape.
1261 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1262 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1263 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1264 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1266 (svdotprod_lane_impl): ...this new class.
1267 (svmmla_impl, svusdot_impl): New classes.
1268 (svdot_lane): Update to use svdotprod_lane_impl.
1269 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1270 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1272 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1273 function, with no types defined.
1274 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1275 AARCH64_FL_I8MM functions.
1276 (svmmla): New AARCH64_FL_F32MM function.
1277 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1278 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1279 AARCH64_FL_F64MM function.
1280 (REQUIRED_EXTENSIONS):
1282 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1284 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1287 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1289 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1290 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1291 (*movti_internal): Do not check for
1292 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1293 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1294 just after check for TARGET_AVX.
1295 (*movdf_internal): Ditto.
1296 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1297 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1298 * config/i386/sse.md (mov<mode>_internal): Only check
1299 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1300 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1301 (<sse>_andnot<mode>3<mask_name>): Move check for
1302 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1303 (<code><mode>3<mask_name>): Ditto.
1304 (*andnot<mode>3): Ditto.
1305 (*andnottf3): Ditto.
1306 (*<code><mode>3): Ditto.
1307 (*<code>tf3): Ditto.
1308 (*andnot<VI:mode>3): Remove
1309 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1310 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1311 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1312 (sse4_1_blendv<ssemodesuffix>): Ditto.
1313 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1314 Explain that tune applies to 128bit instructions only.
1316 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1318 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1319 to definition of hsa_kernel_description. Parse assembly to find SGPR
1320 and VGPR count of kernel and store in hsa_kernel_description.
1322 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1324 PR rtl-optimization/91838
1325 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1326 to truncate if allowed or reject combination.
1328 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1330 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1331 (find_inv_vars_cb): Likewise.
1333 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1335 * calls.c (special_function_p): Split out the check for DECL_NAME
1336 being non-NULL and fndecl being extern at file scope into a
1337 new maybe_special_function_p and call it. Drop check for fndecl
1338 being non-NULL that was after a usage of DECL_NAME (fndecl).
1339 * tree.h (maybe_special_function_p): New inline function.
1341 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1343 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1344 (mask_gather_load<mode>): ... here, and zero-initialize the
1346 (maskload<mode>di): Zero-initialize the destination.
1349 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1352 * doc/analyzer.texi (Limitations): Note that constraints on
1353 floating-point values are currently ignored.
1355 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1358 * symtab.c (symtab_node::noninterposable_alias): If localalias
1359 already exists, but is not usable, append numbers after it until
1360 a unique name is found. Formatting fix.
1363 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1366 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1368 * config/gcn/gcn.c (print_operand): Handle LTGT.
1369 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1371 2020-01-30 Richard Biener <rguenther@suse.de>
1373 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1374 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1376 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1378 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1379 without a DECL in .data.rel.ro.local.
1381 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1384 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1388 * config/i386/sse.md
1389 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1390 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1391 any_extend code iterator instead of always zero_extend.
1392 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1393 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1394 Use any_extend code iterator instead of always zero_extend.
1395 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1396 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1397 Use any_extend code iterator instead of always zero_extend.
1398 (*sse2_pmovmskb_ext): New define_insn.
1399 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1402 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1403 (*popcountsi2_zext_falsedep): New define_insn.
1405 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1407 * config.in: Regenerated.
1408 * configure: Regenerated.
1410 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1413 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1414 LLVM's assembler changed the default in version 9.
1416 2020-01-24 Jeff Law <law@redhat.com>
1418 PR tree-optimization/89689
1419 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1421 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1425 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1427 PR rtl-optimization/87763
1428 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1429 simplification to handle subregs as well as bare regs.
1430 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1432 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1435 * ira.c (ira): Revert use of simplified LRA algorithm.
1437 2020-01-29 Martin Jambor <mjambor@suse.cz>
1439 PR tree-optimization/92706
1440 * tree-sra.c (struct access): Fields first_link, last_link,
1441 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1442 next_rhs_queued and grp_rhs_queued respectively, new fields
1443 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1444 (struct assign_link): Field next renamed to next_rhs, new field
1445 next_lhs. Updated comment.
1446 (work_queue_head): Renamed to rhs_work_queue_head.
1447 (lhs_work_queue_head): New variable.
1448 (add_link_to_lhs): New function.
1449 (relink_to_new_repr): Also relink LHS lists.
1450 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1451 (add_access_to_lhs_work_queue): New function.
1452 (pop_access_from_work_queue): Renamed to
1453 pop_access_from_rhs_work_queue.
1454 (pop_access_from_lhs_work_queue): New function.
1455 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1457 (child_would_conflict_in_lacc): Renamed to
1458 child_would_conflict_in_acc. Adjusted parameter names.
1459 (create_artificial_child_access): New parameter set_grp_read, use it.
1460 (subtree_mark_written_and_enqueue): Renamed to
1461 subtree_mark_written_and_rhs_enqueue.
1462 (propagate_subaccesses_across_link): Renamed to
1463 propagate_subaccesses_from_rhs.
1464 (propagate_subaccesses_from_lhs): New function.
1465 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1468 2020-01-29 Martin Jambor <mjambor@suse.cz>
1470 PR tree-optimization/92706
1471 * tree-sra.c (struct access): Adjust comment of
1472 grp_total_scalarization.
1473 (find_access_in_subtree): Look for single children spanning an entire
1475 (scalarizable_type_p): Allow register accesses, adjust callers.
1476 (completely_scalarize): Remove function.
1477 (scalarize_elem): Likewise.
1478 (create_total_scalarization_access): Likewise.
1479 (sort_and_splice_var_accesses): Do not track total scalarization
1481 (analyze_access_subtree): New parameter totally, adjust to new meaning
1482 of grp_total_scalarization.
1483 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1484 (can_totally_scalarize_forest_p): New function.
1485 (create_total_scalarization_access): Likewise.
1486 (create_total_access_and_reshape): Likewise.
1487 (total_should_skip_creating_access): Likewise.
1488 (totally_scalarize_subtree): Likewise.
1489 (analyze_all_variable_accesses): Perform total scalarization after
1490 subaccess propagation using the new functions above.
1491 (initialize_constant_pool_replacements): Output initializers by
1492 traversing the access tree.
1494 2020-01-29 Martin Jambor <mjambor@suse.cz>
1496 * tree-sra.c (verify_sra_access_forest): New function.
1497 (verify_all_sra_access_forests): Likewise.
1498 (create_artificial_child_access): Set parent.
1499 (analyze_all_variable_accesses): Call the verifier.
1501 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1503 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1504 if called on indirect edge.
1505 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1506 speculative call if needed.
1508 2020-01-29 Richard Biener <rguenther@suse.de>
1510 PR tree-optimization/93428
1511 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1512 permutation when the load node is created.
1513 (vect_analyze_slp_instance): Re-use it here.
1515 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1517 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1519 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1521 PR rtl-optimization/93272
1522 * ira-lives.c (process_out_of_region_eh_regs): New function.
1523 (process_bb_node_lives): Call it.
1525 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1527 * coverage.c (read_counts_file): Make error message lowercase.
1529 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1531 * profile-count.c (profile_quality_display_names): Fix ordering.
1533 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1536 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1537 hash only when edge is first within the sequence.
1538 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1539 (symbol_table::create_edge): Do not set target_prob.
1540 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1542 (cgraph_edge::make_speculative): Drop target_prob parameter.
1543 (cgraph_edge::speculative_call_info): Remove.
1544 (cgraph_edge::first_speculative_call_target): New member function.
1545 (update_call_stmt_hash_for_removing_direct_edge): New function.
1546 (cgraph_edge::resolve_speculation): Rewrite to new API.
1547 (cgraph_edge::speculative_call_for_target): New member function.
1548 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1549 multiple speculation targets.
1550 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1552 (verify_speculative_call): Verify that targets form an interval.
1553 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1554 (cgraph_edge::first_speculative_call_target): New member function.
1555 (cgraph_edge::next_speculative_call_target): New member function.
1556 (cgraph_edge::speculative_call_target_ref): New member function.
1557 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1558 (cgraph_edge): Remove target_prob.
1559 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1560 Fix handling of speculative calls.
1561 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1562 * ipa-fnsummary.c (analyze_function_body): Likewise.
1563 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1564 * ipa-profile.c (dump_histogram): Fix formating.
1565 (ipa_profile_generate_summary): Watch for overflows.
1566 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1567 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1568 (update_indirect_edges_after_inlining): Update to new API.
1569 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1571 * profile-count.h: (profile_probability::adjusted): New.
1572 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1573 updating of profile.
1574 * value-prof.c (gimple_ic_transform): Rename to ...
1575 (dump_ic_profile): ... this one; update dumping.
1576 (stream_in_histogram_value): Fix formating.
1577 (gimple_value_profile_transformations): Update.
1579 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1582 * config/i386/i386.md (*movoi_internal_avx): Remove
1583 TARGET_SSE_TYPELESS_STORES check.
1584 (*movti_internal): Prefer TARGET_AVX over
1585 TARGET_SSE_TYPELESS_STORES.
1586 (*movtf_internal): Likewise.
1587 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1588 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1589 from TARGET_SSE_TYPELESS_STORES.
1591 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1593 * diagnostic-core.h (warning_at): Rename overload to...
1594 (warning_meta): ...this.
1595 (emit_diagnostic_valist): Delete decl of overload taking
1596 diagnostic_metadata.
1597 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1598 (warning_at): Rename overload taking diagnostic_metadata to...
1599 (warning_meta): ...this.
1601 2020-01-28 Richard Biener <rguenther@suse.de>
1603 PR tree-optimization/93439
1604 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1605 * tree-cfg.c (move_sese_region_to_fn): ... here.
1606 (verify_types_in_gimple_reference): Verify used cliques are
1609 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1612 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1613 argument of a pointer to struct gcc_options and pass it to
1614 parse_mtune_ctrl_str.
1615 (ix86_function_specific_restore): Pass opts to
1616 set_ix86_tune_features.
1617 (ix86_option_override_internal): Likewise.
1618 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1619 gcc_options and use it for x_ix86_tune_ctrl_string.
1621 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1623 PR rtl-optimization/87763
1624 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1625 simplification to handle subregs as well as bare regs.
1626 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1628 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1630 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1631 for reduction chains that (now) include a call.
1633 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1635 PR tree-optimization/92822
1636 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1637 out the don't-care elements of a vector whose significant elements
1638 are duplicates, make the don't-care elements duplicates too.
1640 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1642 PR tree-optimization/93434
1643 * tree-predcom.c (split_data_refs_to_components): Record which
1644 components have had aliasing loads removed. Prevent store-store
1645 commoning for all such components.
1647 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1650 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1651 -1 or is_vshift is true, use new_vector with number of elts npatterns
1652 rather than new_unary_operation.
1654 PR tree-optimization/93454
1655 * gimple-fold.c (fold_array_ctor_reference): Perform
1656 elt_size.to_uhwi () just once, instead of calling it in every
1657 iteration. Punt if that value is above size of the temporary
1658 buffer. Decrease third native_encode_expr argument when
1659 bufoff + elt_sz is above size of buf.
1661 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1663 * config/mips/mips.c (mips_declare_object_name)
1664 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1666 2020-01-27 Martin Liska <mliska@suse.cz>
1668 PR gcov-profile/93403
1669 * tree-profile.c (gimple_init_gcov_profiler): Generate
1670 both __gcov_indirect_call_profiler_v4 and
1671 __gcov_indirect_call_profiler_v4_atomic.
1673 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1676 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1678 (@aarch64_split_simd_mov<mode>): Use it.
1679 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1680 Leave the vec_extract patterns to handle 2-element vectors.
1681 (aarch64_simd_mov_from_<mode>high): Likewise.
1682 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1683 (vec_extractv2dfv1df): Likewise.
1685 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1687 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1688 jump conditions for *compare_condjump<GPI:mode>.
1690 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1693 * digraph.cc (test_edge::test_edge): Specify template for base
1696 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1698 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1700 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1702 * config/arc/arc-protos.h (gen_mlo): Remove.
1703 (gen_mhi): Likewise.
1704 * config/arc/arc.c (AUX_MULHI): Define.
1705 (arc_must_save_reister): Special handling for r58/59.
1706 (arc_compute_frame_size): Consider mlo/mhi registers.
1707 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1709 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1710 mlo/mhi name selection.
1711 (arc_restore_callee_saves): Don't early restore blink when ISR.
1712 (arc_expand_prologue): Add mlo/mhi saving.
1713 (arc_expand_epilogue): Add mlo/mhi restoring.
1716 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1717 numbering when MUL64 option is used.
1718 (DWARF2_FRAME_REG_OUT): Define.
1719 * config/arc/arc.md (arc600_stall): New pattern.
1720 (VUNSPEC_ARC_ARC600_STALL): Define.
1721 (mulsi64): Use correct mlo/mhi registers.
1722 (mulsi_600): Clean it up.
1723 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1725 (mhi_operand): Likewise.
1727 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1728 Petro Karashchenko <petro.karashchenko@ring.com>
1730 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1731 attributes if needed.
1732 (prepare_move_operands): Generate special unspec instruction for
1734 (arc_isuncached_mem_p): Propagate uncached attribute to each
1736 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1737 (VUNSPEC_ARC_STDI): Likewise.
1738 (ALLI): New mode iterator.
1739 (mALLI): New mode attribute.
1740 (lddi): New instruction pattern.
1742 (stdidi_split): Split instruction for architectures which are not
1743 supporting ll64 option.
1744 (lddidi_split): Likewise.
1746 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1748 PR rtl-optimization/92989
1749 * lra-lives.c (process_bb_lives): Update the live-in set before
1750 processing additional clobbers.
1752 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1754 PR rtl-optimization/93170
1755 * cselib.c (cselib_invalidate_regno_val): New function, split out
1757 (cselib_invalidate_regno): ...here.
1758 (cselib_invalidated_by_call_p): New function.
1759 (cselib_process_insn): Iterate over all the hard-register entries in
1760 REG_VALUES and invalidate any that cross call-clobbered registers.
1762 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1764 * dojump.c (split_comparison): Use HONOR_NANS rather than
1765 HONOR_SNANS when splitting LTGT.
1767 2020-01-27 Martin Liska <mliska@suse.cz>
1770 * opts.c (print_filtered_help): Exclude language-specific
1771 options from --help=common unless enabled in all FEs.
1773 2020-01-27 Martin Liska <mliska@suse.cz>
1775 * opts.c (print_help): Exclude params from
1776 all except --help=param.
1778 2020-01-27 Martin Liska <mliska@suse.cz>
1781 * config/i386/i386-features.c (make_resolver_func):
1782 Align the code with ppc64 target implementation.
1783 Do not generate a unique name for resolver function.
1785 2020-01-27 Richard Biener <rguenther@suse.de>
1787 PR tree-optimization/93397
1788 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1789 converted reduction chain SLP graph adjustment.
1791 2020-01-26 Marek Polacek <polacek@redhat.com>
1794 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1797 2020-01-26 Jason Merrill <jason@redhat.com>
1800 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1803 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1805 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1806 (rx_setmem): Likewise.
1808 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1811 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1812 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1813 drop <di> from constraint of last operand.
1816 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1817 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1818 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1820 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1823 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1826 2020-01-24 Jeff Law <law@redhat.com>
1828 PR tree-optimization/92788
1829 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1832 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1835 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1836 *avx_vperm_broadcast_<mode>,
1837 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1838 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1839 Move before avx2_perm<mode>/avx512f_perm<mode>.
1842 * simplify-rtx.c (simplify_const_unary_operation,
1843 simplify_const_binary_operation): Punt for mode precision above
1844 MAX_BITSIZE_MODE_ANY_INT.
1846 2020-01-24 Andrew Pinski <apinski@marvell.com>
1848 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1851 2020-01-24 Jeff Law <law@redhat.com>
1854 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1855 for REGs. Call output_operand_lossage to get more reasonable
1858 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1860 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1861 gcn_fp_compare_operator.
1862 (vec_cmpu<mode>di): Use gcn_compare_operator.
1863 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1864 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1865 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1866 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1867 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1868 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1869 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1870 gcn_fp_compare_operator.
1871 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1872 gcn_fp_compare_operator.
1873 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1874 gcn_fp_compare_operator.
1875 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1876 gcn_fp_compare_operator.
1878 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1880 * doc/install.texi (Cross-Compiler-Specific Options): Document
1881 `--with-toolexeclibdir' option.
1883 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1885 * target.def (flags_regnum): Also mention effect on delay slot filling.
1886 * doc/tm.texi: Regenerate.
1888 2020-01-23 Jeff Law <law@redhat.com>
1890 PR translation/90162
1891 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1893 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1896 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1899 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1901 PR rtl-optimization/93402
1902 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1905 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1907 * config.in: Regenerated.
1908 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1909 for TARGET_LIBC_GNUSTACK.
1910 * configure: Regenerated.
1911 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1912 found to be 2.31 or greater.
1914 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1916 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1918 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1919 (mips_asm_file_end): New function. Delegate to
1920 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1921 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1923 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1926 * config/i386/i386-modes.def (POImode): New mode.
1927 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1928 * config/i386/i386.md (DPWI): New mode attribute.
1929 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1931 (QPWI): ... this. Use POI instead of OI for TImode.
1932 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1933 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1936 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1939 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1941 (speculation_tracker_rev): New pattern.
1942 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1943 Use speculation_tracker_rev to track the inverse condition.
1945 2020-01-23 Richard Biener <rguenther@suse.de>
1947 PR tree-optimization/93381
1948 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1949 alias-set of the def as argument and record the first one.
1950 (vn_walk_cb_data::first_set): New member.
1951 (vn_reference_lookup_3): Pass the alias-set of the current def
1952 to push_partial_def. Fix alias-set used in the aggregate copy
1954 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1955 * real.c (clear_significand_below): Fix out-of-bound access.
1957 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1960 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1961 New define_insn patterns.
1963 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1965 * doc/sourcebuild.texi (check-function-bodies): Add an
1966 optional target/xfail selector.
1968 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1970 PR rtl-optimization/93124
1971 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1972 bare USE and CLOBBER insns.
1974 2020-01-22 Andrew Pinski <apinski@marvell.com>
1976 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1978 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1981 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1982 diagnostic_manager into "ana" namespace.
1983 * selftest-run-tests.c (selftest::run_tests): Update for move of
1984 selftest::run_analyzer_selftests to
1985 ana::selftest::run_analyzer_selftests.
1987 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1989 * cfgexpand.c (union_stack_vars): Update the size.
1991 2020-01-22 Richard Biener <rguenther@suse.de>
1993 PR tree-optimization/93381
1994 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1995 throughout, handle all conversions the same.
1997 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2000 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
2001 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
2002 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
2003 Call force_reg on high_in2 unconditionally.
2005 2020-01-22 Martin Liska <mliska@suse.cz>
2007 PR tree-optimization/92924
2008 * profile.c (compute_value_histograms): Divide
2011 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2014 * output.h (assemble_name_resolve): Declare.
2015 * varasm.c (assemble_name_resolve): New function.
2016 (assemble_name): Use it.
2017 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2019 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2021 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2022 update_web_docs_git instead of update_web_docs_svn.
2024 2020-01-21 Andrew Pinski <apinski@marvell.com>
2027 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2028 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2029 (*tlsgd_small_<mode>): Likewise.
2030 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2031 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2032 register. Convert that register back to dest using convert_mode.
2034 2020-01-21 Jim Wilson <jimw@sifive.com>
2036 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2039 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2040 Uros Bizjak <ubizjak@gmail.com>
2043 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2045 (legitimize_tls_address): Do GNU2 TLS address computation in
2046 ptr_mode and zero-extend result to Pmode.
2047 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2048 :P with :PTR and Pmode with ptr_mode.
2049 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2050 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2051 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2053 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2056 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2057 the last two operands are CONST_INT_P before using them as such.
2059 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2061 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2062 to get the integer element types.
2064 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2066 * config/aarch64/aarch64-sve-builtins.h
2067 (function_expander::convert_to_pmode): Declare.
2068 * config/aarch64/aarch64-sve-builtins.cc
2069 (function_expander::convert_to_pmode): New function.
2070 (function_expander::get_contiguous_base): Use it.
2071 (function_expander::prepare_gather_address_operands): Likewise.
2072 * config/aarch64/aarch64-sve-builtins-sve2.cc
2073 (svwhilerw_svwhilewr_impl::expand): Likewise.
2075 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2078 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2079 cfun->machine->label_is_assembled.
2080 (aarch64_print_patchable_function_entry): New.
2081 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2082 * config/aarch64/aarch64.h (struct machine_function): New field,
2085 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2088 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2091 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2094 * cgraph.c (cgraph_edge::resolve_speculation,
2095 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2096 call_stmt_site_hash.
2098 2020-01-21 Martin Liska <mliska@suse.cz>
2100 * config/rs6000/rs6000.c (common_mode_defined): Remove
2103 2020-01-21 Richard Biener <rguenther@suse.de>
2105 PR tree-optimization/92328
2106 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2107 type when value-numbering same-sized store by inserting a
2109 (eliminate_dom_walker::eliminate_stmt): When eliminating
2110 a redundant store handle bit-reinterpretation of the same value.
2112 2020-01-21 Andrew Pinski <apinski@marvel.com>
2115 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2117 (prepare_block_for_update): This. Use a worklist instead of
2120 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2122 * gcc/config/arm/arm.c (clear_operation_p):
2123 Initialise last_regno, skip first iteration
2124 based on the first_set value and use ints instead
2125 of the unnecessary HOST_WIDE_INTs.
2127 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2130 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2131 compare_mode other than SFmode or DFmode.
2133 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2136 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2137 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2138 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2140 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2142 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2144 2020-01-20 Andrew Pinski <apinski@marvell.com>
2147 * targhooks.c (default_print_patchable_function_entry): Use
2148 output_asm_insn to emit the nop instruction.
2150 2020-01-20 Fangrui Song <maskray@google.com>
2153 * targhooks.c (default_print_patchable_function_entry): Align to
2156 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2159 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2160 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2161 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2162 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2163 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2164 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2165 Remove the {q} suffix from lea.
2166 (*tls_dynamic_gnu2_call_64): Renamed to ...
2167 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2168 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2169 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2170 Pass Pmode to gen_tls_dynamic_gnu2_64.
2172 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2174 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2176 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2178 * config/aarch64/aarch64-sve-builtins-base.cc
2179 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2181 2020-01-20 Richard Biener <rguenther@suse.de>
2184 * dwarf2out.c (prune_unused_types): Unconditionally mark
2185 called function DIEs.
2187 2020-01-20 Martin Liska <mliska@suse.cz>
2189 PR tree-optimization/93199
2190 * tree-eh.c (struct leh_state): Add
2191 new field outer_non_cleanup.
2192 (cleanup_is_dead_in): Pass leh_state instead
2193 of eh_region. Add a checking that state->outer_non_cleanup
2194 points to outer non-clean up region.
2195 (lower_try_finally): Record outer_non_cleanup
2197 (lower_catch): Likewise.
2198 (lower_eh_filter): Likewise.
2199 (lower_eh_must_not_throw): Likewise.
2200 (lower_cleanup): Likewise.
2202 2020-01-20 Richard Biener <rguenther@suse.de>
2204 PR tree-optimization/93094
2205 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2206 (vect_transform_loop): Likewise.
2207 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2208 loop_vectorized_call to vect_transform_loop.
2209 * tree-vect-loop.c (vect_transform_loop): Pass down
2210 loop_vectorized_call to vect_loop_versioning.
2211 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2212 the earlier discovered loop_vectorized_call.
2214 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2216 * doc/contribute.texi: Update for SVN -> Git transition.
2217 * doc/install.texi: Likewise.
2219 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2222 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2223 speculative targets.
2224 (verify_speculative_call): New function
2225 (cgraph_node::verify_node): Use it.
2226 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2229 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2232 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2233 (cgraph_edge::make_direct): Remove all indirect targets.
2234 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2235 (cgraph_node::verify_node): Verify that only one call_stmt or
2236 lto_stmt_uid is set.
2237 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2239 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2240 (lto_output_ref): Simplify streaming of stmt.
2241 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2243 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2245 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2246 Mark parameter unused.
2248 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2250 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2252 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2254 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2256 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2258 * Makefile.in: Add coroutine-passes.o.
2259 * builtin-types.def (BT_CONST_SIZE): New.
2260 (BT_FN_BOOL_PTR): New.
2261 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2262 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2263 * coroutine-builtins.def: New file.
2264 * coroutine-passes.cc: New file.
2265 * function.h (struct GTY function): Add a bit to indicate that the
2266 function is a coroutine component.
2267 * internal-fn.c (expand_CO_FRAME): New.
2268 (expand_CO_YIELD): New.
2269 (expand_CO_SUSPN): New.
2270 (expand_CO_ACTOR): New.
2271 * internal-fn.def (CO_ACTOR): New.
2275 * passes.def: Add pass_coroutine_lower_builtins,
2276 pass_coroutine_early_expand_ifns.
2277 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2278 (make_pass_coroutine_early_expand_ifns): New.
2279 * doc/invoke.texi: Document the fcoroutines command line
2282 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2284 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2287 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2288 after checking the argument is a REG. Don't use REGNO (reg)
2289 again to set last_regno, reuse regno variable instead.
2291 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2293 * doc/analyzer.texi (Limitations): Add note about NaN.
2295 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2296 Sudakshina Das <sudi.das@arm.com>
2298 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2299 and valid immediate.
2300 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2301 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2302 * config/arm/constraints.md (Pg): New.
2303 * config/arm/predicates.md (long_shift_imm): New.
2304 (arm_reg_or_long_shift_imm): Likewise.
2305 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2306 (thumb2_lsll): Likewise.
2309 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2310 Sudakshina Das <sudi.das@arm.com>
2312 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2313 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2314 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2315 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2316 * config/arm/thumb2.md (thumb2_asrl): New.
2317 (thumb2_lsll): Likewise.
2319 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2321 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2324 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2326 * gdbinit.in (help-gcc-hooks): New command.
2327 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2328 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2331 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2333 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2334 correct target macro.
2336 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2338 * config/aarch64/aarch64-protos.h
2339 (aarch64_sve_ld1ro_operand_p): New.
2340 * config/aarch64/aarch64-sve-builtins-base.cc
2341 (class load_replicate): New.
2342 (class svld1ro_impl): New.
2343 (class svld1rq_impl): Change to inherit from load_replicate.
2344 (svld1ro): New sve intrinsic function base.
2345 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2346 New DEF_SVE_FUNCTION.
2347 * config/aarch64/aarch64-sve-builtins-base.h
2348 (svld1ro): New decl.
2349 * config/aarch64/aarch64-sve-builtins.cc
2350 (function_expander::add_mem_operand): Modify assert to allow
2352 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2354 * config/aarch64/aarch64.c
2355 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2356 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2357 (aarch64_sve_ld1ro_operand_p): New.
2358 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2359 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2360 * config/aarch64/predicates.md
2361 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2363 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2365 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2366 Introduce this ACLE specified predefined macro.
2367 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2368 (fp): Disabling this disables f64mm.
2369 (simd): Disabling this disables f64mm.
2370 (fp16): Disabling this disables f64mm.
2371 (sve): Disabling this disables f64mm.
2372 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2373 (AARCH64_ISA_F64MM): New.
2374 (TARGET_F64MM): New.
2375 * doc/invoke.texi (f64mm): Document new option.
2377 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2379 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2380 (neoversen1_tunings): Likewise.
2382 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2385 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2386 Add assert to ensure prolog has been emitted.
2387 (aarch64_split_atomic_op): Likewise.
2388 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2389 Use epilogue_completed rather than reload_completed.
2390 (aarch64_atomic_exchange<mode>): Likewise.
2391 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2392 (atomic_nand<mode>): Likewise.
2393 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2394 (atomic_fetch_nand<mode>): Likewise.
2395 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2396 (atomic_nand_fetch<mode>): Likewise.
2398 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2401 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2403 (REVERSE_CONDITION): Delete.
2404 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2405 (CCFP_CCFPE): Likewise.
2406 (e): New mode attribute.
2407 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2408 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2409 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2410 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2411 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2412 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2413 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2414 name of generator from gen_ccmpdi to gen_ccmpccdi.
2415 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2416 the previous comparison but aren't able to, use the new ccmp_rev
2419 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2421 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2422 than testing directly for INTEGER_CST.
2423 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2425 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2427 PR tree-optimization/93292
2428 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2429 get_vectype_for_scalar_type returns NULL.
2431 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2433 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2434 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2436 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2438 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2440 * params.opt: (max-predicted-iterations): Set bounds.
2441 * predict.c (real_almost_one, real_br_prob_base,
2442 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2443 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2444 probabilities; do not truncate to reg_br_prob_bases.
2445 (estimate_loops_at_level): Pass max_cyclic_prob.
2446 (estimate_loops): Compute max_cyclic_prob.
2447 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2449 * profile-count.c (profile_probability::to_sreal): New.
2450 * profile-count.h (class sreal): Move up in file.
2451 (profile_probability::to_sreal): Declare.
2453 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2456 (arm_invalid_conversion): New function for target hook.
2457 (arm_invalid_unary_op): New function for target hook.
2458 (arm_invalid_binary_op): New function for target hook.
2460 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2462 * config.gcc: Add arm_bf16.h.
2463 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2464 (arm_simd_builtin_std_type): Add BFmode.
2465 (arm_init_simd_builtin_types): Define element types for vector types.
2466 (arm_init_bf16_types): New function.
2467 (arm_init_builtins): Add arm_init_bf16_types function call.
2468 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2469 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2470 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2471 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2472 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2473 (arm_mangle_type): Add __bf16.
2474 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2475 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2476 arm_bf16_ptr_type_node.
2477 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2478 define_split between ARM registers.
2479 * config/arm/arm_bf16.h: New file.
2480 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2481 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2483 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2484 * config/arm/vfp.md: Add BFmode to movhf patterns.
2486 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2487 Andre Vieira <andre.simoesdiasvieira@arm.com>
2489 * config/arm/arm-cpus.in (mve, mve_float): New features.
2490 (dsp, mve, mve.fp): New options.
2491 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2492 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2493 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2495 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2496 Thomas Preud'homme <thomas.preudhomme@arm.com>
2498 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2500 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2501 error for using -mcmse when targeting Armv8.1-M Mainline.
2503 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2504 Thomas Preud'homme <thomas.preudhomme@arm.com>
2506 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2507 address in r4 when targeting Armv8.1-M Mainline.
2508 (nonsecure_call_value_internal): Likewise.
2509 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2510 a register match_operand again. Emit BLXNS when targeting
2512 (nonsecure_call_value_reg_thumb2): Likewise.
2514 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2515 Thomas Preud'homme <thomas.preudhomme@arm.com>
2517 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2518 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2519 variable as true when floating-point ABI is not hard. Replace
2520 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2521 Generate VLSTM and VLLDM instruction respectively before and
2522 after a function call to cmse_nonsecure_call function.
2523 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2524 (VUNSPEC_VLLDM): Likewise.
2525 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2526 (lazy_load_multiple_insn): Likewise.
2528 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2529 Thomas Preud'homme <thomas.preudhomme@arm.com>
2531 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2532 (arm_emit_vfp_multi_reg_pop): Likewise.
2533 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2534 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2535 restore callee-saved VFP registers.
2537 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2538 Thomas Preud'homme <thomas.preudhomme@arm.com>
2540 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2541 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2542 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2543 callee-saved GPRs as well as clear ip register before doing a nonsecure
2544 call then restore callee-saved GPRs after it when targeting
2546 (arm_reorg): Adapt to function rename.
2548 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2549 Thomas Preud'homme <thomas.preudhomme@arm.com>
2551 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2552 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2553 clear_vfp_multiple pattern based on a new vfp parameter.
2554 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2555 targeting Armv8.1-M Mainline.
2556 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2557 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2558 whether VFP registers are available before looking call_used_regs for a
2560 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2561 of prototype of clear_operation_p.
2562 (clear_vfp_multiple_operation): New predicate.
2563 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2564 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2566 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2567 Thomas Preud'homme <thomas.preudhomme@arm.com>
2569 * config/arm/arm-protos.h (clear_operation_p): Declare.
2570 * config/arm/arm.c (clear_operation_p): New function.
2571 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2572 targeting Armv8.1-M Mainline or successor.
2573 (output_return_instruction): Only output APSR register clearing if
2574 Armv8.1-M Mainline instructions not available.
2575 (thumb_exit): Likewise.
2576 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2577 * config/arm/thumb2.md (clear_apsr): New define_insn.
2578 (clear_multiple): Likewise.
2579 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2581 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2582 Thomas Preud'homme <thomas.preudhomme@arm.com>
2584 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2585 (use_return_insn): Also return false for Armv8.1-M Mainline.
2586 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2587 Mainline instructions are available.
2588 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2589 when targeting Armv8.1-M Mainline Security Extensions.
2590 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2591 Mainline entry function.
2592 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2593 targeting Armv8.1-M Mainline or successor.
2594 (arm_expand_epilogue): Fix indentation of caller-saved register
2595 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2597 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2598 (FP_SYSREGS): Likewise.
2599 (enum vfp_sysregs_encoding): Define enum.
2600 (fp_sysreg_names): Declare.
2601 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2602 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2603 (pop_fpsysreg_insn): Likewise.
2605 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2606 Thomas Preud'homme <thomas.preudhomme@arm.com>
2608 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2609 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2610 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2611 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2612 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2613 (ARMv8_1m_main): New feature group.
2614 (armv8.1-m.main): New architecture.
2615 * config/arm/arm-tables.opt: Regenerate.
2616 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2617 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2618 (arm_options_perform_arch_sanity_checks): Error out when targeting
2619 Armv8.1-M Mainline Security Extensions.
2620 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2622 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2624 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2625 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2626 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2627 aarch64_bfdot_laneq): New.
2628 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2629 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2630 vbfdotq_laneq_f32): New.
2631 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2632 VBFMLA_W, VBF): New.
2633 (isquadop): Add V4BF, V8BF.
2635 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2637 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2638 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2639 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2640 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2641 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2642 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2643 usdot_laneq, sudot_lane,sudot_laneq): New.
2644 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2645 (aarch64_<sur>dot_lane): New.
2646 * config/aarch64/arm_neon.h (vusdot_s32): New.
2648 (vusdot_lane_s32): New.
2649 (vsudot_lane_s32): New.
2650 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2651 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2653 2020-01-16 Martin Liska <mliska@suse.cz>
2655 * value-prof.c (dump_histogram_value): Fix
2656 obvious spacing issue.
2658 2020-01-16 Andrew Pinski <apinski@marvell.com>
2660 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2661 !storage_order_barrier_p.
2663 2020-01-16 Andrew Pinski <apinski@marvell.com>
2665 * sched-int.h (_dep): Add unused bit-field field for the padding.
2666 * sched-deps.c (init_dep_1): Init unused field.
2668 2020-01-16 Andrew Pinski <apinski@marvell.com>
2670 * optabs.h (create_expand_operand): Initialize target field also.
2672 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2674 PR tree-optimization/92429
2675 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2676 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2678 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2681 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2683 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2684 aarch64_sve_int_mode to each mode.
2686 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2688 * doc/analyzer.texi (Overview): Add note about
2689 -fdump-ipa-analyzer.
2691 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2693 PR tree-optimization/93231
2694 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2695 input_type is unsigned. Use tree_to_shwi for shift constant.
2696 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2697 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2698 inputs more efficiently.
2700 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2702 * config/i386/i386.md (*movsf_internal): Do not require
2703 SSE2 ISA for alternatives 14 and 15.
2705 2020-01-15 Richard Biener <rguenther@suse.de>
2708 * tree-eh.c (sink_clobbers): If we already visited the destination
2709 block do not defer insertion.
2710 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2711 the purpose of defered insertion.
2713 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2715 * BASE-VER: Bump to 10.0.1.
2717 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2719 PR tree-optimization/93247
2720 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2721 type of the stmt that we're going to vectorize.
2723 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2725 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2726 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2729 2020-01-15 Martin Liska <mliska@suse.cz>
2731 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2732 2 calls of streamer_read_hwi in a function call.
2734 2020-01-15 Richard Biener <rguenther@suse.de>
2736 * alias.c (record_alias_subset): Avoid redundant work when
2737 subset is already recorded.
2739 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2741 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2742 the analyzer options provide CWE identifiers.
2744 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2746 * tree-diagnostic-path.cc (path_summary::event_range::print):
2747 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2748 using get_pure_location.
2750 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2752 PR tree-optimization/93262
2753 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2754 perform head trimming only if the last argument is constant,
2755 either all ones, or larger or equal to head trim, in the latter
2756 case decrease the last argument by head_trim.
2758 PR tree-optimization/93249
2759 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2760 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2761 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2762 perform head trim unless we can prove there are no '\0' chars
2763 from the source among the first head_trim chars.
2765 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2767 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2769 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2772 * config/i386/sse.md
2773 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2774 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2775 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2776 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2777 just a single alternative instead of two, make operands 1 and 2
2780 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2783 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2786 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2788 * Makefile.in (lang_opt_files): Add analyzer.opt.
2789 (ANALYZER_OBJS): New.
2790 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2791 tristate.o and ANALYZER_OBJS.
2792 (TEXI_GCCINT_FILES): Add analyzer.texi.
2793 * common.opt (-fanalyzer): New driver option.
2794 * config.in: Regenerate.
2795 * configure: Regenerate.
2796 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2797 (gccdepdir): Also create depdir for "analyzer" subdir.
2798 * digraph.cc: New file.
2799 * digraph.h: New file.
2800 * doc/analyzer.texi: New file.
2801 * doc/gccint.texi ("Static Analyzer") New menu item.
2802 (analyzer.texi): Include it.
2803 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2804 ("Warning Options"): Add static analysis warnings to the list.
2805 (-Wno-analyzer-double-fclose): New option.
2806 (-Wno-analyzer-double-free): New option.
2807 (-Wno-analyzer-exposure-through-output-file): New option.
2808 (-Wno-analyzer-file-leak): New option.
2809 (-Wno-analyzer-free-of-non-heap): New option.
2810 (-Wno-analyzer-malloc-leak): New option.
2811 (-Wno-analyzer-possible-null-argument): New option.
2812 (-Wno-analyzer-possible-null-dereference): New option.
2813 (-Wno-analyzer-null-argument): New option.
2814 (-Wno-analyzer-null-dereference): New option.
2815 (-Wno-analyzer-stale-setjmp-buffer): New option.
2816 (-Wno-analyzer-tainted-array-index): New option.
2817 (-Wno-analyzer-use-after-free): New option.
2818 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2819 (-Wno-analyzer-use-of-uninitialized-value): New option.
2820 (-Wanalyzer-too-complex): New option.
2821 (-fanalyzer-call-summaries): New warning.
2822 (-fanalyzer-checker=): New warning.
2823 (-fanalyzer-fine-grained): New warning.
2824 (-fno-analyzer-state-merge): New warning.
2825 (-fno-analyzer-state-purge): New warning.
2826 (-fanalyzer-transitivity): New warning.
2827 (-fanalyzer-verbose-edges): New warning.
2828 (-fanalyzer-verbose-state-changes): New warning.
2829 (-fanalyzer-verbosity=): New warning.
2830 (-fdump-analyzer): New warning.
2831 (-fdump-analyzer-callgraph): New warning.
2832 (-fdump-analyzer-exploded-graph): New warning.
2833 (-fdump-analyzer-exploded-nodes): New warning.
2834 (-fdump-analyzer-exploded-nodes-2): New warning.
2835 (-fdump-analyzer-exploded-nodes-3): New warning.
2836 (-fdump-analyzer-supergraph): New warning.
2837 * doc/sourcebuild.texi (dg-require-dot): New.
2838 (dg-check-dot): New.
2839 * gdbinit.in (break-on-saved-diagnostic): New command.
2840 * graphviz.cc: New file.
2841 * graphviz.h: New file.
2842 * ordered-hash-map-tests.cc: New file.
2843 * ordered-hash-map.h: New file.
2844 * passes.def (pass_analyzer): Add before
2845 pass_ipa_whole_program_visibility.
2846 * selftest-run-tests.c (selftest::run_tests): Call
2847 selftest::ordered_hash_map_tests_cc_tests.
2848 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2850 * shortest-paths.h: New file.
2851 * timevar.def (TV_ANALYZER): New timevar.
2852 (TV_ANALYZER_SUPERGRAPH): Likewise.
2853 (TV_ANALYZER_STATE_PURGE): Likewise.
2854 (TV_ANALYZER_PLAN): Likewise.
2855 (TV_ANALYZER_SCC): Likewise.
2856 (TV_ANALYZER_WORKLIST): Likewise.
2857 (TV_ANALYZER_DUMP): Likewise.
2858 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2859 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2860 * tree-pass.h (make_pass_analyzer): New decl.
2861 * tristate.cc: New file.
2862 * tristate.h: New file.
2864 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2867 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2868 alternatives 9 and 10.
2870 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2872 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2873 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2874 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2875 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2876 (selftest::hash_map_tests_c_tests): Call it.
2877 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2878 New static constant, using the value of = H::empty_zero_p.
2879 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2880 from default_hash_traits <Value>.
2881 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2883 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2884 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2885 calls to mark_empty with !Descriptor::empty_zero_p.
2886 (hash_table::empty_slow): Conditionalize the memset call with a
2887 check that Descriptor::empty_zero_p; otherwise, loop through the
2888 entries calling mark_empty on them.
2889 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2890 (pointer_hash::empty_zero_p): Likewise.
2891 (pair_hash::empty_zero_p): Likewise.
2892 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2894 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2895 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2896 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2897 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2898 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2899 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2900 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2901 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2903 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2906 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2908 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2909 fix typo on return value.
2911 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2914 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2916 (cgraph_edge::make_speculative): Add param for setting speculative_id
2918 (cgraph_edge::speculative_call_info): Update comments and find reference
2919 by speculative_id for multiple indirect targets.
2920 (cgraph_edge::resolve_speculation): Decrease the speculations
2921 for indirect edge, drop it's speculative if not direct target
2922 left. Update comments.
2923 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2924 (cgraph_node::dump): Print num_speculative_call_targets.
2925 (cgraph_node::verify_node): Don't report error if speculative
2926 edge not include statement.
2927 (cgraph_edge::num_speculative_call_targets_p): New function.
2928 * cgraph.h (int common_target_id): Remove.
2929 (int common_target_probability): Remove.
2930 (num_speculative_call_targets): New variable.
2931 (make_speculative): Add param for setting speculative_id.
2932 (cgraph_edge::num_speculative_call_targets_p): New declare.
2933 (target_prob): New variable.
2934 (speculative_id): New variable.
2935 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2936 call summaries for multiple speculative call targets.
2937 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2938 * ipa-profile.c (struct speculative_call_target): New struct.
2939 (class speculative_call_summary): New class.
2940 (class speculative_call_summaries): New class.
2941 (call_sums): New variable.
2942 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2943 (ipa_profile_write_edge_summary): New function.
2944 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2945 (ipa_profile_dump_all_summaries): New function.
2946 (ipa_profile_read_edge_summary): New function.
2947 (ipa_profile_read_summary_section): New function.
2948 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2949 (ipa_profile): Generate num_speculative_call_targets from
2951 * ipa-ref.h (speculative_id): New variable.
2952 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2953 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2954 common_target_probability. Stream out speculative_id and
2955 num_speculative_call_targets.
2956 (input_edge): Likewise.
2957 * predict.c (dump_prediction): Remove edges count assert to be
2959 * symtab.c (symtab_node::create_reference): Init speculative_id.
2960 (symtab_node::clone_references): Clone speculative_id.
2961 (symtab_node::clone_referring): Clone speculative_id.
2962 (symtab_node::clone_reference): Clone speculative_id.
2963 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2964 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2965 if indirect call contains multiple speculative targets.
2966 * value-prof.h (check_ic_target): Remove.
2967 * value-prof.c (gimple_value_profile_transformations):
2968 Use void function gimple_ic_transform.
2969 * value-prof.c (gimple_ic_transform): Handle topn case.
2970 Fix comment typos. Change it to a void function.
2972 2020-01-13 Andrew Pinski <apinski@marvell.com>
2974 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2975 (octeontx2t98): New define.
2976 (octeontx2t96): New define.
2977 (octeontx2t93): New define.
2978 (octeontx2f95): New define.
2979 (octeontx2f95n): New define.
2980 (octeontx2f95mm): New define.
2981 * config/aarch64/aarch64-tune.md: Regenerate.
2982 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2984 2020-01-13 Jason Merrill <jason@redhat.com>
2986 PR c++/33799 - destroy return value if local cleanup throws.
2987 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2989 2020-01-13 Martin Liska <mliska@suse.cz>
2991 * ipa-cp.c (get_max_overall_size): Use newly
2992 renamed param param_ipa_cp_unit_growth.
2993 * params.opt: Remove legacy param name.
2995 2020-01-13 Martin Sebor <msebor@redhat.com>
2997 PR tree-optimization/93213
2998 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2999 stores to be eliminated.
3001 2020-01-13 Martin Liska <mliska@suse.cz>
3003 * opts.c (print_help): Do not print CL_PARAM
3004 and CL_WARNING for CL_OPTIMIZATION.
3006 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
3009 * doc/invoke.texi (Warning Options): Add caveat about some warnings
3010 depending on optimization settings.
3012 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3014 PR tree-optimization/90838
3015 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3016 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3017 argument rather than to initialize temporary for targets that
3018 don't use the mode argument at all. Initialize ctzval to avoid
3021 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3023 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3024 * tree-core.h: Document it.
3025 * gimplify.c (gimplify_omp_workshare): Set it.
3026 * omp-low.c (lower_omp_target): Use it.
3027 * tree-pretty-print.c (dump_omp_clause): Print it.
3029 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3030 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3032 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3034 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3035 * common.opt (fdiagnostics-path-format=): New option.
3036 (diagnostic_path_format): New enum.
3037 (fdiagnostics-show-path-depths): New option.
3038 * coretypes.h (diagnostic_event_id_t): New forward decl.
3039 * diagnostic-color.c (color_dict): Add "path".
3040 * diagnostic-event-id.h: New file.
3041 * diagnostic-format-json.cc (json_from_expanded_location): Make
3043 (json_end_diagnostic): Call context->make_json_for_path if it
3044 exists and the diagnostic has a path.
3045 (diagnostic_output_format_init): Clear context->print_path.
3046 * diagnostic-path.h: New file.
3047 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3048 when printing a run of events in a diagnostic_path so that they
3049 all get the same color.
3050 (layout::m_diagnostic_path_p): New field.
3051 (layout::layout): Initialize it.
3052 (layout::print_any_labels): Don't colorize the label text for an
3053 event in a diagnostic_path.
3054 (gcc_rich_location::add_location_if_nearby): Add
3055 "restrict_to_current_line_spans" and "label" params. Pass the
3056 former to layout.maybe_add_location_range; pass the latter
3057 when calling add_range.
3058 * diagnostic.c: Include "diagnostic-path.h".
3059 (diagnostic_initialize): Initialize context->path_format and
3060 context->show_path_depths.
3061 (diagnostic_show_any_path): New function.
3062 (diagnostic_path::interprocedural_p): New function.
3063 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3064 (simple_diagnostic_path::num_events): New function.
3065 (simple_diagnostic_path::get_event): New function.
3066 (simple_diagnostic_path::add_event): New function.
3067 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3068 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3069 (debug): New overload taking a diagnostic_path *.
3070 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3071 * diagnostic.h (enum diagnostic_path_format): New enum.
3072 (json::value): New forward decl.
3073 (diagnostic_context::path_format): New field.
3074 (diagnostic_context::show_path_depths): New field.
3075 (diagnostic_context::print_path): New callback field.
3076 (diagnostic_context::make_json_for_path): New callback field.
3077 (diagnostic_show_any_path): New decl.
3078 (json_from_expanded_location): New decl.
3079 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3080 (-fdiagnostics-show-path-depths): New option.
3081 (-fdiagnostics-color): Add "path" to description of default
3082 GCC_COLORS; describe it.
3083 (-fdiagnostics-format=json): Document how diagnostic paths are
3084 represented in the JSON output format.
3085 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3086 Add optional params "restrict_to_current_line_spans" and "label".
3087 * opts.c (common_handle_option): Handle
3088 OPT_fdiagnostics_path_format_ and
3089 OPT_fdiagnostics_show_path_depths.
3090 * pretty-print.c: Include "diagnostic-event-id.h".
3091 (pp_format): Implement "%@" format code for printing
3092 diagnostic_event_id_t *.
3093 (selftest::test_pp_format): Add tests for "%@".
3094 * selftest-run-tests.c (selftest::run_tests): Call
3095 selftest::tree_diagnostic_path_cc_tests.
3096 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3097 * toplev.c (general_init): Initialize global_dc->path_format and
3098 global_dc->show_path_depths.
3099 * tree-diagnostic-path.cc: New file.
3100 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3101 non-static. Drop "diagnostic" param in favor of storing the
3102 original value of "where" and re-using it.
3103 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3104 maybe_unwind_expanded_macro_loc.
3105 (tree_diagnostics_defaults): Initialize context->print_path and
3106 context->make_json_for_path.
3107 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3109 (default_tree_make_json_for_path): New decl.
3110 (maybe_unwind_expanded_macro_loc): New decl.
3112 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3114 PR tree-optimization/93210
3115 * fold-const.h (native_encode_initializer,
3116 can_native_interpret_type_p): Declare.
3117 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3119 (native_encode_initializer): New function, moved from dwarf2out.c.
3120 Adjust to native_encode_expr compatible arguments, including dry-run
3121 and partial extraction modes. Don't handle STRING_CST.
3122 (can_native_interpret_type_p): No longer static.
3123 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3124 offset / BITS_PER_UNIT fits into int and don't call it if
3125 can_native_interpret_type_p fails. If suboff is NULL and for
3126 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3127 native_encode_initializer.
3128 (fold_const_aggregate_ref_1): Formatting fix.
3129 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3130 (tree_add_const_value_attribute): Adjust caller.
3132 PR tree-optimization/90838
3133 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3134 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3135 CTZ_DEFINED_VALUE_AT_ZERO.
3137 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3140 * lra-constraints.c (match_reload): Permit input operands have the
3141 same mode as output while other input operands have a different
3144 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3146 PR tree-optimization/90838
3147 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3148 (check_ctz_string): Likewise.
3149 (optimize_count_trailing_zeroes): Likewise.
3150 (simplify_count_trailing_zeroes): Likewise.
3151 (pass_forwprop::execute): Try ctz simplification.
3152 * match.pd: Add matching for ctz idioms.
3154 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3156 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3158 (aarch64_invalid_unary_op): New function for target hook.
3159 (aarch64_invalid_binary_op): New function for target hook.
3161 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3163 * config.gcc: Add arm_bf16.h.
3164 * config/aarch64/aarch64-builtins.c
3165 (aarch64_simd_builtin_std_type): Add BFmode.
3166 (aarch64_init_simd_builtin_types): Define element types for vector
3168 (aarch64_init_bf16_types): New function.
3169 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3170 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3172 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3173 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3175 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3176 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3177 * config/aarch64/aarch64.c
3178 (aarch64_classify_vector_mode): Add support for BF types.
3179 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3180 (aarch64_vq_mode): Add support for BF types.
3181 (aarch64_simd_container_mode): Add support for BF types.
3182 (aarch64_mangle_type): Add support for BF scalar type.
3183 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3184 * config/aarch64/arm_bf16.h: New file.
3185 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3186 * config/aarch64/iterators.md: Add BF types to mode attributes.
3187 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3189 2020-01-10 Jason Merrill <jason@redhat.com>
3191 PR c++/93173 - incorrect tree sharing.
3192 * gimplify.c (copy_if_shared): No longer static.
3193 * gimplify.h: Declare it.
3195 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3197 * doc/invoke.texi (-msve-vector-bits=): Document that
3198 -msve-vector-bits=128 now generates VL-specific code for
3199 little-endian targets.
3200 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3201 build_vector_type_for_mode to construct the data vector types.
3202 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3203 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3204 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3205 for 128-bit vectors.
3207 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3209 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3212 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3214 * config/aarch64/aarch64-builtins.c
3215 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3216 rather than checking the number of elements and the element mode.
3218 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3220 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3221 get_related_vectype_for_scalar_type rather than build_vector_type
3222 to create the index type for a conditional reduction.
3224 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3226 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3227 for any type of gather or scatter, including strided accesses.
3229 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3231 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3234 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3236 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3238 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3239 parameter and its use to reset DR_OFFSET's.
3240 (vect_transform_loop): Remove orig_drs_init argument.
3241 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3242 member of dr_vec_info rather than the offset of the associated
3243 data_reference's innermost_loop_behavior.
3244 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3245 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3246 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3247 get_dr_vinfo_offset.
3248 (vectorizable_store): Likewise.
3249 (vectorizable_load): Likewise.
3251 2020-01-10 Richard Biener <rguenther@suse.de>
3253 * gimple-ssa-store-merging
3254 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3256 2020-01-10 Martin Liska <mliska@suse.cz>
3259 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3260 encapsulation that was there before r280040.
3262 2020-01-10 Richard Biener <rguenther@suse.de>
3265 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3266 sequences to avoid walking them again for secondary opportunities.
3267 (pass_lower_eh_dispatch::execute): Instead actually insert
3270 2020-01-10 Richard Biener <rguenther@suse.de>
3273 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3274 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3275 avoid quadraticness.
3277 2020-01-10 Martin Jambor <mjambor@suse.cz>
3279 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3280 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3281 to get param_ipa_sra_max_replacements.
3282 (param_splitting_across_edge): Pass the caller to
3283 pull_accesses_from_callee.
3285 2020-01-10 Martin Jambor <mjambor@suse.cz>
3287 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3288 * ipa-cp.c (max_new_size): Removed.
3289 (orig_overall_size): New variable.
3290 (get_max_overall_size): New function.
3291 (estimate_local_effects): Use it. Adjust dump.
3292 (decide_about_value): Likewise.
3293 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3294 orig_overall_size. Adjust dump.
3295 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3297 2020-01-10 Martin Jambor <mjambor@suse.cz>
3299 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3300 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3301 instead of param_ipa_max_agg_items.
3302 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3303 optimization info for the callee.
3305 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3307 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3308 markers if debug_inline_points is false.
3310 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3312 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3314 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3315 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3316 aarch64-sve-builtins-sve2.h.
3317 (aarch64-sve-builtins-sve2.o): New rule.
3318 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3319 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3320 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3321 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3322 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3323 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3325 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3326 instructions that are handled here.
3327 (@cond_asrd<mode>): Generalize to...
3328 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3329 (*cond_asrd<mode>_2): Generalize to...
3330 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3331 (*cond_asrd<mode>_z): Generalize to...
3332 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3333 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3334 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3335 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3336 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3338 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3339 (@aarch64_scatter_stnt<mode>): Likewise.
3340 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3341 (@aarch64_mul_lane_<mode>): Likewise.
3342 (@aarch64_sve_suqadd<mode>_const): Likewise.
3343 (*<sur>h<addsub><mode>): Generalize to...
3344 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3346 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3347 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3348 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3349 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3350 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3351 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3352 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3353 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3354 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3355 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3356 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3357 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3358 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3359 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3360 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3361 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3362 (@aarch64_sve2_xar<mode>): Likewise.
3363 (@aarch64_sve2_bcax<mode>): Likewise.
3364 (*aarch64_sve2_eor3<mode>): Rename to...
3365 (@aarch64_sve2_eor3<mode>): ...this.
3366 (@aarch64_sve2_bsl<mode>): New expander.
3367 (@aarch64_sve2_nbsl<mode>): Likewise.
3368 (@aarch64_sve2_bsl1n<mode>): Likewise.
3369 (@aarch64_sve2_bsl2n<mode>): Likewise.
3370 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3371 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3372 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3373 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3374 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3375 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3376 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3377 (<su>mull<bt><Vwide>): Generalize to...
3378 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3380 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3381 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3382 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3383 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3384 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3385 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3386 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3387 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3388 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3389 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3390 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3391 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3392 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3393 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3394 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3395 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3396 (<SHRNB:r>shrnb<mode>): Generalize to...
3397 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3399 (<SHRNT:r>shrnt<mode>): Generalize to...
3400 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3402 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3403 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3404 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3405 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3406 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3407 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3408 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3409 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3410 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3411 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3412 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3413 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3414 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3415 (@aarch64_sve2_cvtnt<mode>): Likewise.
3416 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3417 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3418 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3419 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3420 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3421 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3422 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3423 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3424 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3425 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3426 (@aarch64_sve2_pmul<mode>): Likewise.
3427 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3428 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3429 (@aarch64_sve2_tbl2<mode>): Likewise.
3430 (@aarch64_sve2_tbx<mode>): Likewise.
3431 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3432 (@aarch64_sve2_histcnt<mode>): Likewise.
3433 (@aarch64_sve2_histseg<mode>): Likewise.
3434 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3435 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3436 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3437 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3438 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3439 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3440 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3441 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3442 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3443 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3444 (SVE2_PMULL_PAIR_I): New mode iterators.
3445 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3446 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3447 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3448 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3449 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3450 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3451 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3452 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3453 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3454 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3455 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3456 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3457 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3458 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3459 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3460 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3461 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3462 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3463 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3464 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3465 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3466 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3467 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3468 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3469 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3470 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3471 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3472 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3473 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3474 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3475 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3477 (VNARROW, Ventype): New mode attributes.
3478 (Vewtype): Handle VNx2DI. Fix typo in comment.
3479 (VDOUBLE): New mode attribute.
3480 (sve_lane_con): Handle VNx8HI.
3481 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3482 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3483 (sve_int_op, sve_int_op_rev): Handle the above codes.
3484 (sve_pred_int_rhs2_operand): Likewise.
3485 (MULLBT, SHRNB, SHRNT): Delete.
3486 (SVE_INT_SHIFT_IMM): New int iterator.
3487 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3488 and UNSPEC_WHILEHS for TARGET_SVE2.
3489 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3490 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3491 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3492 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3493 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3494 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3495 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3496 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3497 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3498 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3499 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3500 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3501 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3502 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3503 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3504 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3505 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3506 (optab): Handle the new unspecs.
3507 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3509 (lr): Handle the new unspecs.
3511 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3512 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3513 (sve_int_qsub_op): New int attributes.
3514 (sve_fp_op, rot): Handle the new unspecs.
3515 * config/aarch64/aarch64-sve-builtins.h
3516 (function_resolver::require_matching_pointer_type): Declare.
3517 (function_resolver::resolve_unary): Add an optional boolean argument.
3518 (function_resolver::finish_opt_n_resolution): Add an optional
3519 type_suffix_index argument.
3520 (gimple_folder::redirect_call): Declare.
3521 (gimple_expander::prepare_gather_address_operands): Add an optional
3523 * config/aarch64/aarch64-sve-builtins.cc: Include
3524 aarch64-sve-builtins-sve2.h.
3525 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3526 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3527 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3528 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3529 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3530 (TYPES_s_unsigned): Likewise.
3531 (TYPES_s_integer): Use TYPES_s_unsigned.
3532 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3533 (TYPES_sd_integer): Use them.
3534 (TYPES_d_unsigned): New macro.
3535 (TYPES_d_integer): Use it.
3536 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3537 (TYPES_cvt_narrow): Likewise.
3538 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3539 (preds_mx): New variable.
3540 (function_builder::add_overloaded_function): Allow the new feature
3541 set to be more restrictive than the original one.
3542 (function_resolver::infer_pointer_type): Remove qualifiers from
3543 the pointer type before printing it.
3544 (function_resolver::require_matching_pointer_type): New function.
3545 (function_resolver::resolve_sv_displacement): Handle functions
3546 that don't support 32-bit vector indices or svint32_t vector offsets.
3547 (function_resolver::finish_opt_n_resolution): Take the inferred type
3548 as a separate argument.
3549 (function_resolver::resolve_unary): Optionally treat all forms in
3550 the same way as normal merging functions.
3551 (gimple_folder::redirect_call): New function.
3552 (function_expander::prepare_gather_address_operands): Add an argument
3553 that says whether scaled forms are available. If they aren't,
3554 handle scaling of vector indices and don't add the extension and
3556 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3557 fall back to using cond_* instead.
3558 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3559 Split out the member variables into...
3560 (rtx_code_function_base): ...this new base class.
3561 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3562 (unspec_based_function): Split out the member variables into...
3563 (unspec_based_function_base): ...this new base class.
3564 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3565 (unspec_based_function_exact_insn): New class.
3566 (unspec_based_add_function, unspec_based_add_lane_function)
3567 (unspec_based_lane_function, unspec_based_pred_function)
3568 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3569 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3570 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3572 (unspec_based_fused_function): New class.
3573 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3574 (unspec_based_fused_lane_function): New class.
3575 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3577 (CODE_FOR_MODE1): New macro.
3578 (fixed_insn_function): New class.
3579 (while_comparison): Likewise.
3580 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3581 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3582 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3583 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3584 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3585 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3586 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3587 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3588 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3589 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3590 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3591 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3592 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3593 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3594 Also add an initial argument for unary_convert_narrowt, regardless
3595 of the predication type.
3596 (build_32_64): Allow loads and stores to specify MODE_none.
3597 (build_sv_index64, build_sv_uint_offset): New functions.
3598 (long_type_suffix): New function.
3599 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3600 (binary_imm_long_base, load_gather_sv_base): Likewise.
3601 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3602 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3603 (unary_narrowb_base, unary_narrowt_base): Likewise.
3604 (binary_long_lane_def, binary_long_lane): New shape.
3605 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3606 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3607 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3608 (binary_to_uint_def, binary_to_uint): Likewise.
3609 (binary_wide_def, binary_wide): Likewise.
3610 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3611 (compare_def, compare): Likewise.
3612 (compare_ptr_def, compare_ptr): Likewise.
3613 (load_ext_gather_index_restricted_def,
3614 load_ext_gather_index_restricted): Likewise.
3615 (load_ext_gather_offset_restricted_def,
3616 load_ext_gather_offset_restricted): Likewise.
3617 (load_gather_sv_def): Inherit from load_gather_sv_base.
3618 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3619 (shift_left_imm_def, shift_left_imm): Likewise.
3620 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3621 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3622 (store_scatter_index_restricted_def,
3623 store_scatter_index_restricted): Likewise.
3624 (store_scatter_offset_restricted_def,
3625 store_scatter_offset_restricted): Likewise.
3626 (tbl_tuple_def, tbl_tuple): Likewise.
3627 (ternary_long_lane_def, ternary_long_lane): Likewise.
3628 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3629 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3630 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3631 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3632 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3633 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3634 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3635 (ternary_uint_def, ternary_uint): Likewise.
3636 (unary_convert): Fix typo in comment.
3637 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3638 (unary_long_def, unary_long): Likewise.
3639 (unary_narrowb_def, unary_narrowb): Likewise.
3640 (unary_narrowt_def, unary_narrowt): Likewise.
3641 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3642 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3643 (unary_to_int_def, unary_to_int): Likewise.
3644 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3645 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3646 (svasrd_impl): Delete.
3647 (svcadd_impl::expand): Handle integer operations too.
3648 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3649 new functions to derive the unspec numbers.
3650 (svmla_svmls_lane_impl): Replace with...
3651 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3652 integer operations too.
3653 (svwhile_impl): Rename to...
3654 (svwhilelx_impl): ...this and inherit from while_comparison.
3655 (svasrd): Use unspec_based_function.
3656 (svmla_lane): Use svmla_lane_impl.
3657 (svmls_lane): Use svmls_lane_impl.
3658 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3659 (svwhilele, svwhilelt): Use svwhilelx_impl.
3660 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3661 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3662 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3663 * config/aarch64/aarch64-sve-builtins.def: Include
3664 aarch64-sve-builtins-sve2.def.
3666 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3668 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3669 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3670 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3671 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3672 immediates as well as vector ones.
3673 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3674 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3675 (aarch64_sve_qsub_immediate): Update calls accordingly.
3677 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3679 * config/aarch64/aarch64-sve2.md: Add banner comments.
3680 (<su>mulh<r>s<mode>3): Move further up file.
3681 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3682 (*aarch64_sve2_sra<mode>): Move further down file.
3683 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3685 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3687 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3689 (while_optab_cmp): Handle them.
3690 * config/aarch64/aarch64-sve.md
3691 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3692 and add a "@" marker.
3693 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3694 instead of gen_aarch64_sve2_while_ptest.
3695 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3697 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3699 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3700 (UNSPEC_WHILELE): ...this.
3701 (UNSPEC_WHILE_LO): Rename to...
3702 (UNSPEC_WHILELO): ...this.
3703 (UNSPEC_WHILE_LS): Rename to...
3704 (UNSPEC_WHILELS): ...this.
3705 (UNSPEC_WHILE_LT): Rename to...
3706 (UNSPEC_WHILELT): ...this.
3707 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3708 (cmp_op, while_optab_cmp): Likewise.
3709 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3710 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3711 (svwhilelt): Likewise.
3713 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3715 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3716 (unary_to_uint): Define.
3717 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3718 (unary_count): Rename to...
3719 (unary_to_uint_def, unary_to_uint): ...this.
3720 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3722 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3724 * config/aarch64/aarch64-sve-builtins-functions.h
3725 (code_for_mode_function): New class.
3726 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3727 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3728 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3729 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3730 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3732 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3734 * config/aarch64/iterators.md (addsub): New code attribute.
3735 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3737 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3738 in the asm string and attributes. Fix indentation.
3739 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3741 (@aarch64_sve_<optab><mode>): ...this.
3742 * config/aarch64/aarch64-sve-builtins.h
3743 (function_expander::expand_signed_unpred_op): Delete.
3744 * config/aarch64/aarch64-sve-builtins.cc
3745 (function_expander::expand_signed_unpred_op): Likewise.
3746 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3747 try using code_for_aarch64_sve instead.
3748 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3749 (svqsub_impl): Likewise.
3750 (svqadd, svqsub): Use rtx_code_function instead.
3752 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3754 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3755 (HADDSUB, sur, addsub): Remove them.
3757 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3759 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3760 functions the same way, rather than singling out those that
3761 aren't mapped directly to optabs.
3763 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3765 * target.def (compatible_vector_types_p): New target hook.
3766 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3767 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3768 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3769 * doc/tm.texi: Regenerate.
3770 * gimple-expr.c: Include target.h.
3771 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3772 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3774 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3775 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3776 Use the original predicate if it already has a suitable type.
3778 2020-01-09 Martin Jambor <mjambor@suse.cz>
3780 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3781 resolve_speculation and redirect_call_stmt_to_callee static. Change
3782 return type of set_call_stmt to cgraph_edge *.
3783 * auto-profile.c (afdo_indirect_call): Adjust call to
3784 redirect_call_stmt_to_callee.
3785 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3786 make the this pointer explicit, adjust self-recursive calls and the
3787 call top make_direct. Return the resulting edge.
3788 (cgraph_edge::remove): Make this pointer explicit.
3789 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3790 (cgraph_edge::make_direct): Likewise, adjust call to
3791 resolve_speculation.
3792 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3793 call to set_call_stmt.
3794 (cgraph_update_edges_for_call_stmt_node): Update call to
3795 set_call_stmt and remove.
3796 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3797 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3798 (cgraph_node::create_edge_including_clones): Moved "first" definition
3799 of edge to the block where it was used. Adjusted calls to
3801 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3802 cgraph_edge::remove.
3803 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3804 make_direct and redirect_call_stmt_to_callee.
3805 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3806 resolve_speculation and make_direct.
3807 * ipa-inline-transform.c (inline_transform): Adjust call to
3808 redirect_call_stmt_to_callee.
3809 (check_speculations_1):: Adjust call to resolve_speculation.
3810 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3811 resolve-speculation.
3812 (inline_small_functions): Adjust call to resolve_speculation.
3813 (ipa_inline): Likewise.
3814 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3816 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3817 safe with regards to edge removal, adjust calls to
3818 redirect_call_stmt_to_callee.
3819 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3820 and redirect_call_stmt_to_callee.
3821 * multiple_target.c (create_dispatcher_calls): Adjust call to
3822 redirect_call_stmt_to_callee
3823 (redirect_to_specific_clone): Likewise.
3824 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3825 Adjust calls to cgraph_edge::remove.
3826 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3827 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3828 (expand_call_inline): Adjust call to cgraph_edge::remove.
3830 2020-01-09 Martin Liska <mliska@suse.cz>
3832 * params.opt: Set Optimization for
3833 param_max_speculative_devirt_maydefs.
3835 2020-01-09 Martin Sebor <msebor@redhat.com>
3839 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3841 2020-01-09 Martin Liska <mliska@suse.cz>
3843 * auto-profile.c (auto_profile): Use opt_for_fn
3845 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3846 (propagate_vals_across_arith_jfunc): Likewise.
3847 (hint_time_bonus): Likewise.
3848 (incorporate_penalties): Likewise.
3849 (good_cloning_opportunity_p): Likewise.
3850 (perform_estimation_of_a_value): Likewise.
3851 (estimate_local_effects): Likewise.
3852 (ipcp_propagate_stage): Likewise.
3853 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3854 (set_switch_stmt_execution_predicate): Likewise.
3855 (analyze_function_body): Likewise.
3856 * ipa-inline-analysis.c (offline_size): Likewise.
3857 * ipa-inline.c (early_inliner): Likewise.
3858 * ipa-prop.c (ipa_analyze_node): Likewise.
3859 (ipcp_transform_function): Likewise.
3860 * ipa-sra.c (process_scan_results): Likewise.
3861 (ipa_sra_summarize_function): Likewise.
3862 * params.opt: Rename ipcp-unit-growth to
3863 ipa-cp-unit-growth. Add Optimization for various
3864 IPA-related parameters.
3866 2020-01-09 Richard Biener <rguenther@suse.de>
3869 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3871 2020-01-09 Richard Biener <rguenther@suse.de>
3873 PR tree-optimization/93040
3874 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3876 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3878 * common/config/avr/avr-common.c (avr_option_optimization_table)
3879 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3881 2020-01-09 Martin Liska <mliska@suse.cz>
3883 * cgraphclones.c (symbol_table::materialize_all_clones):
3884 Use cgraph_node::dump_name.
3886 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3889 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3890 output_operand_lossage instead of gcc_unreachable.
3891 * doc/md.texi (riscv f constraint): Fix typo.
3894 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3895 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3896 CONST_SCALAR_INT_P instead of CONST_INT_P.
3897 (*subv<mode>4_1): Rename to ...
3898 (subv<mode>4_1): ... this.
3899 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3900 define_insn_and_split patterns.
3901 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3904 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3906 * vec.c (class selftest::count_dtor): New class.
3907 (selftest::test_auto_delete_vec): New test.
3908 (selftest::vec_c_tests): Call it.
3909 * vec.h (class auto_delete_vec): New class template.
3910 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3912 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3914 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3916 2020-01-08 Jim Wilson <jimw@sifive.com>
3918 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3919 use of TLS_MODEL_LOCAL_EXEC when not pic.
3921 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3923 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3926 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3929 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3930 *stack_protect_set_3 peephole2): Also check that the second
3931 insns source is general_operand.
3934 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3935 predicate for output operand instead of register_operand.
3936 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3937 memory destination and non-memory operands[2].
3939 2020-01-08 Martin Liska <mliska@suse.cz>
3941 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3942 ::dump_asm_name instead of (::name or ::asm_name).
3943 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3944 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3945 (analyze_functions): Likewise.
3946 (expand_all_functions): Likewise.
3947 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3948 (propagate_bits_across_jump_function): Likewise.
3949 (dump_profile_updates): Likewise.
3950 (ipcp_store_bits_results): Likewise.
3951 (ipcp_store_vr_results): Likewise.
3952 * ipa-devirt.c (dump_targets): Likewise.
3953 * ipa-fnsummary.c (analyze_function_body): Likewise.
3954 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3955 (process_hsa_functions): Likewise.
3956 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3957 (set_alias_uids): Likewise.
3958 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3959 * ipa-inline.c (recursive_inlining): Likewise.
3960 (inline_to_all_callers_1): Likewise.
3961 (ipa_inline): Likewise.
3962 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3963 (ipa_propagate_frequency): Likewise.
3964 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3965 (remove_described_reference): Likewise.
3966 * ipa-pure-const.c (worse_state): Likewise.
3967 (check_retval_uses): Likewise.
3968 (analyze_function): Likewise.
3969 (propagate_pure_const): Likewise.
3970 (propagate_nothrow): Likewise.
3971 (dump_malloc_lattice): Likewise.
3972 (propagate_malloc): Likewise.
3973 (pass_local_pure_const::execute): Likewise.
3974 * ipa-visibility.c (optimize_weakref): Likewise.
3975 (function_and_variable_visibility): Likewise.
3976 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3977 (ipa_discover_variable_flags): Likewise.
3978 * lto-streamer-out.c (output_function): Likewise.
3979 (output_constructor): Likewise.
3980 * tree-inline.c (copy_bb): Likewise.
3981 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3982 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3984 2020-01-08 Richard Biener <rguenther@suse.de>
3987 * tree-eh.c (sink_clobbers): Update virtual operands for
3988 the first and last stmt only. Add a dry-run capability.
3989 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3990 after CFG manipulations and in RPO order to catch all
3991 secondary opportunities reliably.
3993 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3996 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3998 2019-01-08 Richard Biener <rguenther@suse.de>
4001 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
4002 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
4003 virtual operand, also updating SSA use.
4004 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
4005 Update stmt after resetting virtual operand.
4006 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
4007 * gimple-iterator.c (gsi_remove): When not removing the stmt
4008 permanently do not delink immediate uses or mark the stmt modified.
4010 2020-01-08 Martin Liska <mliska@suse.cz>
4012 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4013 (ipa_call_context::estimate_size_and_time): Likewise.
4014 (inline_analyze_function): Likewise.
4016 2020-01-08 Martin Liska <mliska@suse.cz>
4018 * cgraph.c (cgraph_node::dump): Use systematically
4021 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4023 Add -nodevicespecs option for avr.
4026 * config/avr/avr.opt (-nodevicespecs): New driver option.
4027 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4028 "-specs=device-specs/..." if that option is not set.
4029 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4031 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4033 Implement 64-bit double functions for avr.
4036 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4037 --with-double-comparison.
4038 * doc/install.texi: Document them.
4039 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4040 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4041 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4042 * doc/invoke.texi (AVR Built-in Macros): Document them.
4043 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4044 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4045 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4047 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4050 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4051 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4052 when only building rm-profile multilibs.
4054 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4057 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4058 lattice for a value to check.
4059 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4060 finite propagation in self-recursive scc.
4062 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4064 * ipa-inline.c (caller_growth_limits): Restore the AND.
4066 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4068 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4069 (VEC_ALLREG_ALT): New iterator.
4070 (VEC_ALLREG_INT_MODE): New iterator.
4071 (VCMP_MODE): New iterator.
4072 (VCMP_MODE_INT): New iterator.
4073 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4074 (vec_cmp<u>v64qidi): New define_expand.
4075 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4076 (vec_cmpu<mode>di_exec): New define_expand.
4077 (vec_cmp<u>v64qidi_exec): New define_expand.
4078 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4079 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4080 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4081 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4082 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4083 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4084 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4085 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4086 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4087 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4089 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4090 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4092 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4094 * config/gcn/constraints.md (DA): Update description and match.
4096 (Db): New constraint.
4097 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4099 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4100 Implement 'Db' mixed immediate type.
4101 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4102 (addcv64si3_dup<exec_vcc>): Delete.
4103 (subcv64si3<exec_vcc>): Rework constraints.
4104 (addv64di3): Rework constraints.
4105 (addv64di3_exec): Rework constraints.
4106 (subv64di3): Rework constraints.
4107 (addv64di3_dup): Delete.
4108 (addv64di3_dup_exec): Delete.
4109 (addv64di3_zext): Rework constraints.
4110 (addv64di3_zext_exec): Rework constraints.
4111 (addv64di3_zext_dup): Rework constraints.
4112 (addv64di3_zext_dup_exec): Rework constraints.
4113 (addv64di3_zext_dup2): Rework constraints.
4114 (addv64di3_zext_dup2_exec): Rework constraints.
4115 (addv64di3_sext_dup2): Rework constraints.
4116 (addv64di3_sext_dup2_exec): Rework constraints.
4118 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4120 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4121 existing target checks.
4123 2020-01-07 Richard Biener <rguenther@suse.de>
4125 * doc/install.texi: Bump minimal supported MPC version.
4127 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4129 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4130 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4131 * langhooks.c: Include stor-layout.h.
4132 (lhd_simulate_enum_decl): New function.
4133 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4134 handle_arm_sve_h for the LTO frontend.
4135 (register_vector_type): Cope with null returns from pushdecl.
4137 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4139 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4140 (aarch64_sve::nvectors_if_data_type): Replace with...
4141 (aarch64_sve::builtin_type_p): ...this.
4142 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4143 (find_vector_type): Delete.
4144 (add_sve_type_attribute): New function.
4145 (lookup_sve_type_attribute): Likewise.
4146 (register_builtin_types): Add an "SVE type" attribute to each type.
4147 (register_tuple_type): Likewise.
4148 (svbool_type_p, nvectors_if_data_type): Delete.
4149 (mangle_builtin_type): Use lookup_sve_type_attribute.
4150 (builtin_type_p): Likewise. Add an overload that returns the
4151 number of constituent vector and predicate registers.
4152 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4153 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4154 instead of aarch64_sve_argument_p.
4155 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4156 (aarch64_pass_by_reference): Likewise.
4157 (aarch64_function_value_1): Likewise.
4158 (aarch64_return_in_memory): Likewise.
4159 (aarch64_layout_arg): Likewise.
4161 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4163 PR tree-optimization/93156
4164 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4165 least significant bit is always clear.
4167 PR tree-optimization/93118
4168 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4169 simplifier with two intermediate conversions.
4171 2020-01-07 Martin Liska <mliska@suse.cz>
4173 * params.opt: Add Optimization for various parameters.
4175 2020-01-07 Martin Liska <mliska@suse.cz>
4178 * doc/extend.texi: Explain cloning for target_clone
4181 2020-01-07 Martin Liska <mliska@suse.cz>
4183 PR tree-optimization/92860
4184 * common.opt: Make in Optimization option
4185 as it is affected by -O0, which is an Optimization
4187 * tree-inline.c (tree_inlinable_function_p):
4188 Use opt_for_fn for warn_inline.
4189 (expand_call_inline): Likewise.
4191 2020-01-07 Martin Liska <mliska@suse.cz>
4193 PR tree-optimization/92860
4194 * common.opt: Make flag_ree as optimization
4197 2020-01-07 Martin Liska <mliska@suse.cz>
4199 PR optimization/92860
4200 * params.opt: Mark param_min_crossjump_insns with Optimization
4203 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4205 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4206 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4208 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4210 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4211 helper function to return the valid addressing formats for a given
4212 hard register and mode.
4213 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4215 * config/rs6000/constraints.md (Q constraint): Update
4217 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4220 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4221 Use 'Q' for doing vector extract from memory.
4222 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4224 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4225 doing vector extract from memory.
4226 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4227 extract from memory.
4229 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4230 for the offset being 34-bits when -mcpu=future is used.
4232 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4234 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4235 instead of cmpib_comparison_operator in cmpib patterns.
4236 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4237 of cmpib_comparison_operator. Revise comment.
4239 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4241 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4242 in an IFN_DIV_POW2 node to be equal.
4244 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4246 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4247 (vect_check_scalar_mask): ...this.
4248 (vectorizable_store, vectorizable_load): Update call accordingly.
4249 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4250 argument in calls to conditional internal functions.
4252 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4254 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4255 '0' matching inputs.
4256 (subv64di3_exec): Likewise.
4258 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4260 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4261 * doc/md.texi (movstr): Likewise.
4263 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4265 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4268 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4270 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4272 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4273 to a temporary file and use move-if-change to update the real
4274 file where necessary.
4276 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4278 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4279 rather than Upa for CPY /M.
4281 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4283 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4286 2020-01-06 Martin Liska <mliska@suse.cz>
4288 PR tree-optimization/92860
4289 * params.opt: Mark param_max_combine_insns with Optimization
4292 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4295 * config/i386/i386.md (SWIDWI): New mode iterator.
4296 (DWI, dwi): Add TImode variants.
4297 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4298 <general_hilo_operand> instead of <general_operand>. Use
4299 CONST_SCALAR_INT_P instead of CONST_INT_P.
4300 (*addv<mode>4_1): Rename to ...
4301 (addv<mode>4_1): ... this.
4302 (QWI): New mode attribute.
4303 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4304 define_insn_and_split patterns.
4305 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4307 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4308 <general_hilo_operand> instead of <general_operand>.
4309 (*addcarry<mode>_1): New define_insn.
4310 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4312 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4314 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4315 Use "call" instead of "set".
4317 2020-01-03 Martin Jambor <mjambor@suse.cz>
4320 * ipa-cp.c (print_all_lattices): Skip functions without info.
4322 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4325 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4326 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4327 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4328 for 'e' simd clones.
4331 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4333 (mprefer-vector-width=): Add Save.
4334 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4335 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4336 (ix86_debug_options, ix86_function_specific_print): Adjust
4337 ix86_target_string callers.
4338 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4339 (ix86_valid_target_attribute_tree): Likewise.
4340 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4341 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4342 ix86_target_string caller.
4345 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4346 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4347 instead of gen_int_shift_amount + convert_modes.
4349 PR rtl-optimization/93088
4350 * loop-iv.c (find_single_def_src): Punt after looking through
4351 128 reg copies for regs with single definitions. Move definitions
4354 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4356 * config/arm/arm-c.c (arm_cpu_builtins): Define
4357 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4358 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4359 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4360 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4361 * config/arm/arm-tables.opt: Regenerated.
4362 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4363 arm_arch_i8mm and arm_arch_bf16 when enabled.
4364 * config/arm/arm.h (TARGET_I8MM): New macro.
4365 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4366 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4367 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4368 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4369 (v8_6_a_simd_variants): New.
4370 (v8_*_a_simd_variants): Add i8mm and bf16.
4371 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4373 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4376 * predict.c (compute_function_frequency): Don't call
4377 warn_function_cold on functions that already have cold attribute.
4379 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4382 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4383 COMDAT group function labels in .data.rel.ro.local section.
4384 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4387 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4388 comparison_operator in B and S integer comparisons. Likewise, use
4389 ordered_comparison_operator instead of cmpib_comparison_operator in
4391 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4393 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4395 Update copyright years.
4397 * gcc.c (process_command): Update copyright notice dates.
4398 * gcov-dump.c (print_version): Ditto.
4399 * gcov.c (print_version): Ditto.
4400 * gcov-tool.c (print_version): Ditto.
4401 * gengtype.c (create_file): Ditto.
4402 * doc/cpp.texi: Bump @copying's copyright year.
4403 * doc/cppinternals.texi: Ditto.
4404 * doc/gcc.texi: Ditto.
4405 * doc/gccint.texi: Ditto.
4406 * doc/gcov.texi: Ditto.
4407 * doc/install.texi: Ditto.
4408 * doc/invoke.texi: Ditto.
4410 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4412 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4415 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4417 PR tree-optimization/93098
4418 * match.pd (popcount): For shift amounts, use integer_onep
4419 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4420 tests. Make sure that precision is power of two larger than or equal
4421 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4422 instead of ULL suffixed constants. Formatting fixes.
4424 Copyright (C) 2020 Free Software Foundation, Inc.
4426 Copying and distribution of this file, with or without modification,
4427 are permitted in any medium without royalty provided the copyright
4428 notice and this notice are preserved.