1 2020-02-18 Feng Xue <fxue@os.amperecomputing.com>
4 * ipa-cp.c (self_recursively_generated_p): Mark self-dependent value as
5 self-recursively generated.
7 2020-02-21 Iain Sandoe <iain@sandoe.co.uk>
10 * config/darwin-c.c (pop_field_alignment): Adjust quoting of
13 2020-02-21 Mihail Ionescu <mihail.ionescu@arm.com>
15 * doc/sourcebuild.texi (arm_v8_1m_mve_ok):
16 Document new target supports option.
18 2020-02-21 Dennis Zhang <dennis.zhang@arm.com>
20 * config/arm/arm_neon.h (vmmlaq_s32, vmmlaq_u32, vusmmlaq_s32): New.
21 * config/arm/arm_neon_builtins.def (smmla, ummla, usmmla): New.
22 * config/arm/iterators.md (MATMUL): New iterator.
23 (sup): Add UNSPEC_MATMUL_S, UNSPEC_MATMUL_U, and UNSPEC_MATMUL_US.
24 (mmla_sfx): New attribute.
25 * config/arm/neon.md (neon_<sup>mmlav16qi): New.
26 * config/arm/unspecs.md (UNSPEC_MATMUL_S, UNSPEC_MATMUL_U): New.
27 (UNSPEC_MATMUL_US): New.
29 2020-02-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
31 * config/arm/arm.md: Prevent scalar shifts from being used when big
34 2020-02-21 Jan Hubicka <hubicka@ucw.cz>
35 Richard Biener <rguenther@suse.de>
37 PR tree-optimization/93586
38 * tree-ssa-alias.c (nonoverlapping_array_refs_p): Finish array walk
39 after mismatched array refs; do not sure type size information to
40 recover from unmatched referneces with !flag_strict_aliasing_p.
42 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
44 * config/gcn/gcn-valu.md (gather_load<mode>): Rename to ...
45 (gather_load<mode>v64si): ... this and set operand 2 to V64SI.
46 (scatter_store<mode>): Rename to ...
47 (scatter_store<mode>v64si): ... this and set operand 1 to V64SI.
48 (scatter<mode>_exec): Delete. Move contents ...
49 (mask_scatter_store<mode>): ... here, and rename that to ...
50 (mask_gather_load<mode>v64si): ... this. Set operand 2 to V64SI.
51 Remove mode conversion.
52 (mask_gather_load<mode>): Rename to ...
53 (mask_scatter_store<mode>v64si): ... this. Set operand 1 to V64SI.
54 Remove mode conversion.
55 * config/gcn/gcn.c (gcn_expand_scaled_offsets): Remove mode conversion.
57 2020-02-21 Martin Jambor <mjambor@suse.cz>
59 PR tree-optimization/93845
60 * tree-sra.c (verify_sra_access_forest): Only test access size of
63 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
65 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
66 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
67 (addv64di3_exec): Likewise.
68 (subv64di3): Likewise.
69 (subv64di3_exec): Likewise.
70 (addv64di3_zext): Likewise.
71 (addv64di3_zext_exec): Likewise.
72 (addv64di3_zext_dup): Likewise.
73 (addv64di3_zext_dup_exec): Likewise.
74 (addv64di3_zext_dup2): Likewise.
75 (addv64di3_zext_dup2_exec): Likewise.
76 (addv64di3_sext_dup2): Likewise.
77 (addv64di3_sext_dup2_exec): Likewise.
78 (<expander>v64di3): Likewise.
79 (<expander>v64di3_exec): Likewise.
80 (*<reduc_op>_dpp_shr_v64di): Likewise.
81 (*plus_carry_dpp_shr_v64di): Likewise.
82 * config/gcn/gcn.md (adddi3): Likewise.
83 (addptrdi3): Likewise.
84 (<expander>di3): Likewise.
86 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
88 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
90 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
92 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
93 support. Use aarch64_emit_mult instead of emitting multiplication
94 instructions directly.
95 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
96 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
98 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
100 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
101 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
102 instead of emitting multiplication instructions directly.
103 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
104 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
105 (@aarch64_frecps<mode>): New expanders.
107 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
109 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
110 on and produce uint64_ts rather than ints.
111 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
112 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
114 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
116 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
117 an unused xmsk register when handling approximate rsqrt.
119 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
121 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
122 flag_finite_math_only condition.
124 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
127 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
128 to destination operand for shufps alternative.
129 (*vec_extractv2si_1): Ditto.
131 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
134 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
137 2020-02-20 Martin Liska <mliska@suse.cz>
140 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
142 2020-02-20 Martin Liska <mliska@suse.cz>
145 * common/config/avr/avr-common.c: Remote trailing "|".
147 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
149 * collect2.c (maybe_run_lto_and_relink): Fix typo in
152 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
154 PR tree-optimization/93767
155 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
156 access-size bias from the offset calculations for negative strides.
158 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
160 * collect2.c (c_file, o_file): Make const again.
161 (ldout,lderrout, dump_ld_file): Remove.
162 (tool_cleanup): Avoid calling not signal-safe functions.
163 (maybe_run_lto_and_relink): Avoid possible signal handler
164 access to unintialzed memory (lto_o_files).
165 (main): Avoid leaking temp files in $TMPDIR.
166 Initialize c_file/o_file with concat, which avoids exposing
167 uninitialized memory to signal handler, which calls unlink(!).
168 Avoid calling maybe_unlink when the main function returns,
169 since the atexit handler is already doing this.
170 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
172 2020-02-19 Martin Jambor <mjambor@suse.cz>
174 PR tree-optimization/93776
175 * tree-sra.c (create_access): Do not create zero size accesses.
176 (get_access_for_expr): Do not search for zero sized accesses.
178 2020-02-19 Martin Jambor <mjambor@suse.cz>
180 PR tree-optimization/93667
181 * tree-sra.c (scalarizable_type_p): Return false if record fields
182 do not follow wach other.
184 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
186 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
187 rather than fmv.x.s/fmv.s.x.
189 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
191 * config/aarch64/aarch64-simd-builtins.def
192 (intrinsic_vec_smult_lo_): New.
193 (intrinsic_vec_umult_lo_): Likewise.
194 (vec_widen_smult_hi_): Likewise.
195 (vec_widen_umult_hi_): Likewise.
196 * config/aarch64/aarch64-simd.md
197 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
198 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
199 (vmull_high_s16): Likewise.
200 (vmull_high_s32): Likewise.
201 (vmull_high_u8): Likewise.
202 (vmull_high_u16): Likewise.
203 (vmull_high_u32): Likewise.
204 (vmull_s8): Likewise.
205 (vmull_s16): Likewise.
206 (vmull_s32): Likewise.
207 (vmull_u8): Likewise.
208 (vmull_u16): Likewise.
209 (vmull_u32): Likewise.
211 2020-02-18 Martin Liska <mliska@suse.cz>
213 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
214 bootstrap by missing removal of invalid sanity check.
216 2020-02-18 Martin Liska <mliska@suse.cz>
219 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
220 Always compare LHS of gimple_assign.
222 2020-02-18 Martin Liska <mliska@suse.cz>
225 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
226 and return type of functions.
227 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
228 Drop MALLOC attribute for void functions.
229 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
230 malloc_state for a new VOID clone.
232 2020-02-18 Martin Liska <mliska@suse.cz>
235 * common.opt: Add -fprofile-reproducibility.
236 * doc/invoke.texi: Document it.
237 * value-prof.c (dump_histogram_value):
238 Document and support behavior for counters[0]
239 being a negative value.
240 (get_nth_most_common_value): Handle negative
241 counters[0] in respect to flag_profile_reproducible.
243 2020-02-18 Jakub Jelinek <jakub@redhat.com>
246 * cgraph.c (verify_speculative_call): Use speculative_id instead of
247 speculative_uid in messages. Remove trailing whitespace from error
248 message. Use num_speculative_call_targets instead of
249 num_speculative_targets in a message.
250 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
251 edge messages and stmt instead of cal_stmt in reference message.
253 PR tree-optimization/93780
254 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
255 before calling build_vector_type.
256 (execute_update_addresses_taken): Likewise.
259 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
260 typo, functoin -> function.
261 * tree.c (free_lang_data_in_decl): Fix comment typo,
262 functoin -> function.
263 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
265 2020-02-17 David Malcolm <dmalcolm@redhat.com>
267 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
269 (print_option_information): Don't call get_option_url if URLs
272 2020-02-17 Alexandre Oliva <oliva@adacore.com>
274 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
275 handling of register_common-less targets.
277 2020-02-17 Martin Liska <mliska@suse.cz>
280 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
282 2020-02-17 Martin Liska <mliska@suse.cz>
285 * config/rs6000/rs6000.c (rs6000_option_override_internal):
288 2020-02-17 Martin Liska <mliska@suse.cz>
291 * config/rx/elf.opt: Fix typo.
293 2020-02-17 Richard Biener <rguenther@suse.de>
296 * opts-global.c (print_ignored_options): Use inform and
299 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
302 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
304 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
307 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
308 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
310 2020-02-15 Jason Merrill <jason@redhat.com>
312 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
314 2020-02-15 Jakub Jelinek <jakub@redhat.com>
316 PR tree-optimization/93744
317 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
318 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
319 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
320 sure @2 in the first and @1 in the other patterns has no side-effects.
322 2020-02-15 David Malcolm <dmalcolm@redhat.com>
323 Bernd Edlinger <bernd.edlinger@hotmail.de>
327 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
328 * configure.ac (--with-diagnostics-urls): New configuration
329 option, based on --with-diagnostics-color.
330 (DIAGNOSTICS_URLS_DEFAULT): New define.
331 * config.h: Regenerate.
332 * configure: Regenerate.
333 * diagnostic.c (diagnostic_urls_init): Handle -1 for
334 DIAGNOSTICS_URLS_DEFAULT from configure-time
335 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
336 and TERM_URLS environment variable.
337 * diagnostic-url.h (diagnostic_url_format): New enum type.
338 (diagnostic_urls_enabled_p): rename to...
339 (determine_url_format): ... this, and change return type.
340 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
341 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
342 the linux console, and mingw.
343 (diagnostic_urls_enabled_p): rename to...
344 (determine_url_format): ... this, and adjust.
345 * pretty-print.h (pretty_printer::show_urls): rename to...
346 (pretty_printer::url_format): ... this, and change to enum.
347 * pretty-print.c (pretty_printer::pretty_printer,
348 pp_begin_url, pp_end_url, test_urls): Adjust.
349 * doc/install.texi (--with-diagnostics-urls): Document the new
350 configuration option.
351 (--with-diagnostics-color): Document the existing interaction
352 with GCC_COLORS better.
353 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
354 vindex reference. Update description of defaults based on the above.
355 (-fdiagnostics-color): Update description of how -fdiagnostics-color
356 interacts with GCC_COLORS.
358 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
361 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
362 conjunction with TARGET_GNU_TLS in early return.
364 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
366 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
367 the mode is not wider than UNITS_PER_WORD.
369 2020-02-14 Martin Jambor <mjambor@suse.cz>
371 PR tree-optimization/93516
372 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
373 access of the same type as the parent.
374 (propagate_subaccesses_from_lhs): Likewise.
376 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
379 * config/i386/avx512vbmi2intrin.h
380 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
381 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
382 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
383 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
384 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
385 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
386 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
387 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
388 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
389 of lacking a closing parenthesis.
390 * config/i386/avx512vbmi2vlintrin.h
391 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
392 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
393 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
394 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
395 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
396 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
397 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
398 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
399 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
400 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
401 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
402 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
403 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
404 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
405 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
406 _mm_shldi_epi32, _mm_mask_shldi_epi32,
407 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
408 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
410 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
413 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
414 the target function entry.
416 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
418 * common/config/arc/arc-common.c (arc_option_optimization_table):
419 Disable if-conversion step when optimized for size.
421 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
423 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
424 R12-R15 are always in ARCOMPACT16_REGS register class.
425 * config/arc/arc.opt (mq-class): Deprecate.
426 * config/arc/constraint.md ("q"): Remove dependency on mq-class
428 * doc/invoke.texi (mq-class): Update text.
429 * common/config/arc/arc-common.c (arc_option_optimization_table):
432 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
434 * config/arc/arc.c (arc_insn_cost): New function.
435 (TARGET_INSN_COST): Define.
436 * config/arc/arc.md (cost): New attribute.
437 (add_n): Use arc_nonmemory_operand.
438 (ashlsi3_insn): Likewise, also update constraints.
439 (ashrsi3_insn): Likewise.
441 (add_shift): Likewise.
442 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
444 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
446 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
448 (umulsidi_600): Likewise.
450 2020-02-13 Jakub Jelinek <jakub@redhat.com>
453 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
454 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
455 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
456 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
457 pass __A to the builtin followed by __W instead of __A followed by
459 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
460 _mm512_mask_popcnt_epi64): Likewise.
461 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
462 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
463 _mm256_mask_popcnt_epi64): Likewise.
465 PR tree-optimization/93582
466 * fold-const.h (shift_bytes_in_array_left,
467 shift_bytes_in_array_right): Declare.
468 * fold-const.c (shift_bytes_in_array_left,
469 shift_bytes_in_array_right): New function, moved from
470 gimple-ssa-store-merging.c, no longer static.
471 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
472 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
473 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
474 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
475 shift_bytes_in_array.
476 (verify_shift_bytes_in_array): Rename to ...
477 (verify_shift_bytes_in_array_left): ... this. Use
478 shift_bytes_in_array_left instead of shift_bytes_in_array.
479 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
480 instead of verify_shift_bytes_in_array.
481 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
482 / native_interpret_expr where the store covers all needed bits,
483 punt on PDP-endian, otherwise allow all involved offsets and sizes
484 not to be byte-aligned.
487 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
488 use const_0_to_255_operand predicate instead of immediate_operand.
489 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
490 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
491 vgf2p8affineinvqb_<mode><mask_name>,
492 vgf2p8affineqb_<mode><mask_name>): Drop mode from
493 const_0_to_255_operand predicated operands.
495 2020-02-12 Jeff Law <law@redhat.com>
497 * config/h8300/h8300.md (comparison shortening peepholes): Use
498 a mode iterator to merge the HImode and SImode peepholes.
500 2020-02-12 Jakub Jelinek <jakub@redhat.com>
503 * real.c (is_even): Make static. Function comment fix.
504 (is_halfway_below): Make static, don't assert R is not inf/nan,
505 instead return false for those. Small formatting fixes.
507 2020-02-12 Martin Sebor <msebor@redhat.com>
510 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
511 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
512 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
513 (strlen_check_and_optimize_call): Adjust callee name.
515 2020-02-12 Jeff Law <law@redhat.com>
517 * config/h8300/h8300.md (comparison shortening peepholes): Drop
518 (and (xor)) variant. Combine other two into single peephole.
520 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
522 PR rtl-optimization/93565
523 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
525 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
527 * config/aarch64/aarch64-simd.md
528 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
529 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
530 generating separate ADDV and zero_extend patterns.
531 * config/aarch64/iterators.md (VDQV_E): New iterator.
533 2020-02-12 Jeff Law <law@redhat.com>
535 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
536 expanders, splits, etc.
537 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
538 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
539 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
540 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
541 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
542 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
544 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
546 2020-02-12 Jakub Jelinek <jakub@redhat.com>
549 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
550 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
551 TARGET_AVX512DQ from condition.
552 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
553 instead of <mask_mode512bit_condition> in condition. If
554 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
556 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
559 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
562 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
564 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
566 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
567 where strlen is more legible.
568 (rs6000_builtin_vectorized_libmass): Ditto.
569 (rs6000_print_options_internal): Ditto.
571 2020-02-11 Martin Sebor <msebor@redhat.com>
573 PR tree-optimization/93683
574 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
576 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
578 * config/rs6000/predicates.md (cint34_operand): Rename the
579 -mprefixed-addr option to be -mprefixed.
580 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
581 the -mprefixed-addr option to be -mprefixed.
582 (OTHER_FUTURE_MASKS): Likewise.
583 (POWERPC_MASKS): Likewise.
584 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
585 the -mprefixed-addr option to be -mprefixed. Change error
586 messages to refer to -mprefixed.
587 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
589 (rs6000_legitimate_offset_address_p): Likewise.
590 (rs6000_mode_dependent_address): Likewise.
591 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
592 "-mprefixed" for target attributes and pragmas.
593 (address_to_insn_form): Rename the -mprefixed-addr option to be
595 (rs6000_adjust_insn_length): Likewise.
596 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
597 -mprefixed-addr option to be -mprefixed.
598 (ASM_OUTPUT_OPCODE): Likewise.
599 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
600 -mprefixed-addr option to be -mprefixed.
601 * config/rs6000/rs6000.opt (-mprefixed): Rename the
602 -mprefixed-addr option to be prefixed. Change the option from
603 being undocumented to being documented.
604 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
605 -mprefixed option. Update the -mpcrel documentation to mention
608 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
610 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
611 including FIRST_PSEUDO_REGISTER - 1.
612 * ira-color.c (print_hard_reg_set): Ditto.
614 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
616 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
617 (USTERNOP_QUALIFIERS): New define.
618 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
619 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
620 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
621 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
622 * config/arm/arm_neon.h (vusdot_s32): New.
623 (vusdot_lane_s32): New.
624 (vusdotq_lane_s32): New.
625 (vsudot_lane_s32): New.
626 (vsudotq_lane_s32): New.
627 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
628 * config/arm/iterators.md (DOTPROD_I8MM): New.
629 (sup, opsuffix): Add <us/su>.
630 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
631 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
633 2020-02-11 Richard Biener <rguenther@suse.de>
635 PR tree-optimization/93661
636 PR tree-optimization/93662
637 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
639 * tree-sra.c (get_access_for_expr): Likewise.
641 2020-02-10 Jakub Jelinek <jakub@redhat.com>
644 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
645 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
646 Change condition from TARGET_AVX2 to TARGET_AVX.
648 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
651 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
654 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
656 Try to generate zero-based comparisons.
657 * config/cris/cris.c (cris_reduce_compare): New function.
658 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
659 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
660 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
662 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
665 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
666 in Thumb state and also as a destination in Arm state. Add T16
669 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
671 * md.texi (Define Subst): Match closing paren in example.
673 2020-02-10 Jakub Jelinek <jakub@redhat.com>
677 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
678 arguments of strncmp.
680 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
683 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
684 but different source value.
685 (adjust_callers_for_value_intersection): New function.
686 (gather_edges_for_value): Adjust order of callers to let a
687 non-self-recursive caller be the first element.
688 (self_recursive_pass_through_p): Add a new parameter "simple", and
689 check generalized self-recursive pass-through jump function.
690 (self_recursive_agg_pass_through_p): Likewise.
691 (find_more_scalar_values_for_callers_subset): Compute value from
692 pass-through jump function for self-recursive.
693 (intersect_with_plats): Cleanup previous implementation code for value
694 itersection with self-recursive call edge.
695 (intersect_with_agg_replacements): Likewise.
696 (intersect_aggregates_with_edge): Deduce value from pass-through jump
697 function for self-recursive call edge. Cleanup previous implementation
698 code for value intersection with self-recursive call edge.
699 (decide_whether_version_node): Remove dead callers and adjust order
700 to let a non-self-recursive caller be the first element.
702 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
704 * recog.c: Move pass_split_before_sched2 code in front of
705 pass_split_before_regstack.
706 (pass_data_split_before_sched2): Rename pass to split3 from split4.
707 (pass_data_split_before_regstack): Rename pass to split4 from split3.
708 (rest_of_handle_split_before_sched2): Remove.
709 (pass_split_before_sched2::execute): Unconditionally call
711 (enable_split_before_sched2): New function.
712 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
713 (pass_split_before_regstack::gate): Ditto.
714 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
715 Update name check for renamed split4 pass.
716 * config/sh/sh.c (register_sh_passes): Update pass insertion
717 point for renamed split4 pass.
719 2020-02-09 Jakub Jelinek <jakub@redhat.com>
721 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
722 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
723 copying them around between host and target.
725 2020-02-08 Andrew Pinski <apinski@marvell.com>
728 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
729 STRICT_ALIGNMENT also.
731 2020-02-08 Jim Wilson <jimw@sifive.com>
734 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
736 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
737 Jakub Jelinek <jakub@redhat.com>
740 * config/i386/i386.h (CALL_USED_REGISTERS): Make
741 xmm16-xmm31 call-used even in 64-bit ms-abi.
743 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
745 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
746 (simd_ummla, simd_usmmla): Likewise.
747 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
748 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
751 2020-02-07 Richard Biener <rguenther@suse.de>
754 * tree-inline.c (fold_marked_statements): Do a PRE walk,
755 skipping unreachable regions.
756 (optimize_inline_calls): Skip folding stmts when we didn't
759 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
762 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
763 Don't return aggregates with only SFmode and DFmode in SSE
765 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
767 2020-02-07 Jakub Jelinek <jakub@redhat.com>
770 * config/rs6000/rs6000-logue.c
771 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
772 if it fails, move rs into end_addr and retry. Add
773 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
774 the insn pattern doesn't describe well what exactly happens to
778 * config/i386/predicates.md (avx_identity_operand): Remove.
779 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
780 (avx_<castmode><avxsizesuffix>_<castmode>,
781 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
782 a VEC_CONCAT of the operand and UNSPEC_CAST.
783 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
784 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
788 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
789 recog_data.insn if distance_non_agu_define changed it.
791 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
794 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
795 we only had X-FORM (reg+reg) addressing for vectors. Also before
796 ISA 3.0, we only had X-FORM addressing for scalars in the
797 traditional Altivec registers.
799 2020-02-06 <zhongyunde@huawei.com>
800 Vladimir Makarov <vmakarov@redhat.com>
802 PR rtl-optimization/93561
803 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
806 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
808 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
811 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
813 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
814 where the low and the high 32 bits are equal to each other specially,
815 with an rldimi instruction.
817 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
819 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
821 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
823 * config/arm/arm-tables.opt: Regenerate.
825 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
828 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
829 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
830 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
832 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
834 PR rtl-optimization/87763
835 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
837 2020-02-06 Delia Burduv <delia.burduv@arm.com>
839 * config/aarch64/aarch64-simd-builtins.def
840 (bfmlaq): New built-in function.
841 (bfmlalb): New built-in function.
842 (bfmlalt): New built-in function.
843 (bfmlalb_lane): New built-in function.
844 (bfmlalt_lane): New built-in function.
845 * config/aarch64/aarch64-simd.md
846 (aarch64_bfmmlaqv4sf): New pattern.
847 (aarch64_bfmlal<bt>v4sf): New pattern.
848 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
849 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
850 (vbfmlalbq_f32): New intrinsic.
851 (vbfmlaltq_f32): New intrinsic.
852 (vbfmlalbq_lane_f32): New intrinsic.
853 (vbfmlaltq_lane_f32): New intrinsic.
854 (vbfmlalbq_laneq_f32): New intrinsic.
855 (vbfmlaltq_laneq_f32): New intrinsic.
856 * config/aarch64/iterators.md (BF_MLA): New int iterator.
857 (bt): New int attribute.
859 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
861 * config/i386/i386.md (*pushtf): Emit "#" instead of
862 calling gcc_unreachable in insn output.
865 (*pushsf_rex64): Ditto for alternatives other than 1.
866 (*pushsf): Ditto for alternatives other than 1.
868 2020-02-06 Martin Liska <mliska@suse.cz>
870 PR gcov-profile/91971
871 PR gcov-profile/93466
872 * coverage.c (coverage_init): Revert mangling of
873 path into filename. It can lead to huge filename length.
874 Creation of subfolders seem more natural.
876 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
879 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
880 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
881 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
883 2020-02-06 Jakub Jelinek <jakub@redhat.com>
886 * config/i386/predicates.md (avx_identity_operand): New predicate.
887 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
888 define_insn_and_split.
891 * omp-low.c (use_pointer_for_field): For nested constructs, also
892 look for map clauses on target construct.
893 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
894 taskreg_nesting_level.
897 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
898 shared clause, call omp_notice_variable on outer context if any.
900 2020-02-05 Jason Merrill <jason@redhat.com>
903 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
904 non-zero address even if weak and not yet defined.
906 2020-02-05 Martin Sebor <msebor@redhat.com>
908 PR tree-optimization/92765
909 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
910 * tree-ssa-strlen.c (compute_string_length): Remove.
911 (determine_min_objsize): Remove.
912 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
913 Avoid using type size as the upper bound on string length.
914 (handle_builtin_string_cmp): Add an argument. Adjust.
915 (strlen_check_and_optimize_call): Pass additional argument to
916 handle_builtin_string_cmp.
918 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
920 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
921 (*pushdi2_rex64 peephole2): Unconditionally split after
923 (*ashl<mode>3_doubleword): Ditto.
924 (*<shift_insn><mode>3_doubleword): Ditto.
926 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
929 * config/rs6000/rs6000.c (get_vector_offset): Fix
931 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
933 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
935 2020-02-05 David Malcolm <dmalcolm@redhat.com>
938 (Special Functions for Debugging the Analyzer): Update description
939 of __analyzer_dump_exploded_nodes.
941 2020-02-05 Jakub Jelinek <jakub@redhat.com>
944 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
945 include sets and not clobbers in the vzeroupper pattern.
946 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
947 the parallel has 17 (64-bit) or 9 (32-bit) elts.
948 (*avx_vzeroupper_1): New define_insn_and_split.
951 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
952 don't run when !optimize.
953 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
956 2020-02-05 Richard Biener <rguenther@suse.de>
959 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
960 checks before matching calls.
962 2020-02-05 Jakub Jelinek <jakub@redhat.com>
964 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
965 function comment typo.
968 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
969 simd_clone_create failed when i == 0, adjust clone->nargs by
972 2020-02-05 Martin Liska <mliska@suse.cz>
975 * doc/invoke.texi: Document that one should
976 not combine ASLR and -fpch.
978 2020-02-04 Richard Biener <rguenther@suse.de>
980 PR tree-optimization/93538
981 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
983 2020-02-04 Richard Biener <rguenther@suse.de>
985 PR tree-optimization/91123
986 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
987 (vn_walk_cb_data::last_vuse): New member.
988 (vn_walk_cb_data::saved_operands): Likewsie.
989 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
990 (vn_walk_cb_data::push_partial_def): Use finish.
991 (vn_reference_lookup_2): Update last_vuse and use finish if
992 we've saved operands.
993 (vn_reference_lookup_3): Use finish and update calls to
994 push_partial_defs everywhere. When translating through
995 memcpy or aggregate copies save off operands and alias-set.
996 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
997 operation for redundant store removal.
999 2020-02-04 Richard Biener <rguenther@suse.de>
1001 PR tree-optimization/92819
1002 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
1003 generating more stmts than before.
1005 2020-02-04 Martin Liska <mliska@suse.cz>
1007 * config/arm/arm.c (arm_gen_far_branch): Move the function
1008 outside of selftests.
1010 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1012 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
1013 function to adjust PC-relative vector addresses.
1014 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
1015 handle vectors with PC-relative addresses.
1017 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1019 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
1021 (hard_reg_and_mode_to_addr_mask): Delete.
1022 (rs6000_adjust_vec_address): If the original vector address
1023 was REG+REG or REG+OFFSET and the element is not zero, do the add
1024 of the elements in the original address before adding the offset
1025 for the vector element. Use address_to_insn_form to validate the
1026 address using the register being loaded, rather than guessing
1027 whether the address is a DS-FORM or DQ-FORM address.
1029 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1031 * config/rs6000/rs6000.c (get_vector_offset): New helper function
1032 to calculate the offset in memory from the start of a vector of a
1033 particular element. Add code to keep the element number in
1034 bounds if the element number is variable.
1035 (rs6000_adjust_vec_address): Move calculation of offset of the
1036 vector element to get_vector_offset.
1037 (rs6000_split_vec_extract_var): Do not do the initial AND of
1038 element here, move the code to get_vector_offset.
1040 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
1042 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
1045 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
1047 * config/rs6000/constraints.md: Improve documentation.
1049 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
1052 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
1053 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
1055 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1057 * config.gcc: Remove "carrizo" support.
1058 * config/gcn/gcn-opts.h (processor_type): Likewise.
1059 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1060 * config/gcn/gcn.opt (gpu_type): Likewise.
1061 * config/gcn/t-omp-device: Likewise.
1063 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1066 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1067 * config/arm/arm.c (arm_gen_far_branch): New function
1069 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1071 2020-02-03 Julian Brown <julian@codesourcery.com>
1072 Tobias Burnus <tobias@codesourcery.com>
1074 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1076 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1079 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1080 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1083 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1085 PR rtl-optimization/91333
1086 * ira-color.c (struct allocno_color_data): Add member
1088 (init_allocno_threads): Set the member up.
1089 (bucket_allocno_compare_func): Add compare hard reg
1092 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1094 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1096 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1097 * config.in: Regenerated.
1098 * configure: Regenerated.
1099 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1100 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1101 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1103 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1105 * configure: Regenerate.
1107 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1109 PR rtl-optimization/91333
1110 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1111 reg preferences comparison up.
1113 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1115 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1116 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1117 aarch64-sve-builtins-base.h.
1118 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1119 aarch64-sve-builtins-base.cc.
1120 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1121 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1123 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1124 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1125 (svcvtnt): New functions.
1126 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1127 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1128 (svcvtnt): New functions.
1129 (svcvt): Add a form that converts f32 to bf16.
1130 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1131 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1133 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1134 Treat B as bfloat16_t.
1135 (ternary_bfloat_lane_base): New class.
1136 (ternary_bfloat_def): Likewise.
1137 (ternary_bfloat): New shape.
1138 (ternary_bfloat_lane_def): New class.
1139 (ternary_bfloat_lane): New shape.
1140 (ternary_bfloat_lanex2_def): New class.
1141 (ternary_bfloat_lanex2): New shape.
1142 (ternary_bfloat_opt_n_def): New class.
1143 (ternary_bfloat_opt_n): New shape.
1144 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1145 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1146 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1147 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1148 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1149 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1150 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1151 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1152 the pattern off the narrow mode instead of the wider one.
1153 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1154 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1155 (sve_fp_op): Handle them.
1156 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1157 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1159 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1161 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1162 * config/aarch64/aarch64-modes.def (BF): Move definition before
1163 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1164 (SVE_MODES): Handle BF modes.
1165 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1167 (aarch64_full_sve_mode): Likewise.
1168 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1170 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1171 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1172 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1173 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1175 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1177 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1178 (TYPES_all_data): Add bf16.
1179 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1180 (register_tuple_type): Increase buffer size.
1181 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1182 (bf16): New type suffix.
1183 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1184 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1185 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1186 Change type from all_data to all_arith.
1187 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1190 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1191 Matthew Malcomson <matthew.malcomson@arm.com>
1192 Richard Sandiford <richard.sandiford@arm.com>
1194 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1195 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1196 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1197 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1198 __ARM_FEATURE_MATMUL_FP64.
1199 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1200 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1201 be disabled at the same time.
1202 (f32mm): New extension.
1203 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1204 (AARCH64_FL_F64MM): Bump to the next bit up.
1205 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1206 (TARGET_SVE_F64MM): New macros.
1207 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1208 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1209 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1210 (UNSPEC_ZIP2Q): New unspeccs.
1211 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1212 (optab, sur, perm_insn): Handle the new unspecs.
1213 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1214 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1215 TARGET_SVE_F64MM instead of separate tests.
1216 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1217 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1218 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1219 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1220 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1221 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1222 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1223 (TYPES_s_signed): New macro.
1224 (TYPES_s_integer): Use it.
1225 (TYPES_d_float): New macro.
1226 (TYPES_d_data): Use it.
1227 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1228 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1229 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1230 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1231 (svmmla): New shape.
1232 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1233 template parameters.
1234 (ternary_resize2_lane_base): Likewise.
1235 (ternary_resize2_base): New class.
1236 (ternary_qq_lane_base): Likewise.
1237 (ternary_intq_uintq_lane_def): Likewise.
1238 (ternary_intq_uintq_lane): New shape.
1239 (ternary_intq_uintq_opt_n_def): New class
1240 (ternary_intq_uintq_opt_n): New shape.
1241 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1242 (ternary_uintq_intq_def): New class.
1243 (ternary_uintq_intq): New shape.
1244 (ternary_uintq_intq_lane_def): New class.
1245 (ternary_uintq_intq_lane): New shape.
1246 (ternary_uintq_intq_opt_n_def): New class.
1247 (ternary_uintq_intq_opt_n): New shape.
1248 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1249 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1250 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1251 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1253 (svdotprod_lane_impl): ...this new class.
1254 (svmmla_impl, svusdot_impl): New classes.
1255 (svdot_lane): Update to use svdotprod_lane_impl.
1256 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1257 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1259 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1260 function, with no types defined.
1261 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1262 AARCH64_FL_I8MM functions.
1263 (svmmla): New AARCH64_FL_F32MM function.
1264 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1265 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1266 AARCH64_FL_F64MM function.
1267 (REQUIRED_EXTENSIONS):
1269 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1271 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1274 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1276 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1277 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1278 (*movti_internal): Do not check for
1279 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1280 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1281 just after check for TARGET_AVX.
1282 (*movdf_internal): Ditto.
1283 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1284 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1285 * config/i386/sse.md (mov<mode>_internal): Only check
1286 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1287 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1288 (<sse>_andnot<mode>3<mask_name>): Move check for
1289 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1290 (<code><mode>3<mask_name>): Ditto.
1291 (*andnot<mode>3): Ditto.
1292 (*andnottf3): Ditto.
1293 (*<code><mode>3): Ditto.
1294 (*<code>tf3): Ditto.
1295 (*andnot<VI:mode>3): Remove
1296 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1297 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1298 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1299 (sse4_1_blendv<ssemodesuffix>): Ditto.
1300 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1301 Explain that tune applies to 128bit instructions only.
1303 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1305 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1306 to definition of hsa_kernel_description. Parse assembly to find SGPR
1307 and VGPR count of kernel and store in hsa_kernel_description.
1309 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1311 PR rtl-optimization/91838
1312 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1313 to truncate if allowed or reject combination.
1315 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1317 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1318 (find_inv_vars_cb): Likewise.
1320 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1322 * calls.c (special_function_p): Split out the check for DECL_NAME
1323 being non-NULL and fndecl being extern at file scope into a
1324 new maybe_special_function_p and call it. Drop check for fndecl
1325 being non-NULL that was after a usage of DECL_NAME (fndecl).
1326 * tree.h (maybe_special_function_p): New inline function.
1328 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1330 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1331 (mask_gather_load<mode>): ... here, and zero-initialize the
1333 (maskload<mode>di): Zero-initialize the destination.
1336 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1339 * doc/analyzer.texi (Limitations): Note that constraints on
1340 floating-point values are currently ignored.
1342 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1345 * symtab.c (symtab_node::noninterposable_alias): If localalias
1346 already exists, but is not usable, append numbers after it until
1347 a unique name is found. Formatting fix.
1350 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1353 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1355 * config/gcn/gcn.c (print_operand): Handle LTGT.
1356 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1358 2020-01-30 Richard Biener <rguenther@suse.de>
1360 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1361 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1363 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1365 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1366 without a DECL in .data.rel.ro.local.
1368 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1371 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1375 * config/i386/sse.md
1376 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1377 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1378 any_extend code iterator instead of always zero_extend.
1379 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1380 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1381 Use any_extend code iterator instead of always zero_extend.
1382 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1383 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1384 Use any_extend code iterator instead of always zero_extend.
1385 (*sse2_pmovmskb_ext): New define_insn.
1386 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1389 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1390 (*popcountsi2_zext_falsedep): New define_insn.
1392 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1394 * config.in: Regenerated.
1395 * configure: Regenerated.
1397 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1400 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1401 LLVM's assembler changed the default in version 9.
1403 2020-01-24 Jeff Law <law@redhat.com>
1405 PR tree-optimization/89689
1406 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1408 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1412 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1414 PR rtl-optimization/87763
1415 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1416 simplification to handle subregs as well as bare regs.
1417 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1419 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1422 * ira.c (ira): Revert use of simplified LRA algorithm.
1424 2020-01-29 Martin Jambor <mjambor@suse.cz>
1426 PR tree-optimization/92706
1427 * tree-sra.c (struct access): Fields first_link, last_link,
1428 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1429 next_rhs_queued and grp_rhs_queued respectively, new fields
1430 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1431 (struct assign_link): Field next renamed to next_rhs, new field
1432 next_lhs. Updated comment.
1433 (work_queue_head): Renamed to rhs_work_queue_head.
1434 (lhs_work_queue_head): New variable.
1435 (add_link_to_lhs): New function.
1436 (relink_to_new_repr): Also relink LHS lists.
1437 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1438 (add_access_to_lhs_work_queue): New function.
1439 (pop_access_from_work_queue): Renamed to
1440 pop_access_from_rhs_work_queue.
1441 (pop_access_from_lhs_work_queue): New function.
1442 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1444 (child_would_conflict_in_lacc): Renamed to
1445 child_would_conflict_in_acc. Adjusted parameter names.
1446 (create_artificial_child_access): New parameter set_grp_read, use it.
1447 (subtree_mark_written_and_enqueue): Renamed to
1448 subtree_mark_written_and_rhs_enqueue.
1449 (propagate_subaccesses_across_link): Renamed to
1450 propagate_subaccesses_from_rhs.
1451 (propagate_subaccesses_from_lhs): New function.
1452 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1455 2020-01-29 Martin Jambor <mjambor@suse.cz>
1457 PR tree-optimization/92706
1458 * tree-sra.c (struct access): Adjust comment of
1459 grp_total_scalarization.
1460 (find_access_in_subtree): Look for single children spanning an entire
1462 (scalarizable_type_p): Allow register accesses, adjust callers.
1463 (completely_scalarize): Remove function.
1464 (scalarize_elem): Likewise.
1465 (create_total_scalarization_access): Likewise.
1466 (sort_and_splice_var_accesses): Do not track total scalarization
1468 (analyze_access_subtree): New parameter totally, adjust to new meaning
1469 of grp_total_scalarization.
1470 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1471 (can_totally_scalarize_forest_p): New function.
1472 (create_total_scalarization_access): Likewise.
1473 (create_total_access_and_reshape): Likewise.
1474 (total_should_skip_creating_access): Likewise.
1475 (totally_scalarize_subtree): Likewise.
1476 (analyze_all_variable_accesses): Perform total scalarization after
1477 subaccess propagation using the new functions above.
1478 (initialize_constant_pool_replacements): Output initializers by
1479 traversing the access tree.
1481 2020-01-29 Martin Jambor <mjambor@suse.cz>
1483 * tree-sra.c (verify_sra_access_forest): New function.
1484 (verify_all_sra_access_forests): Likewise.
1485 (create_artificial_child_access): Set parent.
1486 (analyze_all_variable_accesses): Call the verifier.
1488 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1490 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1491 if called on indirect edge.
1492 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1493 speculative call if needed.
1495 2020-01-29 Richard Biener <rguenther@suse.de>
1497 PR tree-optimization/93428
1498 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1499 permutation when the load node is created.
1500 (vect_analyze_slp_instance): Re-use it here.
1502 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1504 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1506 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1508 PR rtl-optimization/93272
1509 * ira-lives.c (process_out_of_region_eh_regs): New function.
1510 (process_bb_node_lives): Call it.
1512 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1514 * coverage.c (read_counts_file): Make error message lowercase.
1516 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1518 * profile-count.c (profile_quality_display_names): Fix ordering.
1520 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1523 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1524 hash only when edge is first within the sequence.
1525 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1526 (symbol_table::create_edge): Do not set target_prob.
1527 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1529 (cgraph_edge::make_speculative): Drop target_prob parameter.
1530 (cgraph_edge::speculative_call_info): Remove.
1531 (cgraph_edge::first_speculative_call_target): New member function.
1532 (update_call_stmt_hash_for_removing_direct_edge): New function.
1533 (cgraph_edge::resolve_speculation): Rewrite to new API.
1534 (cgraph_edge::speculative_call_for_target): New member function.
1535 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1536 multiple speculation targets.
1537 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1539 (verify_speculative_call): Verify that targets form an interval.
1540 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1541 (cgraph_edge::first_speculative_call_target): New member function.
1542 (cgraph_edge::next_speculative_call_target): New member function.
1543 (cgraph_edge::speculative_call_target_ref): New member function.
1544 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1545 (cgraph_edge): Remove target_prob.
1546 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1547 Fix handling of speculative calls.
1548 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1549 * ipa-fnsummary.c (analyze_function_body): Likewise.
1550 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1551 * ipa-profile.c (dump_histogram): Fix formating.
1552 (ipa_profile_generate_summary): Watch for overflows.
1553 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1554 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1555 (update_indirect_edges_after_inlining): Update to new API.
1556 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1558 * profile-count.h: (profile_probability::adjusted): New.
1559 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1560 updating of profile.
1561 * value-prof.c (gimple_ic_transform): Rename to ...
1562 (dump_ic_profile): ... this one; update dumping.
1563 (stream_in_histogram_value): Fix formating.
1564 (gimple_value_profile_transformations): Update.
1566 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1569 * config/i386/i386.md (*movoi_internal_avx): Remove
1570 TARGET_SSE_TYPELESS_STORES check.
1571 (*movti_internal): Prefer TARGET_AVX over
1572 TARGET_SSE_TYPELESS_STORES.
1573 (*movtf_internal): Likewise.
1574 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1575 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1576 from TARGET_SSE_TYPELESS_STORES.
1578 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1580 * diagnostic-core.h (warning_at): Rename overload to...
1581 (warning_meta): ...this.
1582 (emit_diagnostic_valist): Delete decl of overload taking
1583 diagnostic_metadata.
1584 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1585 (warning_at): Rename overload taking diagnostic_metadata to...
1586 (warning_meta): ...this.
1588 2020-01-28 Richard Biener <rguenther@suse.de>
1590 PR tree-optimization/93439
1591 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1592 * tree-cfg.c (move_sese_region_to_fn): ... here.
1593 (verify_types_in_gimple_reference): Verify used cliques are
1596 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1599 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1600 argument of a pointer to struct gcc_options and pass it to
1601 parse_mtune_ctrl_str.
1602 (ix86_function_specific_restore): Pass opts to
1603 set_ix86_tune_features.
1604 (ix86_option_override_internal): Likewise.
1605 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1606 gcc_options and use it for x_ix86_tune_ctrl_string.
1608 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1610 PR rtl-optimization/87763
1611 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1612 simplification to handle subregs as well as bare regs.
1613 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1615 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1617 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1618 for reduction chains that (now) include a call.
1620 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1622 PR tree-optimization/92822
1623 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1624 out the don't-care elements of a vector whose significant elements
1625 are duplicates, make the don't-care elements duplicates too.
1627 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1629 PR tree-optimization/93434
1630 * tree-predcom.c (split_data_refs_to_components): Record which
1631 components have had aliasing loads removed. Prevent store-store
1632 commoning for all such components.
1634 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1637 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1638 -1 or is_vshift is true, use new_vector with number of elts npatterns
1639 rather than new_unary_operation.
1641 PR tree-optimization/93454
1642 * gimple-fold.c (fold_array_ctor_reference): Perform
1643 elt_size.to_uhwi () just once, instead of calling it in every
1644 iteration. Punt if that value is above size of the temporary
1645 buffer. Decrease third native_encode_expr argument when
1646 bufoff + elt_sz is above size of buf.
1648 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1650 * config/mips/mips.c (mips_declare_object_name)
1651 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1653 2020-01-27 Martin Liska <mliska@suse.cz>
1655 PR gcov-profile/93403
1656 * tree-profile.c (gimple_init_gcov_profiler): Generate
1657 both __gcov_indirect_call_profiler_v4 and
1658 __gcov_indirect_call_profiler_v4_atomic.
1660 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1663 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1665 (@aarch64_split_simd_mov<mode>): Use it.
1666 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1667 Leave the vec_extract patterns to handle 2-element vectors.
1668 (aarch64_simd_mov_from_<mode>high): Likewise.
1669 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1670 (vec_extractv2dfv1df): Likewise.
1672 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1674 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1675 jump conditions for *compare_condjump<GPI:mode>.
1677 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1680 * digraph.cc (test_edge::test_edge): Specify template for base
1683 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1685 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1687 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1689 * config/arc/arc-protos.h (gen_mlo): Remove.
1690 (gen_mhi): Likewise.
1691 * config/arc/arc.c (AUX_MULHI): Define.
1692 (arc_must_save_reister): Special handling for r58/59.
1693 (arc_compute_frame_size): Consider mlo/mhi registers.
1694 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1696 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1697 mlo/mhi name selection.
1698 (arc_restore_callee_saves): Don't early restore blink when ISR.
1699 (arc_expand_prologue): Add mlo/mhi saving.
1700 (arc_expand_epilogue): Add mlo/mhi restoring.
1703 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1704 numbering when MUL64 option is used.
1705 (DWARF2_FRAME_REG_OUT): Define.
1706 * config/arc/arc.md (arc600_stall): New pattern.
1707 (VUNSPEC_ARC_ARC600_STALL): Define.
1708 (mulsi64): Use correct mlo/mhi registers.
1709 (mulsi_600): Clean it up.
1710 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1712 (mhi_operand): Likewise.
1714 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1715 Petro Karashchenko <petro.karashchenko@ring.com>
1717 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1718 attributes if needed.
1719 (prepare_move_operands): Generate special unspec instruction for
1721 (arc_isuncached_mem_p): Propagate uncached attribute to each
1723 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1724 (VUNSPEC_ARC_STDI): Likewise.
1725 (ALLI): New mode iterator.
1726 (mALLI): New mode attribute.
1727 (lddi): New instruction pattern.
1729 (stdidi_split): Split instruction for architectures which are not
1730 supporting ll64 option.
1731 (lddidi_split): Likewise.
1733 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1735 PR rtl-optimization/92989
1736 * lra-lives.c (process_bb_lives): Update the live-in set before
1737 processing additional clobbers.
1739 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1741 PR rtl-optimization/93170
1742 * cselib.c (cselib_invalidate_regno_val): New function, split out
1744 (cselib_invalidate_regno): ...here.
1745 (cselib_invalidated_by_call_p): New function.
1746 (cselib_process_insn): Iterate over all the hard-register entries in
1747 REG_VALUES and invalidate any that cross call-clobbered registers.
1749 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1751 * dojump.c (split_comparison): Use HONOR_NANS rather than
1752 HONOR_SNANS when splitting LTGT.
1754 2020-01-27 Martin Liska <mliska@suse.cz>
1757 * opts.c (print_filtered_help): Exclude language-specific
1758 options from --help=common unless enabled in all FEs.
1760 2020-01-27 Martin Liska <mliska@suse.cz>
1762 * opts.c (print_help): Exclude params from
1763 all except --help=param.
1765 2020-01-27 Martin Liska <mliska@suse.cz>
1768 * config/i386/i386-features.c (make_resolver_func):
1769 Align the code with ppc64 target implementation.
1770 Do not generate a unique name for resolver function.
1772 2020-01-27 Richard Biener <rguenther@suse.de>
1774 PR tree-optimization/93397
1775 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1776 converted reduction chain SLP graph adjustment.
1778 2020-01-26 Marek Polacek <polacek@redhat.com>
1781 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1784 2020-01-26 Jason Merrill <jason@redhat.com>
1787 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1790 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1792 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1793 (rx_setmem): Likewise.
1795 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1798 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1799 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1800 drop <di> from constraint of last operand.
1803 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1804 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1805 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1807 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1810 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1813 2020-01-24 Jeff Law <law@redhat.com>
1815 PR tree-optimization/92788
1816 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1819 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1822 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1823 *avx_vperm_broadcast_<mode>,
1824 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1825 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1826 Move before avx2_perm<mode>/avx512f_perm<mode>.
1829 * simplify-rtx.c (simplify_const_unary_operation,
1830 simplify_const_binary_operation): Punt for mode precision above
1831 MAX_BITSIZE_MODE_ANY_INT.
1833 2020-01-24 Andrew Pinski <apinski@marvell.com>
1835 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1838 2020-01-24 Jeff Law <law@redhat.com>
1841 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1842 for REGs. Call output_operand_lossage to get more reasonable
1845 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1847 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1848 gcn_fp_compare_operator.
1849 (vec_cmpu<mode>di): Use gcn_compare_operator.
1850 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1851 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1852 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1853 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1854 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1855 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1856 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1857 gcn_fp_compare_operator.
1858 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1859 gcn_fp_compare_operator.
1860 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1861 gcn_fp_compare_operator.
1862 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1863 gcn_fp_compare_operator.
1865 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1867 * doc/install.texi (Cross-Compiler-Specific Options): Document
1868 `--with-toolexeclibdir' option.
1870 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1872 * target.def (flags_regnum): Also mention effect on delay slot filling.
1873 * doc/tm.texi: Regenerate.
1875 2020-01-23 Jeff Law <law@redhat.com>
1877 PR translation/90162
1878 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1880 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1883 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1886 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1888 PR rtl-optimization/93402
1889 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1892 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1894 * config.in: Regenerated.
1895 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1896 for TARGET_LIBC_GNUSTACK.
1897 * configure: Regenerated.
1898 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1899 found to be 2.31 or greater.
1901 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1903 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1905 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1906 (mips_asm_file_end): New function. Delegate to
1907 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1908 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1910 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1913 * config/i386/i386-modes.def (POImode): New mode.
1914 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1915 * config/i386/i386.md (DPWI): New mode attribute.
1916 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1918 (QPWI): ... this. Use POI instead of OI for TImode.
1919 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1920 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1923 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1926 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1928 (speculation_tracker_rev): New pattern.
1929 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1930 Use speculation_tracker_rev to track the inverse condition.
1932 2020-01-23 Richard Biener <rguenther@suse.de>
1934 PR tree-optimization/93381
1935 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1936 alias-set of the def as argument and record the first one.
1937 (vn_walk_cb_data::first_set): New member.
1938 (vn_reference_lookup_3): Pass the alias-set of the current def
1939 to push_partial_def. Fix alias-set used in the aggregate copy
1941 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1942 * real.c (clear_significand_below): Fix out-of-bound access.
1944 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1947 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1948 New define_insn patterns.
1950 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1952 * doc/sourcebuild.texi (check-function-bodies): Add an
1953 optional target/xfail selector.
1955 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1957 PR rtl-optimization/93124
1958 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1959 bare USE and CLOBBER insns.
1961 2020-01-22 Andrew Pinski <apinski@marvell.com>
1963 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1965 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1968 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1969 diagnostic_manager into "ana" namespace.
1970 * selftest-run-tests.c (selftest::run_tests): Update for move of
1971 selftest::run_analyzer_selftests to
1972 ana::selftest::run_analyzer_selftests.
1974 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1976 * cfgexpand.c (union_stack_vars): Update the size.
1978 2020-01-22 Richard Biener <rguenther@suse.de>
1980 PR tree-optimization/93381
1981 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1982 throughout, handle all conversions the same.
1984 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1987 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1988 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1989 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1990 Call force_reg on high_in2 unconditionally.
1992 2020-01-22 Martin Liska <mliska@suse.cz>
1994 PR tree-optimization/92924
1995 * profile.c (compute_value_histograms): Divide
1998 2020-01-22 Jakub Jelinek <jakub@redhat.com>
2001 * output.h (assemble_name_resolve): Declare.
2002 * varasm.c (assemble_name_resolve): New function.
2003 (assemble_name): Use it.
2004 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
2006 2020-01-22 Joseph Myers <joseph@codesourcery.com>
2008 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
2009 update_web_docs_git instead of update_web_docs_svn.
2011 2020-01-21 Andrew Pinski <apinski@marvell.com>
2014 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
2015 as PTR mode. Have operand 1 as being modeless, it can be P mode.
2016 (*tlsgd_small_<mode>): Likewise.
2017 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
2018 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
2019 register. Convert that register back to dest using convert_mode.
2021 2020-01-21 Jim Wilson <jimw@sifive.com>
2023 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
2026 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
2027 Uros Bizjak <ubizjak@gmail.com>
2030 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
2032 (legitimize_tls_address): Do GNU2 TLS address computation in
2033 ptr_mode and zero-extend result to Pmode.
2034 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
2035 :P with :PTR and Pmode with ptr_mode.
2036 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
2037 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
2038 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
2040 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2043 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
2044 the last two operands are CONST_INT_P before using them as such.
2046 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2048 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
2049 to get the integer element types.
2051 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
2053 * config/aarch64/aarch64-sve-builtins.h
2054 (function_expander::convert_to_pmode): Declare.
2055 * config/aarch64/aarch64-sve-builtins.cc
2056 (function_expander::convert_to_pmode): New function.
2057 (function_expander::get_contiguous_base): Use it.
2058 (function_expander::prepare_gather_address_operands): Likewise.
2059 * config/aarch64/aarch64-sve-builtins-sve2.cc
2060 (svwhilerw_svwhilewr_impl::expand): Likewise.
2062 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2065 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2066 cfun->machine->label_is_assembled.
2067 (aarch64_print_patchable_function_entry): New.
2068 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2069 * config/aarch64/aarch64.h (struct machine_function): New field,
2072 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2075 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2078 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2081 * cgraph.c (cgraph_edge::resolve_speculation,
2082 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2083 call_stmt_site_hash.
2085 2020-01-21 Martin Liska <mliska@suse.cz>
2087 * config/rs6000/rs6000.c (common_mode_defined): Remove
2090 2020-01-21 Richard Biener <rguenther@suse.de>
2092 PR tree-optimization/92328
2093 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2094 type when value-numbering same-sized store by inserting a
2096 (eliminate_dom_walker::eliminate_stmt): When eliminating
2097 a redundant store handle bit-reinterpretation of the same value.
2099 2020-01-21 Andrew Pinski <apinski@marvel.com>
2102 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2104 (prepare_block_for_update): This. Use a worklist instead of
2107 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2109 * gcc/config/arm/arm.c (clear_operation_p):
2110 Initialise last_regno, skip first iteration
2111 based on the first_set value and use ints instead
2112 of the unnecessary HOST_WIDE_INTs.
2114 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2117 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2118 compare_mode other than SFmode or DFmode.
2120 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2123 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2124 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2125 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2127 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2129 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2131 2020-01-20 Andrew Pinski <apinski@marvell.com>
2134 * targhooks.c (default_print_patchable_function_entry): Use
2135 output_asm_insn to emit the nop instruction.
2137 2020-01-20 Fangrui Song <maskray@google.com>
2140 * targhooks.c (default_print_patchable_function_entry): Align to
2143 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2146 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2147 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2148 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2149 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2150 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2151 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2152 Remove the {q} suffix from lea.
2153 (*tls_dynamic_gnu2_call_64): Renamed to ...
2154 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2155 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2156 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2157 Pass Pmode to gen_tls_dynamic_gnu2_64.
2159 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2161 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2163 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2165 * config/aarch64/aarch64-sve-builtins-base.cc
2166 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2168 2020-01-20 Richard Biener <rguenther@suse.de>
2171 * dwarf2out.c (prune_unused_types): Unconditionally mark
2172 called function DIEs.
2174 2020-01-20 Martin Liska <mliska@suse.cz>
2176 PR tree-optimization/93199
2177 * tree-eh.c (struct leh_state): Add
2178 new field outer_non_cleanup.
2179 (cleanup_is_dead_in): Pass leh_state instead
2180 of eh_region. Add a checking that state->outer_non_cleanup
2181 points to outer non-clean up region.
2182 (lower_try_finally): Record outer_non_cleanup
2184 (lower_catch): Likewise.
2185 (lower_eh_filter): Likewise.
2186 (lower_eh_must_not_throw): Likewise.
2187 (lower_cleanup): Likewise.
2189 2020-01-20 Richard Biener <rguenther@suse.de>
2191 PR tree-optimization/93094
2192 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2193 (vect_transform_loop): Likewise.
2194 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2195 loop_vectorized_call to vect_transform_loop.
2196 * tree-vect-loop.c (vect_transform_loop): Pass down
2197 loop_vectorized_call to vect_loop_versioning.
2198 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2199 the earlier discovered loop_vectorized_call.
2201 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2203 * doc/contribute.texi: Update for SVN -> Git transition.
2204 * doc/install.texi: Likewise.
2206 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2209 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2210 speculative targets.
2211 (verify_speculative_call): New function
2212 (cgraph_node::verify_node): Use it.
2213 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2216 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2219 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2220 (cgraph_edge::make_direct): Remove all indirect targets.
2221 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2222 (cgraph_node::verify_node): Verify that only one call_stmt or
2223 lto_stmt_uid is set.
2224 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2226 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2227 (lto_output_ref): Simplify streaming of stmt.
2228 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2230 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2232 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2233 Mark parameter unused.
2235 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2237 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2239 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2241 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2243 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2245 * Makefile.in: Add coroutine-passes.o.
2246 * builtin-types.def (BT_CONST_SIZE): New.
2247 (BT_FN_BOOL_PTR): New.
2248 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2249 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2250 * coroutine-builtins.def: New file.
2251 * coroutine-passes.cc: New file.
2252 * function.h (struct GTY function): Add a bit to indicate that the
2253 function is a coroutine component.
2254 * internal-fn.c (expand_CO_FRAME): New.
2255 (expand_CO_YIELD): New.
2256 (expand_CO_SUSPN): New.
2257 (expand_CO_ACTOR): New.
2258 * internal-fn.def (CO_ACTOR): New.
2262 * passes.def: Add pass_coroutine_lower_builtins,
2263 pass_coroutine_early_expand_ifns.
2264 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2265 (make_pass_coroutine_early_expand_ifns): New.
2266 * doc/invoke.texi: Document the fcoroutines command line
2269 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2271 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2274 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2275 after checking the argument is a REG. Don't use REGNO (reg)
2276 again to set last_regno, reuse regno variable instead.
2278 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2280 * doc/analyzer.texi (Limitations): Add note about NaN.
2282 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2283 Sudakshina Das <sudi.das@arm.com>
2285 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2286 and valid immediate.
2287 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2288 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2289 * config/arm/constraints.md (Pg): New.
2290 * config/arm/predicates.md (long_shift_imm): New.
2291 (arm_reg_or_long_shift_imm): Likewise.
2292 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2293 (thumb2_lsll): Likewise.
2296 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2297 Sudakshina Das <sudi.das@arm.com>
2299 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2300 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2301 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2302 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2303 * config/arm/thumb2.md (thumb2_asrl): New.
2304 (thumb2_lsll): Likewise.
2306 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2308 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2311 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2313 * gdbinit.in (help-gcc-hooks): New command.
2314 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2315 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2318 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2320 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2321 correct target macro.
2323 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2325 * config/aarch64/aarch64-protos.h
2326 (aarch64_sve_ld1ro_operand_p): New.
2327 * config/aarch64/aarch64-sve-builtins-base.cc
2328 (class load_replicate): New.
2329 (class svld1ro_impl): New.
2330 (class svld1rq_impl): Change to inherit from load_replicate.
2331 (svld1ro): New sve intrinsic function base.
2332 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2333 New DEF_SVE_FUNCTION.
2334 * config/aarch64/aarch64-sve-builtins-base.h
2335 (svld1ro): New decl.
2336 * config/aarch64/aarch64-sve-builtins.cc
2337 (function_expander::add_mem_operand): Modify assert to allow
2339 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2341 * config/aarch64/aarch64.c
2342 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2343 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2344 (aarch64_sve_ld1ro_operand_p): New.
2345 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2346 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2347 * config/aarch64/predicates.md
2348 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2350 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2352 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2353 Introduce this ACLE specified predefined macro.
2354 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2355 (fp): Disabling this disables f64mm.
2356 (simd): Disabling this disables f64mm.
2357 (fp16): Disabling this disables f64mm.
2358 (sve): Disabling this disables f64mm.
2359 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2360 (AARCH64_ISA_F64MM): New.
2361 (TARGET_F64MM): New.
2362 * doc/invoke.texi (f64mm): Document new option.
2364 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2366 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2367 (neoversen1_tunings): Likewise.
2369 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2372 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2373 Add assert to ensure prolog has been emitted.
2374 (aarch64_split_atomic_op): Likewise.
2375 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2376 Use epilogue_completed rather than reload_completed.
2377 (aarch64_atomic_exchange<mode>): Likewise.
2378 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2379 (atomic_nand<mode>): Likewise.
2380 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2381 (atomic_fetch_nand<mode>): Likewise.
2382 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2383 (atomic_nand_fetch<mode>): Likewise.
2385 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2388 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2390 (REVERSE_CONDITION): Delete.
2391 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2392 (CCFP_CCFPE): Likewise.
2393 (e): New mode attribute.
2394 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2395 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2396 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2397 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2398 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2399 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2400 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2401 name of generator from gen_ccmpdi to gen_ccmpccdi.
2402 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2403 the previous comparison but aren't able to, use the new ccmp_rev
2406 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2408 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2409 than testing directly for INTEGER_CST.
2410 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2412 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2414 PR tree-optimization/93292
2415 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2416 get_vectype_for_scalar_type returns NULL.
2418 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2420 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2421 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2423 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2425 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2427 * params.opt: (max-predicted-iterations): Set bounds.
2428 * predict.c (real_almost_one, real_br_prob_base,
2429 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2430 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2431 probabilities; do not truncate to reg_br_prob_bases.
2432 (estimate_loops_at_level): Pass max_cyclic_prob.
2433 (estimate_loops): Compute max_cyclic_prob.
2434 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2436 * profile-count.c (profile_probability::to_sreal): New.
2437 * profile-count.h (class sreal): Move up in file.
2438 (profile_probability::to_sreal): Declare.
2440 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2443 (arm_invalid_conversion): New function for target hook.
2444 (arm_invalid_unary_op): New function for target hook.
2445 (arm_invalid_binary_op): New function for target hook.
2447 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2449 * config.gcc: Add arm_bf16.h.
2450 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2451 (arm_simd_builtin_std_type): Add BFmode.
2452 (arm_init_simd_builtin_types): Define element types for vector types.
2453 (arm_init_bf16_types): New function.
2454 (arm_init_builtins): Add arm_init_bf16_types function call.
2455 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2456 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2457 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2458 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2459 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2460 (arm_mangle_type): Add __bf16.
2461 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2462 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2463 arm_bf16_ptr_type_node.
2464 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2465 define_split between ARM registers.
2466 * config/arm/arm_bf16.h: New file.
2467 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2468 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2470 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2471 * config/arm/vfp.md: Add BFmode to movhf patterns.
2473 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2474 Andre Vieira <andre.simoesdiasvieira@arm.com>
2476 * config/arm/arm-cpus.in (mve, mve_float): New features.
2477 (dsp, mve, mve.fp): New options.
2478 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2479 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2480 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2482 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2483 Thomas Preud'homme <thomas.preudhomme@arm.com>
2485 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2487 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2488 error for using -mcmse when targeting Armv8.1-M Mainline.
2490 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2491 Thomas Preud'homme <thomas.preudhomme@arm.com>
2493 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2494 address in r4 when targeting Armv8.1-M Mainline.
2495 (nonsecure_call_value_internal): Likewise.
2496 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2497 a register match_operand again. Emit BLXNS when targeting
2499 (nonsecure_call_value_reg_thumb2): Likewise.
2501 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2502 Thomas Preud'homme <thomas.preudhomme@arm.com>
2504 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2505 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2506 variable as true when floating-point ABI is not hard. Replace
2507 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2508 Generate VLSTM and VLLDM instruction respectively before and
2509 after a function call to cmse_nonsecure_call function.
2510 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2511 (VUNSPEC_VLLDM): Likewise.
2512 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2513 (lazy_load_multiple_insn): Likewise.
2515 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2516 Thomas Preud'homme <thomas.preudhomme@arm.com>
2518 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2519 (arm_emit_vfp_multi_reg_pop): Likewise.
2520 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2521 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2522 restore callee-saved VFP registers.
2524 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2525 Thomas Preud'homme <thomas.preudhomme@arm.com>
2527 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2528 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2529 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2530 callee-saved GPRs as well as clear ip register before doing a nonsecure
2531 call then restore callee-saved GPRs after it when targeting
2533 (arm_reorg): Adapt to function rename.
2535 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2536 Thomas Preud'homme <thomas.preudhomme@arm.com>
2538 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2539 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2540 clear_vfp_multiple pattern based on a new vfp parameter.
2541 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2542 targeting Armv8.1-M Mainline.
2543 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2544 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2545 whether VFP registers are available before looking call_used_regs for a
2547 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2548 of prototype of clear_operation_p.
2549 (clear_vfp_multiple_operation): New predicate.
2550 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2551 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2553 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2554 Thomas Preud'homme <thomas.preudhomme@arm.com>
2556 * config/arm/arm-protos.h (clear_operation_p): Declare.
2557 * config/arm/arm.c (clear_operation_p): New function.
2558 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2559 targeting Armv8.1-M Mainline or successor.
2560 (output_return_instruction): Only output APSR register clearing if
2561 Armv8.1-M Mainline instructions not available.
2562 (thumb_exit): Likewise.
2563 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2564 * config/arm/thumb2.md (clear_apsr): New define_insn.
2565 (clear_multiple): Likewise.
2566 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2568 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2569 Thomas Preud'homme <thomas.preudhomme@arm.com>
2571 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2572 (use_return_insn): Also return false for Armv8.1-M Mainline.
2573 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2574 Mainline instructions are available.
2575 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2576 when targeting Armv8.1-M Mainline Security Extensions.
2577 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2578 Mainline entry function.
2579 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2580 targeting Armv8.1-M Mainline or successor.
2581 (arm_expand_epilogue): Fix indentation of caller-saved register
2582 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2584 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2585 (FP_SYSREGS): Likewise.
2586 (enum vfp_sysregs_encoding): Define enum.
2587 (fp_sysreg_names): Declare.
2588 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2589 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2590 (pop_fpsysreg_insn): Likewise.
2592 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2593 Thomas Preud'homme <thomas.preudhomme@arm.com>
2595 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2596 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2597 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2598 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2599 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2600 (ARMv8_1m_main): New feature group.
2601 (armv8.1-m.main): New architecture.
2602 * config/arm/arm-tables.opt: Regenerate.
2603 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2604 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2605 (arm_options_perform_arch_sanity_checks): Error out when targeting
2606 Armv8.1-M Mainline Security Extensions.
2607 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2609 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2611 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2612 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2613 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2614 aarch64_bfdot_laneq): New.
2615 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2616 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2617 vbfdotq_laneq_f32): New.
2618 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2619 VBFMLA_W, VBF): New.
2620 (isquadop): Add V4BF, V8BF.
2622 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2624 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2625 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2626 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2627 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2628 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2629 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2630 usdot_laneq, sudot_lane,sudot_laneq): New.
2631 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2632 (aarch64_<sur>dot_lane): New.
2633 * config/aarch64/arm_neon.h (vusdot_s32): New.
2635 (vusdot_lane_s32): New.
2636 (vsudot_lane_s32): New.
2637 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2638 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2640 2020-01-16 Martin Liska <mliska@suse.cz>
2642 * value-prof.c (dump_histogram_value): Fix
2643 obvious spacing issue.
2645 2020-01-16 Andrew Pinski <apinski@marvell.com>
2647 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2648 !storage_order_barrier_p.
2650 2020-01-16 Andrew Pinski <apinski@marvell.com>
2652 * sched-int.h (_dep): Add unused bit-field field for the padding.
2653 * sched-deps.c (init_dep_1): Init unused field.
2655 2020-01-16 Andrew Pinski <apinski@marvell.com>
2657 * optabs.h (create_expand_operand): Initialize target field also.
2659 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2661 PR tree-optimization/92429
2662 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2663 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2665 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2668 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2670 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2671 aarch64_sve_int_mode to each mode.
2673 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2675 * doc/analyzer.texi (Overview): Add note about
2676 -fdump-ipa-analyzer.
2678 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2680 PR tree-optimization/93231
2681 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2682 input_type is unsigned. Use tree_to_shwi for shift constant.
2683 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2684 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2685 inputs more efficiently.
2687 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2689 * config/i386/i386.md (*movsf_internal): Do not require
2690 SSE2 ISA for alternatives 14 and 15.
2692 2020-01-15 Richard Biener <rguenther@suse.de>
2695 * tree-eh.c (sink_clobbers): If we already visited the destination
2696 block do not defer insertion.
2697 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2698 the purpose of defered insertion.
2700 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2702 * BASE-VER: Bump to 10.0.1.
2704 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2706 PR tree-optimization/93247
2707 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2708 type of the stmt that we're going to vectorize.
2710 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2712 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2713 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2716 2020-01-15 Martin Liska <mliska@suse.cz>
2718 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2719 2 calls of streamer_read_hwi in a function call.
2721 2020-01-15 Richard Biener <rguenther@suse.de>
2723 * alias.c (record_alias_subset): Avoid redundant work when
2724 subset is already recorded.
2726 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2728 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2729 the analyzer options provide CWE identifiers.
2731 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2733 * tree-diagnostic-path.cc (path_summary::event_range::print):
2734 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2735 using get_pure_location.
2737 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2739 PR tree-optimization/93262
2740 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2741 perform head trimming only if the last argument is constant,
2742 either all ones, or larger or equal to head trim, in the latter
2743 case decrease the last argument by head_trim.
2745 PR tree-optimization/93249
2746 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2747 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2748 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2749 perform head trim unless we can prove there are no '\0' chars
2750 from the source among the first head_trim chars.
2752 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2754 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2756 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2759 * config/i386/sse.md
2760 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2761 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2762 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2763 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2764 just a single alternative instead of two, make operands 1 and 2
2767 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2770 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2773 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2775 * Makefile.in (lang_opt_files): Add analyzer.opt.
2776 (ANALYZER_OBJS): New.
2777 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2778 tristate.o and ANALYZER_OBJS.
2779 (TEXI_GCCINT_FILES): Add analyzer.texi.
2780 * common.opt (-fanalyzer): New driver option.
2781 * config.in: Regenerate.
2782 * configure: Regenerate.
2783 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2784 (gccdepdir): Also create depdir for "analyzer" subdir.
2785 * digraph.cc: New file.
2786 * digraph.h: New file.
2787 * doc/analyzer.texi: New file.
2788 * doc/gccint.texi ("Static Analyzer") New menu item.
2789 (analyzer.texi): Include it.
2790 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2791 ("Warning Options"): Add static analysis warnings to the list.
2792 (-Wno-analyzer-double-fclose): New option.
2793 (-Wno-analyzer-double-free): New option.
2794 (-Wno-analyzer-exposure-through-output-file): New option.
2795 (-Wno-analyzer-file-leak): New option.
2796 (-Wno-analyzer-free-of-non-heap): New option.
2797 (-Wno-analyzer-malloc-leak): New option.
2798 (-Wno-analyzer-possible-null-argument): New option.
2799 (-Wno-analyzer-possible-null-dereference): New option.
2800 (-Wno-analyzer-null-argument): New option.
2801 (-Wno-analyzer-null-dereference): New option.
2802 (-Wno-analyzer-stale-setjmp-buffer): New option.
2803 (-Wno-analyzer-tainted-array-index): New option.
2804 (-Wno-analyzer-use-after-free): New option.
2805 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2806 (-Wno-analyzer-use-of-uninitialized-value): New option.
2807 (-Wanalyzer-too-complex): New option.
2808 (-fanalyzer-call-summaries): New warning.
2809 (-fanalyzer-checker=): New warning.
2810 (-fanalyzer-fine-grained): New warning.
2811 (-fno-analyzer-state-merge): New warning.
2812 (-fno-analyzer-state-purge): New warning.
2813 (-fanalyzer-transitivity): New warning.
2814 (-fanalyzer-verbose-edges): New warning.
2815 (-fanalyzer-verbose-state-changes): New warning.
2816 (-fanalyzer-verbosity=): New warning.
2817 (-fdump-analyzer): New warning.
2818 (-fdump-analyzer-callgraph): New warning.
2819 (-fdump-analyzer-exploded-graph): New warning.
2820 (-fdump-analyzer-exploded-nodes): New warning.
2821 (-fdump-analyzer-exploded-nodes-2): New warning.
2822 (-fdump-analyzer-exploded-nodes-3): New warning.
2823 (-fdump-analyzer-supergraph): New warning.
2824 * doc/sourcebuild.texi (dg-require-dot): New.
2825 (dg-check-dot): New.
2826 * gdbinit.in (break-on-saved-diagnostic): New command.
2827 * graphviz.cc: New file.
2828 * graphviz.h: New file.
2829 * ordered-hash-map-tests.cc: New file.
2830 * ordered-hash-map.h: New file.
2831 * passes.def (pass_analyzer): Add before
2832 pass_ipa_whole_program_visibility.
2833 * selftest-run-tests.c (selftest::run_tests): Call
2834 selftest::ordered_hash_map_tests_cc_tests.
2835 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2837 * shortest-paths.h: New file.
2838 * timevar.def (TV_ANALYZER): New timevar.
2839 (TV_ANALYZER_SUPERGRAPH): Likewise.
2840 (TV_ANALYZER_STATE_PURGE): Likewise.
2841 (TV_ANALYZER_PLAN): Likewise.
2842 (TV_ANALYZER_SCC): Likewise.
2843 (TV_ANALYZER_WORKLIST): Likewise.
2844 (TV_ANALYZER_DUMP): Likewise.
2845 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2846 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2847 * tree-pass.h (make_pass_analyzer): New decl.
2848 * tristate.cc: New file.
2849 * tristate.h: New file.
2851 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2854 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2855 alternatives 9 and 10.
2857 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2859 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2860 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2861 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2862 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2863 (selftest::hash_map_tests_c_tests): Call it.
2864 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2865 New static constant, using the value of = H::empty_zero_p.
2866 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2867 from default_hash_traits <Value>.
2868 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2870 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2871 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2872 calls to mark_empty with !Descriptor::empty_zero_p.
2873 (hash_table::empty_slow): Conditionalize the memset call with a
2874 check that Descriptor::empty_zero_p; otherwise, loop through the
2875 entries calling mark_empty on them.
2876 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2877 (pointer_hash::empty_zero_p): Likewise.
2878 (pair_hash::empty_zero_p): Likewise.
2879 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2881 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2882 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2883 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2884 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2885 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2886 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2887 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2888 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2890 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2893 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2895 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2896 fix typo on return value.
2898 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2901 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2903 (cgraph_edge::make_speculative): Add param for setting speculative_id
2905 (cgraph_edge::speculative_call_info): Update comments and find reference
2906 by speculative_id for multiple indirect targets.
2907 (cgraph_edge::resolve_speculation): Decrease the speculations
2908 for indirect edge, drop it's speculative if not direct target
2909 left. Update comments.
2910 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2911 (cgraph_node::dump): Print num_speculative_call_targets.
2912 (cgraph_node::verify_node): Don't report error if speculative
2913 edge not include statement.
2914 (cgraph_edge::num_speculative_call_targets_p): New function.
2915 * cgraph.h (int common_target_id): Remove.
2916 (int common_target_probability): Remove.
2917 (num_speculative_call_targets): New variable.
2918 (make_speculative): Add param for setting speculative_id.
2919 (cgraph_edge::num_speculative_call_targets_p): New declare.
2920 (target_prob): New variable.
2921 (speculative_id): New variable.
2922 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2923 call summaries for multiple speculative call targets.
2924 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2925 * ipa-profile.c (struct speculative_call_target): New struct.
2926 (class speculative_call_summary): New class.
2927 (class speculative_call_summaries): New class.
2928 (call_sums): New variable.
2929 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2930 (ipa_profile_write_edge_summary): New function.
2931 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2932 (ipa_profile_dump_all_summaries): New function.
2933 (ipa_profile_read_edge_summary): New function.
2934 (ipa_profile_read_summary_section): New function.
2935 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2936 (ipa_profile): Generate num_speculative_call_targets from
2938 * ipa-ref.h (speculative_id): New variable.
2939 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2940 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2941 common_target_probability. Stream out speculative_id and
2942 num_speculative_call_targets.
2943 (input_edge): Likewise.
2944 * predict.c (dump_prediction): Remove edges count assert to be
2946 * symtab.c (symtab_node::create_reference): Init speculative_id.
2947 (symtab_node::clone_references): Clone speculative_id.
2948 (symtab_node::clone_referring): Clone speculative_id.
2949 (symtab_node::clone_reference): Clone speculative_id.
2950 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2951 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2952 if indirect call contains multiple speculative targets.
2953 * value-prof.h (check_ic_target): Remove.
2954 * value-prof.c (gimple_value_profile_transformations):
2955 Use void function gimple_ic_transform.
2956 * value-prof.c (gimple_ic_transform): Handle topn case.
2957 Fix comment typos. Change it to a void function.
2959 2020-01-13 Andrew Pinski <apinski@marvell.com>
2961 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2962 (octeontx2t98): New define.
2963 (octeontx2t96): New define.
2964 (octeontx2t93): New define.
2965 (octeontx2f95): New define.
2966 (octeontx2f95n): New define.
2967 (octeontx2f95mm): New define.
2968 * config/aarch64/aarch64-tune.md: Regenerate.
2969 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2971 2020-01-13 Jason Merrill <jason@redhat.com>
2973 PR c++/33799 - destroy return value if local cleanup throws.
2974 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2976 2020-01-13 Martin Liska <mliska@suse.cz>
2978 * ipa-cp.c (get_max_overall_size): Use newly
2979 renamed param param_ipa_cp_unit_growth.
2980 * params.opt: Remove legacy param name.
2982 2020-01-13 Martin Sebor <msebor@redhat.com>
2984 PR tree-optimization/93213
2985 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2986 stores to be eliminated.
2988 2020-01-13 Martin Liska <mliska@suse.cz>
2990 * opts.c (print_help): Do not print CL_PARAM
2991 and CL_WARNING for CL_OPTIMIZATION.
2993 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2996 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2997 depending on optimization settings.
2999 2020-01-13 Jakub Jelinek <jakub@redhat.com>
3001 PR tree-optimization/90838
3002 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3003 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
3004 argument rather than to initialize temporary for targets that
3005 don't use the mode argument at all. Initialize ctzval to avoid
3008 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
3010 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
3011 * tree-core.h: Document it.
3012 * gimplify.c (gimplify_omp_workshare): Set it.
3013 * omp-low.c (lower_omp_target): Use it.
3014 * tree-pretty-print.c (dump_omp_clause): Print it.
3016 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
3017 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
3019 2020-01-10 David Malcolm <dmalcolm@redhat.com>
3021 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
3022 * common.opt (fdiagnostics-path-format=): New option.
3023 (diagnostic_path_format): New enum.
3024 (fdiagnostics-show-path-depths): New option.
3025 * coretypes.h (diagnostic_event_id_t): New forward decl.
3026 * diagnostic-color.c (color_dict): Add "path".
3027 * diagnostic-event-id.h: New file.
3028 * diagnostic-format-json.cc (json_from_expanded_location): Make
3030 (json_end_diagnostic): Call context->make_json_for_path if it
3031 exists and the diagnostic has a path.
3032 (diagnostic_output_format_init): Clear context->print_path.
3033 * diagnostic-path.h: New file.
3034 * diagnostic-show-locus.c (colorizer::set_range): Special-case
3035 when printing a run of events in a diagnostic_path so that they
3036 all get the same color.
3037 (layout::m_diagnostic_path_p): New field.
3038 (layout::layout): Initialize it.
3039 (layout::print_any_labels): Don't colorize the label text for an
3040 event in a diagnostic_path.
3041 (gcc_rich_location::add_location_if_nearby): Add
3042 "restrict_to_current_line_spans" and "label" params. Pass the
3043 former to layout.maybe_add_location_range; pass the latter
3044 when calling add_range.
3045 * diagnostic.c: Include "diagnostic-path.h".
3046 (diagnostic_initialize): Initialize context->path_format and
3047 context->show_path_depths.
3048 (diagnostic_show_any_path): New function.
3049 (diagnostic_path::interprocedural_p): New function.
3050 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
3051 (simple_diagnostic_path::num_events): New function.
3052 (simple_diagnostic_path::get_event): New function.
3053 (simple_diagnostic_path::add_event): New function.
3054 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
3055 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3056 (debug): New overload taking a diagnostic_path *.
3057 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3058 * diagnostic.h (enum diagnostic_path_format): New enum.
3059 (json::value): New forward decl.
3060 (diagnostic_context::path_format): New field.
3061 (diagnostic_context::show_path_depths): New field.
3062 (diagnostic_context::print_path): New callback field.
3063 (diagnostic_context::make_json_for_path): New callback field.
3064 (diagnostic_show_any_path): New decl.
3065 (json_from_expanded_location): New decl.
3066 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3067 (-fdiagnostics-show-path-depths): New option.
3068 (-fdiagnostics-color): Add "path" to description of default
3069 GCC_COLORS; describe it.
3070 (-fdiagnostics-format=json): Document how diagnostic paths are
3071 represented in the JSON output format.
3072 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3073 Add optional params "restrict_to_current_line_spans" and "label".
3074 * opts.c (common_handle_option): Handle
3075 OPT_fdiagnostics_path_format_ and
3076 OPT_fdiagnostics_show_path_depths.
3077 * pretty-print.c: Include "diagnostic-event-id.h".
3078 (pp_format): Implement "%@" format code for printing
3079 diagnostic_event_id_t *.
3080 (selftest::test_pp_format): Add tests for "%@".
3081 * selftest-run-tests.c (selftest::run_tests): Call
3082 selftest::tree_diagnostic_path_cc_tests.
3083 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3084 * toplev.c (general_init): Initialize global_dc->path_format and
3085 global_dc->show_path_depths.
3086 * tree-diagnostic-path.cc: New file.
3087 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3088 non-static. Drop "diagnostic" param in favor of storing the
3089 original value of "where" and re-using it.
3090 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3091 maybe_unwind_expanded_macro_loc.
3092 (tree_diagnostics_defaults): Initialize context->print_path and
3093 context->make_json_for_path.
3094 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3096 (default_tree_make_json_for_path): New decl.
3097 (maybe_unwind_expanded_macro_loc): New decl.
3099 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3101 PR tree-optimization/93210
3102 * fold-const.h (native_encode_initializer,
3103 can_native_interpret_type_p): Declare.
3104 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3106 (native_encode_initializer): New function, moved from dwarf2out.c.
3107 Adjust to native_encode_expr compatible arguments, including dry-run
3108 and partial extraction modes. Don't handle STRING_CST.
3109 (can_native_interpret_type_p): No longer static.
3110 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3111 offset / BITS_PER_UNIT fits into int and don't call it if
3112 can_native_interpret_type_p fails. If suboff is NULL and for
3113 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3114 native_encode_initializer.
3115 (fold_const_aggregate_ref_1): Formatting fix.
3116 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3117 (tree_add_const_value_attribute): Adjust caller.
3119 PR tree-optimization/90838
3120 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3121 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3122 CTZ_DEFINED_VALUE_AT_ZERO.
3124 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3127 * lra-constraints.c (match_reload): Permit input operands have the
3128 same mode as output while other input operands have a different
3131 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3133 PR tree-optimization/90838
3134 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3135 (check_ctz_string): Likewise.
3136 (optimize_count_trailing_zeroes): Likewise.
3137 (simplify_count_trailing_zeroes): Likewise.
3138 (pass_forwprop::execute): Try ctz simplification.
3139 * match.pd: Add matching for ctz idioms.
3141 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3143 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3145 (aarch64_invalid_unary_op): New function for target hook.
3146 (aarch64_invalid_binary_op): New function for target hook.
3148 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3150 * config.gcc: Add arm_bf16.h.
3151 * config/aarch64/aarch64-builtins.c
3152 (aarch64_simd_builtin_std_type): Add BFmode.
3153 (aarch64_init_simd_builtin_types): Define element types for vector
3155 (aarch64_init_bf16_types): New function.
3156 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3157 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3159 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3160 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3162 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3163 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3164 * config/aarch64/aarch64.c
3165 (aarch64_classify_vector_mode): Add support for BF types.
3166 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3167 (aarch64_vq_mode): Add support for BF types.
3168 (aarch64_simd_container_mode): Add support for BF types.
3169 (aarch64_mangle_type): Add support for BF scalar type.
3170 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3171 * config/aarch64/arm_bf16.h: New file.
3172 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3173 * config/aarch64/iterators.md: Add BF types to mode attributes.
3174 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3176 2020-01-10 Jason Merrill <jason@redhat.com>
3178 PR c++/93173 - incorrect tree sharing.
3179 * gimplify.c (copy_if_shared): No longer static.
3180 * gimplify.h: Declare it.
3182 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3184 * doc/invoke.texi (-msve-vector-bits=): Document that
3185 -msve-vector-bits=128 now generates VL-specific code for
3186 little-endian targets.
3187 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3188 build_vector_type_for_mode to construct the data vector types.
3189 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3190 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3191 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3192 for 128-bit vectors.
3194 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3196 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3199 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3201 * config/aarch64/aarch64-builtins.c
3202 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3203 rather than checking the number of elements and the element mode.
3205 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3207 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3208 get_related_vectype_for_scalar_type rather than build_vector_type
3209 to create the index type for a conditional reduction.
3211 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3213 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3214 for any type of gather or scatter, including strided accesses.
3216 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3218 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3221 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3223 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3225 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3226 parameter and its use to reset DR_OFFSET's.
3227 (vect_transform_loop): Remove orig_drs_init argument.
3228 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3229 member of dr_vec_info rather than the offset of the associated
3230 data_reference's innermost_loop_behavior.
3231 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3232 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3233 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3234 get_dr_vinfo_offset.
3235 (vectorizable_store): Likewise.
3236 (vectorizable_load): Likewise.
3238 2020-01-10 Richard Biener <rguenther@suse.de>
3240 * gimple-ssa-store-merging
3241 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3243 2020-01-10 Martin Liska <mliska@suse.cz>
3246 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3247 encapsulation that was there before r280040.
3249 2020-01-10 Richard Biener <rguenther@suse.de>
3252 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3253 sequences to avoid walking them again for secondary opportunities.
3254 (pass_lower_eh_dispatch::execute): Instead actually insert
3257 2020-01-10 Richard Biener <rguenther@suse.de>
3260 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3261 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3262 avoid quadraticness.
3264 2020-01-10 Martin Jambor <mjambor@suse.cz>
3266 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3267 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3268 to get param_ipa_sra_max_replacements.
3269 (param_splitting_across_edge): Pass the caller to
3270 pull_accesses_from_callee.
3272 2020-01-10 Martin Jambor <mjambor@suse.cz>
3274 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3275 * ipa-cp.c (max_new_size): Removed.
3276 (orig_overall_size): New variable.
3277 (get_max_overall_size): New function.
3278 (estimate_local_effects): Use it. Adjust dump.
3279 (decide_about_value): Likewise.
3280 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3281 orig_overall_size. Adjust dump.
3282 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3284 2020-01-10 Martin Jambor <mjambor@suse.cz>
3286 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3287 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3288 instead of param_ipa_max_agg_items.
3289 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3290 optimization info for the callee.
3292 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3294 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3295 markers if debug_inline_points is false.
3297 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3299 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3301 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3302 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3303 aarch64-sve-builtins-sve2.h.
3304 (aarch64-sve-builtins-sve2.o): New rule.
3305 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3306 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3307 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3308 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3309 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3310 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3312 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3313 instructions that are handled here.
3314 (@cond_asrd<mode>): Generalize to...
3315 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3316 (*cond_asrd<mode>_2): Generalize to...
3317 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3318 (*cond_asrd<mode>_z): Generalize to...
3319 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3320 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3321 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3322 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3323 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3325 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3326 (@aarch64_scatter_stnt<mode>): Likewise.
3327 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3328 (@aarch64_mul_lane_<mode>): Likewise.
3329 (@aarch64_sve_suqadd<mode>_const): Likewise.
3330 (*<sur>h<addsub><mode>): Generalize to...
3331 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3333 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3334 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3335 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3336 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3337 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3338 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3339 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3340 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3341 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3342 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3343 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3344 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3345 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3346 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3347 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3348 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3349 (@aarch64_sve2_xar<mode>): Likewise.
3350 (@aarch64_sve2_bcax<mode>): Likewise.
3351 (*aarch64_sve2_eor3<mode>): Rename to...
3352 (@aarch64_sve2_eor3<mode>): ...this.
3353 (@aarch64_sve2_bsl<mode>): New expander.
3354 (@aarch64_sve2_nbsl<mode>): Likewise.
3355 (@aarch64_sve2_bsl1n<mode>): Likewise.
3356 (@aarch64_sve2_bsl2n<mode>): Likewise.
3357 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3358 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3359 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3360 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3361 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3362 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3363 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3364 (<su>mull<bt><Vwide>): Generalize to...
3365 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3367 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3368 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3369 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3370 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3371 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3372 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3373 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3374 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3375 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3376 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3377 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3378 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3379 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3380 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3381 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3382 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3383 (<SHRNB:r>shrnb<mode>): Generalize to...
3384 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3386 (<SHRNT:r>shrnt<mode>): Generalize to...
3387 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3389 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3390 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3391 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3392 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3393 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3394 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3395 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3396 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3397 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3398 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3399 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3400 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3401 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3402 (@aarch64_sve2_cvtnt<mode>): Likewise.
3403 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3404 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3405 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3406 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3407 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3408 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3409 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3410 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3411 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3412 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3413 (@aarch64_sve2_pmul<mode>): Likewise.
3414 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3415 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3416 (@aarch64_sve2_tbl2<mode>): Likewise.
3417 (@aarch64_sve2_tbx<mode>): Likewise.
3418 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3419 (@aarch64_sve2_histcnt<mode>): Likewise.
3420 (@aarch64_sve2_histseg<mode>): Likewise.
3421 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3422 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3423 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3424 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3425 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3426 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3427 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3428 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3429 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3430 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3431 (SVE2_PMULL_PAIR_I): New mode iterators.
3432 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3433 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3434 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3435 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3436 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3437 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3438 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3439 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3440 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3441 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3442 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3443 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3444 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3445 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3446 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3447 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3448 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3449 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3450 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3451 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3452 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3453 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3454 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3455 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3456 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3457 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3458 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3459 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3460 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3461 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3462 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3464 (VNARROW, Ventype): New mode attributes.
3465 (Vewtype): Handle VNx2DI. Fix typo in comment.
3466 (VDOUBLE): New mode attribute.
3467 (sve_lane_con): Handle VNx8HI.
3468 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3469 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3470 (sve_int_op, sve_int_op_rev): Handle the above codes.
3471 (sve_pred_int_rhs2_operand): Likewise.
3472 (MULLBT, SHRNB, SHRNT): Delete.
3473 (SVE_INT_SHIFT_IMM): New int iterator.
3474 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3475 and UNSPEC_WHILEHS for TARGET_SVE2.
3476 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3477 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3478 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3479 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3480 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3481 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3482 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3483 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3484 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3485 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3486 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3487 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3488 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3489 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3490 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3491 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3492 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3493 (optab): Handle the new unspecs.
3494 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3496 (lr): Handle the new unspecs.
3498 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3499 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3500 (sve_int_qsub_op): New int attributes.
3501 (sve_fp_op, rot): Handle the new unspecs.
3502 * config/aarch64/aarch64-sve-builtins.h
3503 (function_resolver::require_matching_pointer_type): Declare.
3504 (function_resolver::resolve_unary): Add an optional boolean argument.
3505 (function_resolver::finish_opt_n_resolution): Add an optional
3506 type_suffix_index argument.
3507 (gimple_folder::redirect_call): Declare.
3508 (gimple_expander::prepare_gather_address_operands): Add an optional
3510 * config/aarch64/aarch64-sve-builtins.cc: Include
3511 aarch64-sve-builtins-sve2.h.
3512 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3513 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3514 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3515 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3516 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3517 (TYPES_s_unsigned): Likewise.
3518 (TYPES_s_integer): Use TYPES_s_unsigned.
3519 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3520 (TYPES_sd_integer): Use them.
3521 (TYPES_d_unsigned): New macro.
3522 (TYPES_d_integer): Use it.
3523 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3524 (TYPES_cvt_narrow): Likewise.
3525 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3526 (preds_mx): New variable.
3527 (function_builder::add_overloaded_function): Allow the new feature
3528 set to be more restrictive than the original one.
3529 (function_resolver::infer_pointer_type): Remove qualifiers from
3530 the pointer type before printing it.
3531 (function_resolver::require_matching_pointer_type): New function.
3532 (function_resolver::resolve_sv_displacement): Handle functions
3533 that don't support 32-bit vector indices or svint32_t vector offsets.
3534 (function_resolver::finish_opt_n_resolution): Take the inferred type
3535 as a separate argument.
3536 (function_resolver::resolve_unary): Optionally treat all forms in
3537 the same way as normal merging functions.
3538 (gimple_folder::redirect_call): New function.
3539 (function_expander::prepare_gather_address_operands): Add an argument
3540 that says whether scaled forms are available. If they aren't,
3541 handle scaling of vector indices and don't add the extension and
3543 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3544 fall back to using cond_* instead.
3545 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3546 Split out the member variables into...
3547 (rtx_code_function_base): ...this new base class.
3548 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3549 (unspec_based_function): Split out the member variables into...
3550 (unspec_based_function_base): ...this new base class.
3551 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3552 (unspec_based_function_exact_insn): New class.
3553 (unspec_based_add_function, unspec_based_add_lane_function)
3554 (unspec_based_lane_function, unspec_based_pred_function)
3555 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3556 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3557 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3559 (unspec_based_fused_function): New class.
3560 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3561 (unspec_based_fused_lane_function): New class.
3562 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3564 (CODE_FOR_MODE1): New macro.
3565 (fixed_insn_function): New class.
3566 (while_comparison): Likewise.
3567 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3568 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3569 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3570 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3571 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3572 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3573 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3574 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3575 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3576 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3577 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3578 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3579 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3580 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3581 Also add an initial argument for unary_convert_narrowt, regardless
3582 of the predication type.
3583 (build_32_64): Allow loads and stores to specify MODE_none.
3584 (build_sv_index64, build_sv_uint_offset): New functions.
3585 (long_type_suffix): New function.
3586 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3587 (binary_imm_long_base, load_gather_sv_base): Likewise.
3588 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3589 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3590 (unary_narrowb_base, unary_narrowt_base): Likewise.
3591 (binary_long_lane_def, binary_long_lane): New shape.
3592 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3593 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3594 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3595 (binary_to_uint_def, binary_to_uint): Likewise.
3596 (binary_wide_def, binary_wide): Likewise.
3597 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3598 (compare_def, compare): Likewise.
3599 (compare_ptr_def, compare_ptr): Likewise.
3600 (load_ext_gather_index_restricted_def,
3601 load_ext_gather_index_restricted): Likewise.
3602 (load_ext_gather_offset_restricted_def,
3603 load_ext_gather_offset_restricted): Likewise.
3604 (load_gather_sv_def): Inherit from load_gather_sv_base.
3605 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3606 (shift_left_imm_def, shift_left_imm): Likewise.
3607 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3608 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3609 (store_scatter_index_restricted_def,
3610 store_scatter_index_restricted): Likewise.
3611 (store_scatter_offset_restricted_def,
3612 store_scatter_offset_restricted): Likewise.
3613 (tbl_tuple_def, tbl_tuple): Likewise.
3614 (ternary_long_lane_def, ternary_long_lane): Likewise.
3615 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3616 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3617 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3618 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3619 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3620 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3621 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3622 (ternary_uint_def, ternary_uint): Likewise.
3623 (unary_convert): Fix typo in comment.
3624 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3625 (unary_long_def, unary_long): Likewise.
3626 (unary_narrowb_def, unary_narrowb): Likewise.
3627 (unary_narrowt_def, unary_narrowt): Likewise.
3628 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3629 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3630 (unary_to_int_def, unary_to_int): Likewise.
3631 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3632 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3633 (svasrd_impl): Delete.
3634 (svcadd_impl::expand): Handle integer operations too.
3635 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3636 new functions to derive the unspec numbers.
3637 (svmla_svmls_lane_impl): Replace with...
3638 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3639 integer operations too.
3640 (svwhile_impl): Rename to...
3641 (svwhilelx_impl): ...this and inherit from while_comparison.
3642 (svasrd): Use unspec_based_function.
3643 (svmla_lane): Use svmla_lane_impl.
3644 (svmls_lane): Use svmls_lane_impl.
3645 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3646 (svwhilele, svwhilelt): Use svwhilelx_impl.
3647 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3648 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3649 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3650 * config/aarch64/aarch64-sve-builtins.def: Include
3651 aarch64-sve-builtins-sve2.def.
3653 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3655 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3656 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3657 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3658 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3659 immediates as well as vector ones.
3660 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3661 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3662 (aarch64_sve_qsub_immediate): Update calls accordingly.
3664 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3666 * config/aarch64/aarch64-sve2.md: Add banner comments.
3667 (<su>mulh<r>s<mode>3): Move further up file.
3668 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3669 (*aarch64_sve2_sra<mode>): Move further down file.
3670 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3672 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3674 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3676 (while_optab_cmp): Handle them.
3677 * config/aarch64/aarch64-sve.md
3678 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3679 and add a "@" marker.
3680 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3681 instead of gen_aarch64_sve2_while_ptest.
3682 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3684 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3686 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3687 (UNSPEC_WHILELE): ...this.
3688 (UNSPEC_WHILE_LO): Rename to...
3689 (UNSPEC_WHILELO): ...this.
3690 (UNSPEC_WHILE_LS): Rename to...
3691 (UNSPEC_WHILELS): ...this.
3692 (UNSPEC_WHILE_LT): Rename to...
3693 (UNSPEC_WHILELT): ...this.
3694 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3695 (cmp_op, while_optab_cmp): Likewise.
3696 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3697 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3698 (svwhilelt): Likewise.
3700 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3702 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3703 (unary_to_uint): Define.
3704 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3705 (unary_count): Rename to...
3706 (unary_to_uint_def, unary_to_uint): ...this.
3707 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3709 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3711 * config/aarch64/aarch64-sve-builtins-functions.h
3712 (code_for_mode_function): New class.
3713 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3714 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3715 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3716 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3717 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3719 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3721 * config/aarch64/iterators.md (addsub): New code attribute.
3722 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3724 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3725 in the asm string and attributes. Fix indentation.
3726 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3728 (@aarch64_sve_<optab><mode>): ...this.
3729 * config/aarch64/aarch64-sve-builtins.h
3730 (function_expander::expand_signed_unpred_op): Delete.
3731 * config/aarch64/aarch64-sve-builtins.cc
3732 (function_expander::expand_signed_unpred_op): Likewise.
3733 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3734 try using code_for_aarch64_sve instead.
3735 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3736 (svqsub_impl): Likewise.
3737 (svqadd, svqsub): Use rtx_code_function instead.
3739 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3741 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3742 (HADDSUB, sur, addsub): Remove them.
3744 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3746 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3747 functions the same way, rather than singling out those that
3748 aren't mapped directly to optabs.
3750 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3752 * target.def (compatible_vector_types_p): New target hook.
3753 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3754 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3755 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3756 * doc/tm.texi: Regenerate.
3757 * gimple-expr.c: Include target.h.
3758 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3759 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3761 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3762 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3763 Use the original predicate if it already has a suitable type.
3765 2020-01-09 Martin Jambor <mjambor@suse.cz>
3767 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3768 resolve_speculation and redirect_call_stmt_to_callee static. Change
3769 return type of set_call_stmt to cgraph_edge *.
3770 * auto-profile.c (afdo_indirect_call): Adjust call to
3771 redirect_call_stmt_to_callee.
3772 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3773 make the this pointer explicit, adjust self-recursive calls and the
3774 call top make_direct. Return the resulting edge.
3775 (cgraph_edge::remove): Make this pointer explicit.
3776 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3777 (cgraph_edge::make_direct): Likewise, adjust call to
3778 resolve_speculation.
3779 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3780 call to set_call_stmt.
3781 (cgraph_update_edges_for_call_stmt_node): Update call to
3782 set_call_stmt and remove.
3783 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3784 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3785 (cgraph_node::create_edge_including_clones): Moved "first" definition
3786 of edge to the block where it was used. Adjusted calls to
3788 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3789 cgraph_edge::remove.
3790 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3791 make_direct and redirect_call_stmt_to_callee.
3792 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3793 resolve_speculation and make_direct.
3794 * ipa-inline-transform.c (inline_transform): Adjust call to
3795 redirect_call_stmt_to_callee.
3796 (check_speculations_1):: Adjust call to resolve_speculation.
3797 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3798 resolve-speculation.
3799 (inline_small_functions): Adjust call to resolve_speculation.
3800 (ipa_inline): Likewise.
3801 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3803 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3804 safe with regards to edge removal, adjust calls to
3805 redirect_call_stmt_to_callee.
3806 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3807 and redirect_call_stmt_to_callee.
3808 * multiple_target.c (create_dispatcher_calls): Adjust call to
3809 redirect_call_stmt_to_callee
3810 (redirect_to_specific_clone): Likewise.
3811 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3812 Adjust calls to cgraph_edge::remove.
3813 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3814 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3815 (expand_call_inline): Adjust call to cgraph_edge::remove.
3817 2020-01-09 Martin Liska <mliska@suse.cz>
3819 * params.opt: Set Optimization for
3820 param_max_speculative_devirt_maydefs.
3822 2020-01-09 Martin Sebor <msebor@redhat.com>
3826 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3828 2020-01-09 Martin Liska <mliska@suse.cz>
3830 * auto-profile.c (auto_profile): Use opt_for_fn
3832 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3833 (propagate_vals_across_arith_jfunc): Likewise.
3834 (hint_time_bonus): Likewise.
3835 (incorporate_penalties): Likewise.
3836 (good_cloning_opportunity_p): Likewise.
3837 (perform_estimation_of_a_value): Likewise.
3838 (estimate_local_effects): Likewise.
3839 (ipcp_propagate_stage): Likewise.
3840 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3841 (set_switch_stmt_execution_predicate): Likewise.
3842 (analyze_function_body): Likewise.
3843 * ipa-inline-analysis.c (offline_size): Likewise.
3844 * ipa-inline.c (early_inliner): Likewise.
3845 * ipa-prop.c (ipa_analyze_node): Likewise.
3846 (ipcp_transform_function): Likewise.
3847 * ipa-sra.c (process_scan_results): Likewise.
3848 (ipa_sra_summarize_function): Likewise.
3849 * params.opt: Rename ipcp-unit-growth to
3850 ipa-cp-unit-growth. Add Optimization for various
3851 IPA-related parameters.
3853 2020-01-09 Richard Biener <rguenther@suse.de>
3856 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3858 2020-01-09 Richard Biener <rguenther@suse.de>
3860 PR tree-optimization/93040
3861 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3863 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3865 * common/config/avr/avr-common.c (avr_option_optimization_table)
3866 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3868 2020-01-09 Martin Liska <mliska@suse.cz>
3870 * cgraphclones.c (symbol_table::materialize_all_clones):
3871 Use cgraph_node::dump_name.
3873 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3876 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3877 output_operand_lossage instead of gcc_unreachable.
3878 * doc/md.texi (riscv f constraint): Fix typo.
3881 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3882 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3883 CONST_SCALAR_INT_P instead of CONST_INT_P.
3884 (*subv<mode>4_1): Rename to ...
3885 (subv<mode>4_1): ... this.
3886 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3887 define_insn_and_split patterns.
3888 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3891 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3893 * vec.c (class selftest::count_dtor): New class.
3894 (selftest::test_auto_delete_vec): New test.
3895 (selftest::vec_c_tests): Call it.
3896 * vec.h (class auto_delete_vec): New class template.
3897 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3899 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3901 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3903 2020-01-08 Jim Wilson <jimw@sifive.com>
3905 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3906 use of TLS_MODEL_LOCAL_EXEC when not pic.
3908 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3910 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3913 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3916 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3917 *stack_protect_set_3 peephole2): Also check that the second
3918 insns source is general_operand.
3921 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3922 predicate for output operand instead of register_operand.
3923 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3924 memory destination and non-memory operands[2].
3926 2020-01-08 Martin Liska <mliska@suse.cz>
3928 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3929 ::dump_asm_name instead of (::name or ::asm_name).
3930 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3931 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3932 (analyze_functions): Likewise.
3933 (expand_all_functions): Likewise.
3934 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3935 (propagate_bits_across_jump_function): Likewise.
3936 (dump_profile_updates): Likewise.
3937 (ipcp_store_bits_results): Likewise.
3938 (ipcp_store_vr_results): Likewise.
3939 * ipa-devirt.c (dump_targets): Likewise.
3940 * ipa-fnsummary.c (analyze_function_body): Likewise.
3941 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3942 (process_hsa_functions): Likewise.
3943 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3944 (set_alias_uids): Likewise.
3945 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3946 * ipa-inline.c (recursive_inlining): Likewise.
3947 (inline_to_all_callers_1): Likewise.
3948 (ipa_inline): Likewise.
3949 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3950 (ipa_propagate_frequency): Likewise.
3951 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3952 (remove_described_reference): Likewise.
3953 * ipa-pure-const.c (worse_state): Likewise.
3954 (check_retval_uses): Likewise.
3955 (analyze_function): Likewise.
3956 (propagate_pure_const): Likewise.
3957 (propagate_nothrow): Likewise.
3958 (dump_malloc_lattice): Likewise.
3959 (propagate_malloc): Likewise.
3960 (pass_local_pure_const::execute): Likewise.
3961 * ipa-visibility.c (optimize_weakref): Likewise.
3962 (function_and_variable_visibility): Likewise.
3963 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3964 (ipa_discover_variable_flags): Likewise.
3965 * lto-streamer-out.c (output_function): Likewise.
3966 (output_constructor): Likewise.
3967 * tree-inline.c (copy_bb): Likewise.
3968 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3969 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3971 2020-01-08 Richard Biener <rguenther@suse.de>
3974 * tree-eh.c (sink_clobbers): Update virtual operands for
3975 the first and last stmt only. Add a dry-run capability.
3976 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3977 after CFG manipulations and in RPO order to catch all
3978 secondary opportunities reliably.
3980 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3983 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3985 2019-01-08 Richard Biener <rguenther@suse.de>
3988 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3989 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3990 virtual operand, also updating SSA use.
3991 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3992 Update stmt after resetting virtual operand.
3993 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3994 * gimple-iterator.c (gsi_remove): When not removing the stmt
3995 permanently do not delink immediate uses or mark the stmt modified.
3997 2020-01-08 Martin Liska <mliska@suse.cz>
3999 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
4000 (ipa_call_context::estimate_size_and_time): Likewise.
4001 (inline_analyze_function): Likewise.
4003 2020-01-08 Martin Liska <mliska@suse.cz>
4005 * cgraph.c (cgraph_node::dump): Use systematically
4008 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4010 Add -nodevicespecs option for avr.
4013 * config/avr/avr.opt (-nodevicespecs): New driver option.
4014 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
4015 "-specs=device-specs/..." if that option is not set.
4016 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
4018 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
4020 Implement 64-bit double functions for avr.
4023 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
4024 --with-double-comparison.
4025 * doc/install.texi: Document them.
4026 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
4027 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
4028 <WITH_DOUBLE_COMPARISON>: New built-in defines.
4029 * doc/invoke.texi (AVR Built-in Macros): Document them.
4030 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
4031 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
4032 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
4034 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
4037 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
4038 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
4039 when only building rm-profile multilibs.
4041 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
4044 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
4045 lattice for a value to check.
4046 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
4047 finite propagation in self-recursive scc.
4049 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
4051 * ipa-inline.c (caller_growth_limits): Restore the AND.
4053 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4055 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4056 (VEC_ALLREG_ALT): New iterator.
4057 (VEC_ALLREG_INT_MODE): New iterator.
4058 (VCMP_MODE): New iterator.
4059 (VCMP_MODE_INT): New iterator.
4060 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4061 (vec_cmp<u>v64qidi): New define_expand.
4062 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4063 (vec_cmpu<mode>di_exec): New define_expand.
4064 (vec_cmp<u>v64qidi_exec): New define_expand.
4065 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4066 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4067 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4068 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4069 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4070 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4071 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4072 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4073 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4074 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4076 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4077 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4079 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4081 * config/gcn/constraints.md (DA): Update description and match.
4083 (Db): New constraint.
4084 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4086 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4087 Implement 'Db' mixed immediate type.
4088 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4089 (addcv64si3_dup<exec_vcc>): Delete.
4090 (subcv64si3<exec_vcc>): Rework constraints.
4091 (addv64di3): Rework constraints.
4092 (addv64di3_exec): Rework constraints.
4093 (subv64di3): Rework constraints.
4094 (addv64di3_dup): Delete.
4095 (addv64di3_dup_exec): Delete.
4096 (addv64di3_zext): Rework constraints.
4097 (addv64di3_zext_exec): Rework constraints.
4098 (addv64di3_zext_dup): Rework constraints.
4099 (addv64di3_zext_dup_exec): Rework constraints.
4100 (addv64di3_zext_dup2): Rework constraints.
4101 (addv64di3_zext_dup2_exec): Rework constraints.
4102 (addv64di3_sext_dup2): Rework constraints.
4103 (addv64di3_sext_dup2_exec): Rework constraints.
4105 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4107 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4108 existing target checks.
4110 2020-01-07 Richard Biener <rguenther@suse.de>
4112 * doc/install.texi: Bump minimal supported MPC version.
4114 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4116 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4117 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4118 * langhooks.c: Include stor-layout.h.
4119 (lhd_simulate_enum_decl): New function.
4120 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4121 handle_arm_sve_h for the LTO frontend.
4122 (register_vector_type): Cope with null returns from pushdecl.
4124 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4126 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4127 (aarch64_sve::nvectors_if_data_type): Replace with...
4128 (aarch64_sve::builtin_type_p): ...this.
4129 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4130 (find_vector_type): Delete.
4131 (add_sve_type_attribute): New function.
4132 (lookup_sve_type_attribute): Likewise.
4133 (register_builtin_types): Add an "SVE type" attribute to each type.
4134 (register_tuple_type): Likewise.
4135 (svbool_type_p, nvectors_if_data_type): Delete.
4136 (mangle_builtin_type): Use lookup_sve_type_attribute.
4137 (builtin_type_p): Likewise. Add an overload that returns the
4138 number of constituent vector and predicate registers.
4139 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4140 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4141 instead of aarch64_sve_argument_p.
4142 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4143 (aarch64_pass_by_reference): Likewise.
4144 (aarch64_function_value_1): Likewise.
4145 (aarch64_return_in_memory): Likewise.
4146 (aarch64_layout_arg): Likewise.
4148 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4150 PR tree-optimization/93156
4151 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4152 least significant bit is always clear.
4154 PR tree-optimization/93118
4155 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4156 simplifier with two intermediate conversions.
4158 2020-01-07 Martin Liska <mliska@suse.cz>
4160 * params.opt: Add Optimization for various parameters.
4162 2020-01-07 Martin Liska <mliska@suse.cz>
4165 * doc/extend.texi: Explain cloning for target_clone
4168 2020-01-07 Martin Liska <mliska@suse.cz>
4170 PR tree-optimization/92860
4171 * common.opt: Make in Optimization option
4172 as it is affected by -O0, which is an Optimization
4174 * tree-inline.c (tree_inlinable_function_p):
4175 Use opt_for_fn for warn_inline.
4176 (expand_call_inline): Likewise.
4178 2020-01-07 Martin Liska <mliska@suse.cz>
4180 PR tree-optimization/92860
4181 * common.opt: Make flag_ree as optimization
4184 2020-01-07 Martin Liska <mliska@suse.cz>
4186 PR optimization/92860
4187 * params.opt: Mark param_min_crossjump_insns with Optimization
4190 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4192 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4193 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4195 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4197 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4198 helper function to return the valid addressing formats for a given
4199 hard register and mode.
4200 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4202 * config/rs6000/constraints.md (Q constraint): Update
4204 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4207 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4208 Use 'Q' for doing vector extract from memory.
4209 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4211 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4212 doing vector extract from memory.
4213 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4214 extract from memory.
4216 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4217 for the offset being 34-bits when -mcpu=future is used.
4219 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4221 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4222 instead of cmpib_comparison_operator in cmpib patterns.
4223 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4224 of cmpib_comparison_operator. Revise comment.
4226 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4228 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4229 in an IFN_DIV_POW2 node to be equal.
4231 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4233 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4234 (vect_check_scalar_mask): ...this.
4235 (vectorizable_store, vectorizable_load): Update call accordingly.
4236 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4237 argument in calls to conditional internal functions.
4239 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4241 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4242 '0' matching inputs.
4243 (subv64di3_exec): Likewise.
4245 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4247 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4248 * doc/md.texi (movstr): Likewise.
4250 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4252 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4255 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4257 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4259 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4260 to a temporary file and use move-if-change to update the real
4261 file where necessary.
4263 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4265 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4266 rather than Upa for CPY /M.
4268 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4270 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4273 2020-01-06 Martin Liska <mliska@suse.cz>
4275 PR tree-optimization/92860
4276 * params.opt: Mark param_max_combine_insns with Optimization
4279 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4282 * config/i386/i386.md (SWIDWI): New mode iterator.
4283 (DWI, dwi): Add TImode variants.
4284 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4285 <general_hilo_operand> instead of <general_operand>. Use
4286 CONST_SCALAR_INT_P instead of CONST_INT_P.
4287 (*addv<mode>4_1): Rename to ...
4288 (addv<mode>4_1): ... this.
4289 (QWI): New mode attribute.
4290 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4291 define_insn_and_split patterns.
4292 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4294 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4295 <general_hilo_operand> instead of <general_operand>.
4296 (*addcarry<mode>_1): New define_insn.
4297 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4299 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4301 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4302 Use "call" instead of "set".
4304 2020-01-03 Martin Jambor <mjambor@suse.cz>
4307 * ipa-cp.c (print_all_lattices): Skip functions without info.
4309 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4312 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4313 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4314 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4315 for 'e' simd clones.
4318 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4320 (mprefer-vector-width=): Add Save.
4321 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4322 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4323 (ix86_debug_options, ix86_function_specific_print): Adjust
4324 ix86_target_string callers.
4325 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4326 (ix86_valid_target_attribute_tree): Likewise.
4327 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4328 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4329 ix86_target_string caller.
4332 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4333 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4334 instead of gen_int_shift_amount + convert_modes.
4336 PR rtl-optimization/93088
4337 * loop-iv.c (find_single_def_src): Punt after looking through
4338 128 reg copies for regs with single definitions. Move definitions
4341 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4343 * config/arm/arm-c.c (arm_cpu_builtins): Define
4344 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4345 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4346 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4347 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4348 * config/arm/arm-tables.opt: Regenerated.
4349 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4350 arm_arch_i8mm and arm_arch_bf16 when enabled.
4351 * config/arm/arm.h (TARGET_I8MM): New macro.
4352 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4353 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4354 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4355 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4356 (v8_6_a_simd_variants): New.
4357 (v8_*_a_simd_variants): Add i8mm and bf16.
4358 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4360 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4363 * predict.c (compute_function_frequency): Don't call
4364 warn_function_cold on functions that already have cold attribute.
4366 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4369 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4370 COMDAT group function labels in .data.rel.ro.local section.
4371 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4374 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4375 comparison_operator in B and S integer comparisons. Likewise, use
4376 ordered_comparison_operator instead of cmpib_comparison_operator in
4378 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4380 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4382 Update copyright years.
4384 * gcc.c (process_command): Update copyright notice dates.
4385 * gcov-dump.c (print_version): Ditto.
4386 * gcov.c (print_version): Ditto.
4387 * gcov-tool.c (print_version): Ditto.
4388 * gengtype.c (create_file): Ditto.
4389 * doc/cpp.texi: Bump @copying's copyright year.
4390 * doc/cppinternals.texi: Ditto.
4391 * doc/gcc.texi: Ditto.
4392 * doc/gccint.texi: Ditto.
4393 * doc/gcov.texi: Ditto.
4394 * doc/install.texi: Ditto.
4395 * doc/invoke.texi: Ditto.
4397 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4399 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4402 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4404 PR tree-optimization/93098
4405 * match.pd (popcount): For shift amounts, use integer_onep
4406 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4407 tests. Make sure that precision is power of two larger than or equal
4408 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4409 instead of ULL suffixed constants. Formatting fixes.
4411 Copyright (C) 2020 Free Software Foundation, Inc.
4413 Copying and distribution of this file, with or without modification,
4414 are permitted in any medium without royalty provided the copyright
4415 notice and this notice are preserved.