1 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
3 * collect2.c (c_file, o_file): Make const again.
4 (ldout,lderrout, dump_ld_file): Remove.
5 (tool_cleanup): Avoid calling not signal-safe functions.
6 (maybe_run_lto_and_relink): Avoid possible signal handler
7 access to unintialzed memory (lto_o_files).
8 (main): Avoid leaking temp files in $TMPDIR.
9 Initialize c_file/o_file with concat, which avoids exposing
10 uninitialized memory to signal handler, which calls unlink(!).
11 Avoid calling maybe_unlink when the main function returns,
12 since the atexit handler is already doing this.
13 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
15 2020-02-19 Martin Jambor <mjambor@suse.cz>
17 PR tree-optimization/93776
18 * tree-sra.c (create_access): Do not create zero size accesses.
19 (get_access_for_expr): Do not search for zero sized accesses.
21 2020-02-19 Martin Jambor <mjambor@suse.cz>
23 PR tree-optimization/93667
24 * tree-sra.c (scalarizable_type_p): Return false if record fields
25 do not follow wach other.
27 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
29 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
30 rather than fmv.x.s/fmv.s.x.
32 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
34 * config/aarch64/aarch64-simd-builtins.def
35 (intrinsic_vec_smult_lo_): New.
36 (intrinsic_vec_umult_lo_): Likewise.
37 (vec_widen_smult_hi_): Likewise.
38 (vec_widen_umult_hi_): Likewise.
39 * config/aarch64/aarch64-simd.md
40 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
41 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
42 (vmull_high_s16): Likewise.
43 (vmull_high_s32): Likewise.
44 (vmull_high_u8): Likewise.
45 (vmull_high_u16): Likewise.
46 (vmull_high_u32): Likewise.
48 (vmull_s16): Likewise.
49 (vmull_s32): Likewise.
51 (vmull_u16): Likewise.
52 (vmull_u32): Likewise.
54 2020-02-18 Martin Liska <mliska@suse.cz>
56 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
57 bootstrap by missing removal of invalid sanity check.
59 2020-02-18 Martin Liska <mliska@suse.cz>
62 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
63 Always compare LHS of gimple_assign.
65 2020-02-18 Martin Liska <mliska@suse.cz>
68 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
69 and return type of functions.
70 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
71 Drop MALLOC attribute for void functions.
72 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
73 malloc_state for a new VOID clone.
75 2020-02-18 Martin Liska <mliska@suse.cz>
78 * common.opt: Add -fprofile-reproducibility.
79 * doc/invoke.texi: Document it.
80 * value-prof.c (dump_histogram_value):
81 Document and support behavior for counters[0]
82 being a negative value.
83 (get_nth_most_common_value): Handle negative
84 counters[0] in respect to flag_profile_reproducible.
86 2020-02-18 Jakub Jelinek <jakub@redhat.com>
89 * cgraph.c (verify_speculative_call): Use speculative_id instead of
90 speculative_uid in messages. Remove trailing whitespace from error
91 message. Use num_speculative_call_targets instead of
92 num_speculative_targets in a message.
93 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
94 edge messages and stmt instead of cal_stmt in reference message.
96 PR tree-optimization/93780
97 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
98 before calling build_vector_type.
99 (execute_update_addresses_taken): Likewise.
102 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
103 typo, functoin -> function.
104 * tree.c (free_lang_data_in_decl): Fix comment typo,
105 functoin -> function.
106 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
108 2020-02-17 David Malcolm <dmalcolm@redhat.com>
110 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
112 (print_option_information): Don't call get_option_url if URLs
115 2020-02-17 Alexandre Oliva <oliva@adacore.com>
117 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
118 handling of register_common-less targets.
120 2020-02-17 Martin Liska <mliska@suse.cz>
123 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
125 2020-02-17 Martin Liska <mliska@suse.cz>
128 * config/rs6000/rs6000.c (rs6000_option_override_internal):
131 2020-02-17 Martin Liska <mliska@suse.cz>
134 * config/rx/elf.opt: Fix typo.
136 2020-02-17 Richard Biener <rguenther@suse.de>
139 * opts-global.c (print_ignored_options): Use inform and
142 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
145 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
147 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
150 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
151 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
153 2020-02-15 Jason Merrill <jason@redhat.com>
155 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
157 2020-02-15 Jakub Jelinek <jakub@redhat.com>
159 PR tree-optimization/93744
160 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
161 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
162 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
163 sure @2 in the first and @1 in the other patterns has no side-effects.
165 2020-02-15 David Malcolm <dmalcolm@redhat.com>
166 Bernd Edlinger <bernd.edlinger@hotmail.de>
170 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
171 * configure.ac (--with-diagnostics-urls): New configuration
172 option, based on --with-diagnostics-color.
173 (DIAGNOSTICS_URLS_DEFAULT): New define.
174 * config.h: Regenerate.
175 * configure: Regenerate.
176 * diagnostic.c (diagnostic_urls_init): Handle -1 for
177 DIAGNOSTICS_URLS_DEFAULT from configure-time
178 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
179 and TERM_URLS environment variable.
180 * diagnostic-url.h (diagnostic_url_format): New enum type.
181 (diagnostic_urls_enabled_p): rename to...
182 (determine_url_format): ... this, and change return type.
183 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
184 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
185 the linux console, and mingw.
186 (diagnostic_urls_enabled_p): rename to...
187 (determine_url_format): ... this, and adjust.
188 * pretty-print.h (pretty_printer::show_urls): rename to...
189 (pretty_printer::url_format): ... this, and change to enum.
190 * pretty-print.c (pretty_printer::pretty_printer,
191 pp_begin_url, pp_end_url, test_urls): Adjust.
192 * doc/install.texi (--with-diagnostics-urls): Document the new
193 configuration option.
194 (--with-diagnostics-color): Document the existing interaction
195 with GCC_COLORS better.
196 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
197 vindex reference. Update description of defaults based on the above.
198 (-fdiagnostics-color): Update description of how -fdiagnostics-color
199 interacts with GCC_COLORS.
201 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
204 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
205 conjunction with TARGET_GNU_TLS in early return.
207 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
209 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
210 the mode is not wider than UNITS_PER_WORD.
212 2020-02-14 Martin Jambor <mjambor@suse.cz>
214 PR tree-optimization/93516
215 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
216 access of the same type as the parent.
217 (propagate_subaccesses_from_lhs): Likewise.
219 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
222 * config/i386/avx512vbmi2intrin.h
223 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
224 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
225 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
226 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
227 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
228 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
229 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
230 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
231 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
232 of lacking a closing parenthesis.
233 * config/i386/avx512vbmi2vlintrin.h
234 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
235 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
236 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
237 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
238 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
239 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
240 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
241 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
242 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
243 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
244 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
245 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
246 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
247 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
248 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
249 _mm_shldi_epi32, _mm_mask_shldi_epi32,
250 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
251 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
253 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
256 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
257 the target function entry.
259 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
261 * common/config/arc/arc-common.c (arc_option_optimization_table):
262 Disable if-conversion step when optimized for size.
264 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
266 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
267 R12-R15 are always in ARCOMPACT16_REGS register class.
268 * config/arc/arc.opt (mq-class): Deprecate.
269 * config/arc/constraint.md ("q"): Remove dependency on mq-class
271 * doc/invoke.texi (mq-class): Update text.
272 * common/config/arc/arc-common.c (arc_option_optimization_table):
275 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
277 * config/arc/arc.c (arc_insn_cost): New function.
278 (TARGET_INSN_COST): Define.
279 * config/arc/arc.md (cost): New attribute.
280 (add_n): Use arc_nonmemory_operand.
281 (ashlsi3_insn): Likewise, also update constraints.
282 (ashrsi3_insn): Likewise.
284 (add_shift): Likewise.
285 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
287 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
289 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
291 (umulsidi_600): Likewise.
293 2020-02-13 Jakub Jelinek <jakub@redhat.com>
296 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
297 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
298 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
299 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
300 pass __A to the builtin followed by __W instead of __A followed by
302 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
303 _mm512_mask_popcnt_epi64): Likewise.
304 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
305 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
306 _mm256_mask_popcnt_epi64): Likewise.
308 PR tree-optimization/93582
309 * fold-const.h (shift_bytes_in_array_left,
310 shift_bytes_in_array_right): Declare.
311 * fold-const.c (shift_bytes_in_array_left,
312 shift_bytes_in_array_right): New function, moved from
313 gimple-ssa-store-merging.c, no longer static.
314 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
315 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
316 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
317 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
318 shift_bytes_in_array.
319 (verify_shift_bytes_in_array): Rename to ...
320 (verify_shift_bytes_in_array_left): ... this. Use
321 shift_bytes_in_array_left instead of shift_bytes_in_array.
322 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
323 instead of verify_shift_bytes_in_array.
324 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
325 / native_interpret_expr where the store covers all needed bits,
326 punt on PDP-endian, otherwise allow all involved offsets and sizes
327 not to be byte-aligned.
330 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
331 use const_0_to_255_operand predicate instead of immediate_operand.
332 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
333 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
334 vgf2p8affineinvqb_<mode><mask_name>,
335 vgf2p8affineqb_<mode><mask_name>): Drop mode from
336 const_0_to_255_operand predicated operands.
338 2020-02-12 Jeff Law <law@redhat.com>
340 * config/h8300/h8300.md (comparison shortening peepholes): Use
341 a mode iterator to merge the HImode and SImode peepholes.
343 2020-02-12 Jakub Jelinek <jakub@redhat.com>
346 * real.c (is_even): Make static. Function comment fix.
347 (is_halfway_below): Make static, don't assert R is not inf/nan,
348 instead return false for those. Small formatting fixes.
350 2020-02-12 Martin Sebor <msebor@redhat.com>
353 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
354 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
355 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
356 (strlen_check_and_optimize_call): Adjust callee name.
358 2020-02-12 Jeff Law <law@redhat.com>
360 * config/h8300/h8300.md (comparison shortening peepholes): Drop
361 (and (xor)) variant. Combine other two into single peephole.
363 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
365 PR rtl-optimization/93565
366 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
368 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
370 * config/aarch64/aarch64-simd.md
371 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
372 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
373 generating separate ADDV and zero_extend patterns.
374 * config/aarch64/iterators.md (VDQV_E): New iterator.
376 2020-02-12 Jeff Law <law@redhat.com>
378 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
379 expanders, splits, etc.
380 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
381 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
382 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
383 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
384 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
385 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
387 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
389 2020-02-12 Jakub Jelinek <jakub@redhat.com>
392 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
393 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
394 TARGET_AVX512DQ from condition.
395 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
396 instead of <mask_mode512bit_condition> in condition. If
397 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
399 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
402 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
405 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
407 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
409 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
410 where strlen is more legible.
411 (rs6000_builtin_vectorized_libmass): Ditto.
412 (rs6000_print_options_internal): Ditto.
414 2020-02-11 Martin Sebor <msebor@redhat.com>
416 PR tree-optimization/93683
417 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
419 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
421 * config/rs6000/predicates.md (cint34_operand): Rename the
422 -mprefixed-addr option to be -mprefixed.
423 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
424 the -mprefixed-addr option to be -mprefixed.
425 (OTHER_FUTURE_MASKS): Likewise.
426 (POWERPC_MASKS): Likewise.
427 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
428 the -mprefixed-addr option to be -mprefixed. Change error
429 messages to refer to -mprefixed.
430 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
432 (rs6000_legitimate_offset_address_p): Likewise.
433 (rs6000_mode_dependent_address): Likewise.
434 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
435 "-mprefixed" for target attributes and pragmas.
436 (address_to_insn_form): Rename the -mprefixed-addr option to be
438 (rs6000_adjust_insn_length): Likewise.
439 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
440 -mprefixed-addr option to be -mprefixed.
441 (ASM_OUTPUT_OPCODE): Likewise.
442 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
443 -mprefixed-addr option to be -mprefixed.
444 * config/rs6000/rs6000.opt (-mprefixed): Rename the
445 -mprefixed-addr option to be prefixed. Change the option from
446 being undocumented to being documented.
447 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
448 -mprefixed option. Update the -mpcrel documentation to mention
451 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
453 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
454 including FIRST_PSEUDO_REGISTER - 1.
455 * ira-color.c (print_hard_reg_set): Ditto.
457 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
459 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
460 (USTERNOP_QUALIFIERS): New define.
461 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
462 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
463 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
464 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
465 * config/arm/arm_neon.h (vusdot_s32): New.
466 (vusdot_lane_s32): New.
467 (vusdotq_lane_s32): New.
468 (vsudot_lane_s32): New.
469 (vsudotq_lane_s32): New.
470 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
471 * config/arm/iterators.md (DOTPROD_I8MM): New.
472 (sup, opsuffix): Add <us/su>.
473 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
474 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
476 2020-02-11 Richard Biener <rguenther@suse.de>
478 PR tree-optimization/93661
479 PR tree-optimization/93662
480 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
482 * tree-sra.c (get_access_for_expr): Likewise.
484 2020-02-10 Jakub Jelinek <jakub@redhat.com>
487 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
488 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
489 Change condition from TARGET_AVX2 to TARGET_AVX.
491 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
494 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
497 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
499 Try to generate zero-based comparisons.
500 * config/cris/cris.c (cris_reduce_compare): New function.
501 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
502 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
503 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
505 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
508 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
509 in Thumb state and also as a destination in Arm state. Add T16
512 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
514 * md.texi (Define Subst): Match closing paren in example.
516 2020-02-10 Jakub Jelinek <jakub@redhat.com>
520 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
521 arguments of strncmp.
523 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
526 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
527 but different source value.
528 (adjust_callers_for_value_intersection): New function.
529 (gather_edges_for_value): Adjust order of callers to let a
530 non-self-recursive caller be the first element.
531 (self_recursive_pass_through_p): Add a new parameter "simple", and
532 check generalized self-recursive pass-through jump function.
533 (self_recursive_agg_pass_through_p): Likewise.
534 (find_more_scalar_values_for_callers_subset): Compute value from
535 pass-through jump function for self-recursive.
536 (intersect_with_plats): Cleanup previous implementation code for value
537 itersection with self-recursive call edge.
538 (intersect_with_agg_replacements): Likewise.
539 (intersect_aggregates_with_edge): Deduce value from pass-through jump
540 function for self-recursive call edge. Cleanup previous implementation
541 code for value intersection with self-recursive call edge.
542 (decide_whether_version_node): Remove dead callers and adjust order
543 to let a non-self-recursive caller be the first element.
545 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
547 * recog.c: Move pass_split_before_sched2 code in front of
548 pass_split_before_regstack.
549 (pass_data_split_before_sched2): Rename pass to split3 from split4.
550 (pass_data_split_before_regstack): Rename pass to split4 from split3.
551 (rest_of_handle_split_before_sched2): Remove.
552 (pass_split_before_sched2::execute): Unconditionally call
554 (enable_split_before_sched2): New function.
555 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
556 (pass_split_before_regstack::gate): Ditto.
557 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
558 Update name check for renamed split4 pass.
559 * config/sh/sh.c (register_sh_passes): Update pass insertion
560 point for renamed split4 pass.
562 2020-02-09 Jakub Jelinek <jakub@redhat.com>
564 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
565 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
566 copying them around between host and target.
568 2020-02-08 Andrew Pinski <apinski@marvell.com>
571 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
572 STRICT_ALIGNMENT also.
574 2020-02-08 Jim Wilson <jimw@sifive.com>
577 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
579 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
580 Jakub Jelinek <jakub@redhat.com>
583 * config/i386/i386.h (CALL_USED_REGISTERS): Make
584 xmm16-xmm31 call-used even in 64-bit ms-abi.
586 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
588 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
589 (simd_ummla, simd_usmmla): Likewise.
590 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
591 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
594 2020-02-07 Richard Biener <rguenther@suse.de>
597 * tree-inline.c (fold_marked_statements): Do a PRE walk,
598 skipping unreachable regions.
599 (optimize_inline_calls): Skip folding stmts when we didn't
602 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
605 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
606 Don't return aggregates with only SFmode and DFmode in SSE
608 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
610 2020-02-07 Jakub Jelinek <jakub@redhat.com>
613 * config/rs6000/rs6000-logue.c
614 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
615 if it fails, move rs into end_addr and retry. Add
616 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
617 the insn pattern doesn't describe well what exactly happens to
621 * config/i386/predicates.md (avx_identity_operand): Remove.
622 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
623 (avx_<castmode><avxsizesuffix>_<castmode>,
624 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
625 a VEC_CONCAT of the operand and UNSPEC_CAST.
626 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
627 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
631 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
632 recog_data.insn if distance_non_agu_define changed it.
634 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
637 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
638 we only had X-FORM (reg+reg) addressing for vectors. Also before
639 ISA 3.0, we only had X-FORM addressing for scalars in the
640 traditional Altivec registers.
642 2020-02-06 <zhongyunde@huawei.com>
643 Vladimir Makarov <vmakarov@redhat.com>
645 PR rtl-optimization/93561
646 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
649 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
651 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
654 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
656 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
657 where the low and the high 32 bits are equal to each other specially,
658 with an rldimi instruction.
660 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
662 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
664 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
666 * config/arm/arm-tables.opt: Regenerate.
668 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
671 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
672 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
673 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
675 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
677 PR rtl-optimization/87763
678 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
680 2020-02-06 Delia Burduv <delia.burduv@arm.com>
682 * config/aarch64/aarch64-simd-builtins.def
683 (bfmlaq): New built-in function.
684 (bfmlalb): New built-in function.
685 (bfmlalt): New built-in function.
686 (bfmlalb_lane): New built-in function.
687 (bfmlalt_lane): New built-in function.
688 * config/aarch64/aarch64-simd.md
689 (aarch64_bfmmlaqv4sf): New pattern.
690 (aarch64_bfmlal<bt>v4sf): New pattern.
691 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
692 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
693 (vbfmlalbq_f32): New intrinsic.
694 (vbfmlaltq_f32): New intrinsic.
695 (vbfmlalbq_lane_f32): New intrinsic.
696 (vbfmlaltq_lane_f32): New intrinsic.
697 (vbfmlalbq_laneq_f32): New intrinsic.
698 (vbfmlaltq_laneq_f32): New intrinsic.
699 * config/aarch64/iterators.md (BF_MLA): New int iterator.
700 (bt): New int attribute.
702 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
704 * config/i386/i386.md (*pushtf): Emit "#" instead of
705 calling gcc_unreachable in insn output.
708 (*pushsf_rex64): Ditto for alternatives other than 1.
709 (*pushsf): Ditto for alternatives other than 1.
711 2020-02-06 Martin Liska <mliska@suse.cz>
713 PR gcov-profile/91971
714 PR gcov-profile/93466
715 * coverage.c (coverage_init): Revert mangling of
716 path into filename. It can lead to huge filename length.
717 Creation of subfolders seem more natural.
719 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
722 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
723 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
724 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
726 2020-02-06 Jakub Jelinek <jakub@redhat.com>
729 * config/i386/predicates.md (avx_identity_operand): New predicate.
730 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
731 define_insn_and_split.
734 * omp-low.c (use_pointer_for_field): For nested constructs, also
735 look for map clauses on target construct.
736 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
737 taskreg_nesting_level.
740 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
741 shared clause, call omp_notice_variable on outer context if any.
743 2020-02-05 Jason Merrill <jason@redhat.com>
746 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
747 non-zero address even if weak and not yet defined.
749 2020-02-05 Martin Sebor <msebor@redhat.com>
751 PR tree-optimization/92765
752 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
753 * tree-ssa-strlen.c (compute_string_length): Remove.
754 (determine_min_objsize): Remove.
755 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
756 Avoid using type size as the upper bound on string length.
757 (handle_builtin_string_cmp): Add an argument. Adjust.
758 (strlen_check_and_optimize_call): Pass additional argument to
759 handle_builtin_string_cmp.
761 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
763 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
764 (*pushdi2_rex64 peephole2): Unconditionally split after
766 (*ashl<mode>3_doubleword): Ditto.
767 (*<shift_insn><mode>3_doubleword): Ditto.
769 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
772 * config/rs6000/rs6000.c (get_vector_offset): Fix
774 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
776 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
778 2020-02-05 David Malcolm <dmalcolm@redhat.com>
781 (Special Functions for Debugging the Analyzer): Update description
782 of __analyzer_dump_exploded_nodes.
784 2020-02-05 Jakub Jelinek <jakub@redhat.com>
787 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
788 include sets and not clobbers in the vzeroupper pattern.
789 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
790 the parallel has 17 (64-bit) or 9 (32-bit) elts.
791 (*avx_vzeroupper_1): New define_insn_and_split.
794 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
795 don't run when !optimize.
796 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
799 2020-02-05 Richard Biener <rguenther@suse.de>
802 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
803 checks before matching calls.
805 2020-02-05 Jakub Jelinek <jakub@redhat.com>
807 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
808 function comment typo.
811 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
812 simd_clone_create failed when i == 0, adjust clone->nargs by
815 2020-02-05 Martin Liska <mliska@suse.cz>
818 * doc/invoke.texi: Document that one should
819 not combine ASLR and -fpch.
821 2020-02-04 Richard Biener <rguenther@suse.de>
823 PR tree-optimization/93538
824 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
826 2020-02-04 Richard Biener <rguenther@suse.de>
828 PR tree-optimization/91123
829 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
830 (vn_walk_cb_data::last_vuse): New member.
831 (vn_walk_cb_data::saved_operands): Likewsie.
832 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
833 (vn_walk_cb_data::push_partial_def): Use finish.
834 (vn_reference_lookup_2): Update last_vuse and use finish if
835 we've saved operands.
836 (vn_reference_lookup_3): Use finish and update calls to
837 push_partial_defs everywhere. When translating through
838 memcpy or aggregate copies save off operands and alias-set.
839 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
840 operation for redundant store removal.
842 2020-02-04 Richard Biener <rguenther@suse.de>
844 PR tree-optimization/92819
845 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
846 generating more stmts than before.
848 2020-02-04 Martin Liska <mliska@suse.cz>
850 * config/arm/arm.c (arm_gen_far_branch): Move the function
851 outside of selftests.
853 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
855 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
856 function to adjust PC-relative vector addresses.
857 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
858 handle vectors with PC-relative addresses.
860 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
862 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
864 (hard_reg_and_mode_to_addr_mask): Delete.
865 (rs6000_adjust_vec_address): If the original vector address
866 was REG+REG or REG+OFFSET and the element is not zero, do the add
867 of the elements in the original address before adding the offset
868 for the vector element. Use address_to_insn_form to validate the
869 address using the register being loaded, rather than guessing
870 whether the address is a DS-FORM or DQ-FORM address.
872 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
874 * config/rs6000/rs6000.c (get_vector_offset): New helper function
875 to calculate the offset in memory from the start of a vector of a
876 particular element. Add code to keep the element number in
877 bounds if the element number is variable.
878 (rs6000_adjust_vec_address): Move calculation of offset of the
879 vector element to get_vector_offset.
880 (rs6000_split_vec_extract_var): Do not do the initial AND of
881 element here, move the code to get_vector_offset.
883 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
885 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
888 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
890 * config/rs6000/constraints.md: Improve documentation.
892 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
895 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
896 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
898 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
900 * config.gcc: Remove "carrizo" support.
901 * config/gcn/gcn-opts.h (processor_type): Likewise.
902 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
903 * config/gcn/gcn.opt (gpu_type): Likewise.
904 * config/gcn/t-omp-device: Likewise.
906 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
909 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
910 * config/arm/arm.c (arm_gen_far_branch): New function
912 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
914 2020-02-03 Julian Brown <julian@codesourcery.com>
915 Tobias Burnus <tobias@codesourcery.com>
917 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
919 2020-02-03 Jakub Jelinek <jakub@redhat.com>
922 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
923 valid RTL to sum up the lowest and second lowest bytes of the popcnt
926 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
928 PR rtl-optimization/91333
929 * ira-color.c (struct allocno_color_data): Add member
931 (init_allocno_threads): Set the member up.
932 (bucket_allocno_compare_func): Add compare hard reg
935 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
937 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
939 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
940 * config.in: Regenerated.
941 * configure: Regenerated.
942 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
943 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
944 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
946 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
948 * configure: Regenerate.
950 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
952 PR rtl-optimization/91333
953 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
954 reg preferences comparison up.
956 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
958 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
959 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
960 aarch64-sve-builtins-base.h.
961 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
962 aarch64-sve-builtins-base.cc.
963 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
964 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
966 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
967 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
968 (svcvtnt): New functions.
969 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
970 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
971 (svcvtnt): New functions.
972 (svcvt): Add a form that converts f32 to bf16.
973 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
974 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
976 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
977 Treat B as bfloat16_t.
978 (ternary_bfloat_lane_base): New class.
979 (ternary_bfloat_def): Likewise.
980 (ternary_bfloat): New shape.
981 (ternary_bfloat_lane_def): New class.
982 (ternary_bfloat_lane): New shape.
983 (ternary_bfloat_lanex2_def): New class.
984 (ternary_bfloat_lanex2): New shape.
985 (ternary_bfloat_opt_n_def): New class.
986 (ternary_bfloat_opt_n): New shape.
987 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
988 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
989 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
990 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
991 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
992 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
993 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
994 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
995 the pattern off the narrow mode instead of the wider one.
996 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
997 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
998 (sve_fp_op): Handle them.
999 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1000 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1002 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1004 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1005 * config/aarch64/aarch64-modes.def (BF): Move definition before
1006 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1007 (SVE_MODES): Handle BF modes.
1008 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1010 (aarch64_full_sve_mode): Likewise.
1011 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1013 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1014 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1015 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1016 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1018 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1020 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1021 (TYPES_all_data): Add bf16.
1022 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1023 (register_tuple_type): Increase buffer size.
1024 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1025 (bf16): New type suffix.
1026 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1027 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1028 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1029 Change type from all_data to all_arith.
1030 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1033 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1034 Matthew Malcomson <matthew.malcomson@arm.com>
1035 Richard Sandiford <richard.sandiford@arm.com>
1037 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1038 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1039 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1040 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1041 __ARM_FEATURE_MATMUL_FP64.
1042 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1043 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1044 be disabled at the same time.
1045 (f32mm): New extension.
1046 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1047 (AARCH64_FL_F64MM): Bump to the next bit up.
1048 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1049 (TARGET_SVE_F64MM): New macros.
1050 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1051 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1052 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1053 (UNSPEC_ZIP2Q): New unspeccs.
1054 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1055 (optab, sur, perm_insn): Handle the new unspecs.
1056 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1057 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1058 TARGET_SVE_F64MM instead of separate tests.
1059 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1060 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1061 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1062 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1063 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1064 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1065 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1066 (TYPES_s_signed): New macro.
1067 (TYPES_s_integer): Use it.
1068 (TYPES_d_float): New macro.
1069 (TYPES_d_data): Use it.
1070 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1071 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1072 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1073 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1074 (svmmla): New shape.
1075 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1076 template parameters.
1077 (ternary_resize2_lane_base): Likewise.
1078 (ternary_resize2_base): New class.
1079 (ternary_qq_lane_base): Likewise.
1080 (ternary_intq_uintq_lane_def): Likewise.
1081 (ternary_intq_uintq_lane): New shape.
1082 (ternary_intq_uintq_opt_n_def): New class
1083 (ternary_intq_uintq_opt_n): New shape.
1084 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1085 (ternary_uintq_intq_def): New class.
1086 (ternary_uintq_intq): New shape.
1087 (ternary_uintq_intq_lane_def): New class.
1088 (ternary_uintq_intq_lane): New shape.
1089 (ternary_uintq_intq_opt_n_def): New class.
1090 (ternary_uintq_intq_opt_n): New shape.
1091 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1092 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1093 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1094 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1096 (svdotprod_lane_impl): ...this new class.
1097 (svmmla_impl, svusdot_impl): New classes.
1098 (svdot_lane): Update to use svdotprod_lane_impl.
1099 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1100 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1102 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1103 function, with no types defined.
1104 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1105 AARCH64_FL_I8MM functions.
1106 (svmmla): New AARCH64_FL_F32MM function.
1107 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1108 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1109 AARCH64_FL_F64MM function.
1110 (REQUIRED_EXTENSIONS):
1112 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1114 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1117 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1119 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1120 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1121 (*movti_internal): Do not check for
1122 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1123 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1124 just after check for TARGET_AVX.
1125 (*movdf_internal): Ditto.
1126 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1127 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1128 * config/i386/sse.md (mov<mode>_internal): Only check
1129 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1130 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1131 (<sse>_andnot<mode>3<mask_name>): Move check for
1132 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1133 (<code><mode>3<mask_name>): Ditto.
1134 (*andnot<mode>3): Ditto.
1135 (*andnottf3): Ditto.
1136 (*<code><mode>3): Ditto.
1137 (*<code>tf3): Ditto.
1138 (*andnot<VI:mode>3): Remove
1139 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1140 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1141 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1142 (sse4_1_blendv<ssemodesuffix>): Ditto.
1143 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1144 Explain that tune applies to 128bit instructions only.
1146 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1148 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1149 to definition of hsa_kernel_description. Parse assembly to find SGPR
1150 and VGPR count of kernel and store in hsa_kernel_description.
1152 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1154 PR rtl-optimization/91838
1155 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1156 to truncate if allowed or reject combination.
1158 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1160 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1161 (find_inv_vars_cb): Likewise.
1163 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1165 * calls.c (special_function_p): Split out the check for DECL_NAME
1166 being non-NULL and fndecl being extern at file scope into a
1167 new maybe_special_function_p and call it. Drop check for fndecl
1168 being non-NULL that was after a usage of DECL_NAME (fndecl).
1169 * tree.h (maybe_special_function_p): New inline function.
1171 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1173 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1174 (mask_gather_load<mode>): ... here, and zero-initialize the
1176 (maskload<mode>di): Zero-initialize the destination.
1179 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1182 * doc/analyzer.texi (Limitations): Note that constraints on
1183 floating-point values are currently ignored.
1185 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1188 * symtab.c (symtab_node::noninterposable_alias): If localalias
1189 already exists, but is not usable, append numbers after it until
1190 a unique name is found. Formatting fix.
1193 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1196 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1198 * config/gcn/gcn.c (print_operand): Handle LTGT.
1199 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1201 2020-01-30 Richard Biener <rguenther@suse.de>
1203 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1204 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1206 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1208 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1209 without a DECL in .data.rel.ro.local.
1211 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1214 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1218 * config/i386/sse.md
1219 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1220 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1221 any_extend code iterator instead of always zero_extend.
1222 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1223 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1224 Use any_extend code iterator instead of always zero_extend.
1225 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1226 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1227 Use any_extend code iterator instead of always zero_extend.
1228 (*sse2_pmovmskb_ext): New define_insn.
1229 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1232 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1233 (*popcountsi2_zext_falsedep): New define_insn.
1235 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1237 * config.in: Regenerated.
1238 * configure: Regenerated.
1240 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1243 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1244 LLVM's assembler changed the default in version 9.
1246 2020-01-24 Jeff Law <law@redhat.com>
1248 PR tree-optimization/89689
1249 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1251 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1255 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1257 PR rtl-optimization/87763
1258 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1259 simplification to handle subregs as well as bare regs.
1260 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1262 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1265 * ira.c (ira): Revert use of simplified LRA algorithm.
1267 2020-01-29 Martin Jambor <mjambor@suse.cz>
1269 PR tree-optimization/92706
1270 * tree-sra.c (struct access): Fields first_link, last_link,
1271 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1272 next_rhs_queued and grp_rhs_queued respectively, new fields
1273 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1274 (struct assign_link): Field next renamed to next_rhs, new field
1275 next_lhs. Updated comment.
1276 (work_queue_head): Renamed to rhs_work_queue_head.
1277 (lhs_work_queue_head): New variable.
1278 (add_link_to_lhs): New function.
1279 (relink_to_new_repr): Also relink LHS lists.
1280 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1281 (add_access_to_lhs_work_queue): New function.
1282 (pop_access_from_work_queue): Renamed to
1283 pop_access_from_rhs_work_queue.
1284 (pop_access_from_lhs_work_queue): New function.
1285 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1287 (child_would_conflict_in_lacc): Renamed to
1288 child_would_conflict_in_acc. Adjusted parameter names.
1289 (create_artificial_child_access): New parameter set_grp_read, use it.
1290 (subtree_mark_written_and_enqueue): Renamed to
1291 subtree_mark_written_and_rhs_enqueue.
1292 (propagate_subaccesses_across_link): Renamed to
1293 propagate_subaccesses_from_rhs.
1294 (propagate_subaccesses_from_lhs): New function.
1295 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1298 2020-01-29 Martin Jambor <mjambor@suse.cz>
1300 PR tree-optimization/92706
1301 * tree-sra.c (struct access): Adjust comment of
1302 grp_total_scalarization.
1303 (find_access_in_subtree): Look for single children spanning an entire
1305 (scalarizable_type_p): Allow register accesses, adjust callers.
1306 (completely_scalarize): Remove function.
1307 (scalarize_elem): Likewise.
1308 (create_total_scalarization_access): Likewise.
1309 (sort_and_splice_var_accesses): Do not track total scalarization
1311 (analyze_access_subtree): New parameter totally, adjust to new meaning
1312 of grp_total_scalarization.
1313 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1314 (can_totally_scalarize_forest_p): New function.
1315 (create_total_scalarization_access): Likewise.
1316 (create_total_access_and_reshape): Likewise.
1317 (total_should_skip_creating_access): Likewise.
1318 (totally_scalarize_subtree): Likewise.
1319 (analyze_all_variable_accesses): Perform total scalarization after
1320 subaccess propagation using the new functions above.
1321 (initialize_constant_pool_replacements): Output initializers by
1322 traversing the access tree.
1324 2020-01-29 Martin Jambor <mjambor@suse.cz>
1326 * tree-sra.c (verify_sra_access_forest): New function.
1327 (verify_all_sra_access_forests): Likewise.
1328 (create_artificial_child_access): Set parent.
1329 (analyze_all_variable_accesses): Call the verifier.
1331 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1333 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1334 if called on indirect edge.
1335 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1336 speculative call if needed.
1338 2020-01-29 Richard Biener <rguenther@suse.de>
1340 PR tree-optimization/93428
1341 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1342 permutation when the load node is created.
1343 (vect_analyze_slp_instance): Re-use it here.
1345 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1347 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1349 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1351 PR rtl-optimization/93272
1352 * ira-lives.c (process_out_of_region_eh_regs): New function.
1353 (process_bb_node_lives): Call it.
1355 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1357 * coverage.c (read_counts_file): Make error message lowercase.
1359 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1361 * profile-count.c (profile_quality_display_names): Fix ordering.
1363 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1366 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1367 hash only when edge is first within the sequence.
1368 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1369 (symbol_table::create_edge): Do not set target_prob.
1370 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1372 (cgraph_edge::make_speculative): Drop target_prob parameter.
1373 (cgraph_edge::speculative_call_info): Remove.
1374 (cgraph_edge::first_speculative_call_target): New member function.
1375 (update_call_stmt_hash_for_removing_direct_edge): New function.
1376 (cgraph_edge::resolve_speculation): Rewrite to new API.
1377 (cgraph_edge::speculative_call_for_target): New member function.
1378 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1379 multiple speculation targets.
1380 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1382 (verify_speculative_call): Verify that targets form an interval.
1383 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1384 (cgraph_edge::first_speculative_call_target): New member function.
1385 (cgraph_edge::next_speculative_call_target): New member function.
1386 (cgraph_edge::speculative_call_target_ref): New member function.
1387 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1388 (cgraph_edge): Remove target_prob.
1389 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1390 Fix handling of speculative calls.
1391 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1392 * ipa-fnsummary.c (analyze_function_body): Likewise.
1393 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1394 * ipa-profile.c (dump_histogram): Fix formating.
1395 (ipa_profile_generate_summary): Watch for overflows.
1396 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1397 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1398 (update_indirect_edges_after_inlining): Update to new API.
1399 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1401 * profile-count.h: (profile_probability::adjusted): New.
1402 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1403 updating of profile.
1404 * value-prof.c (gimple_ic_transform): Rename to ...
1405 (dump_ic_profile): ... this one; update dumping.
1406 (stream_in_histogram_value): Fix formating.
1407 (gimple_value_profile_transformations): Update.
1409 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1412 * config/i386/i386.md (*movoi_internal_avx): Remove
1413 TARGET_SSE_TYPELESS_STORES check.
1414 (*movti_internal): Prefer TARGET_AVX over
1415 TARGET_SSE_TYPELESS_STORES.
1416 (*movtf_internal): Likewise.
1417 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1418 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1419 from TARGET_SSE_TYPELESS_STORES.
1421 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1423 * diagnostic-core.h (warning_at): Rename overload to...
1424 (warning_meta): ...this.
1425 (emit_diagnostic_valist): Delete decl of overload taking
1426 diagnostic_metadata.
1427 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1428 (warning_at): Rename overload taking diagnostic_metadata to...
1429 (warning_meta): ...this.
1431 2020-01-28 Richard Biener <rguenther@suse.de>
1433 PR tree-optimization/93439
1434 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1435 * tree-cfg.c (move_sese_region_to_fn): ... here.
1436 (verify_types_in_gimple_reference): Verify used cliques are
1439 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1442 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1443 argument of a pointer to struct gcc_options and pass it to
1444 parse_mtune_ctrl_str.
1445 (ix86_function_specific_restore): Pass opts to
1446 set_ix86_tune_features.
1447 (ix86_option_override_internal): Likewise.
1448 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1449 gcc_options and use it for x_ix86_tune_ctrl_string.
1451 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1453 PR rtl-optimization/87763
1454 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1455 simplification to handle subregs as well as bare regs.
1456 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1458 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1460 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1461 for reduction chains that (now) include a call.
1463 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1465 PR tree-optimization/92822
1466 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1467 out the don't-care elements of a vector whose significant elements
1468 are duplicates, make the don't-care elements duplicates too.
1470 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1472 PR tree-optimization/93434
1473 * tree-predcom.c (split_data_refs_to_components): Record which
1474 components have had aliasing loads removed. Prevent store-store
1475 commoning for all such components.
1477 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1480 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1481 -1 or is_vshift is true, use new_vector with number of elts npatterns
1482 rather than new_unary_operation.
1484 PR tree-optimization/93454
1485 * gimple-fold.c (fold_array_ctor_reference): Perform
1486 elt_size.to_uhwi () just once, instead of calling it in every
1487 iteration. Punt if that value is above size of the temporary
1488 buffer. Decrease third native_encode_expr argument when
1489 bufoff + elt_sz is above size of buf.
1491 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1493 * config/mips/mips.c (mips_declare_object_name)
1494 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1496 2020-01-27 Martin Liska <mliska@suse.cz>
1498 PR gcov-profile/93403
1499 * tree-profile.c (gimple_init_gcov_profiler): Generate
1500 both __gcov_indirect_call_profiler_v4 and
1501 __gcov_indirect_call_profiler_v4_atomic.
1503 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1506 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1508 (@aarch64_split_simd_mov<mode>): Use it.
1509 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1510 Leave the vec_extract patterns to handle 2-element vectors.
1511 (aarch64_simd_mov_from_<mode>high): Likewise.
1512 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1513 (vec_extractv2dfv1df): Likewise.
1515 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1517 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1518 jump conditions for *compare_condjump<GPI:mode>.
1520 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1523 * digraph.cc (test_edge::test_edge): Specify template for base
1526 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1528 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1530 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1532 * config/arc/arc-protos.h (gen_mlo): Remove.
1533 (gen_mhi): Likewise.
1534 * config/arc/arc.c (AUX_MULHI): Define.
1535 (arc_must_save_reister): Special handling for r58/59.
1536 (arc_compute_frame_size): Consider mlo/mhi registers.
1537 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1539 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1540 mlo/mhi name selection.
1541 (arc_restore_callee_saves): Don't early restore blink when ISR.
1542 (arc_expand_prologue): Add mlo/mhi saving.
1543 (arc_expand_epilogue): Add mlo/mhi restoring.
1546 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1547 numbering when MUL64 option is used.
1548 (DWARF2_FRAME_REG_OUT): Define.
1549 * config/arc/arc.md (arc600_stall): New pattern.
1550 (VUNSPEC_ARC_ARC600_STALL): Define.
1551 (mulsi64): Use correct mlo/mhi registers.
1552 (mulsi_600): Clean it up.
1553 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1555 (mhi_operand): Likewise.
1557 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1558 Petro Karashchenko <petro.karashchenko@ring.com>
1560 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1561 attributes if needed.
1562 (prepare_move_operands): Generate special unspec instruction for
1564 (arc_isuncached_mem_p): Propagate uncached attribute to each
1566 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1567 (VUNSPEC_ARC_STDI): Likewise.
1568 (ALLI): New mode iterator.
1569 (mALLI): New mode attribute.
1570 (lddi): New instruction pattern.
1572 (stdidi_split): Split instruction for architectures which are not
1573 supporting ll64 option.
1574 (lddidi_split): Likewise.
1576 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1578 PR rtl-optimization/92989
1579 * lra-lives.c (process_bb_lives): Update the live-in set before
1580 processing additional clobbers.
1582 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1584 PR rtl-optimization/93170
1585 * cselib.c (cselib_invalidate_regno_val): New function, split out
1587 (cselib_invalidate_regno): ...here.
1588 (cselib_invalidated_by_call_p): New function.
1589 (cselib_process_insn): Iterate over all the hard-register entries in
1590 REG_VALUES and invalidate any that cross call-clobbered registers.
1592 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1594 * dojump.c (split_comparison): Use HONOR_NANS rather than
1595 HONOR_SNANS when splitting LTGT.
1597 2020-01-27 Martin Liska <mliska@suse.cz>
1600 * opts.c (print_filtered_help): Exclude language-specific
1601 options from --help=common unless enabled in all FEs.
1603 2020-01-27 Martin Liska <mliska@suse.cz>
1605 * opts.c (print_help): Exclude params from
1606 all except --help=param.
1608 2020-01-27 Martin Liska <mliska@suse.cz>
1611 * config/i386/i386-features.c (make_resolver_func):
1612 Align the code with ppc64 target implementation.
1613 Do not generate a unique name for resolver function.
1615 2020-01-27 Richard Biener <rguenther@suse.de>
1617 PR tree-optimization/93397
1618 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1619 converted reduction chain SLP graph adjustment.
1621 2020-01-26 Marek Polacek <polacek@redhat.com>
1624 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1627 2020-01-26 Jason Merrill <jason@redhat.com>
1630 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1633 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1635 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1636 (rx_setmem): Likewise.
1638 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1641 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1642 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1643 drop <di> from constraint of last operand.
1646 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1647 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1648 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1650 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1653 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1656 2020-01-24 Jeff Law <law@redhat.com>
1658 PR tree-optimization/92788
1659 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1662 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1665 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1666 *avx_vperm_broadcast_<mode>,
1667 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1668 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1669 Move before avx2_perm<mode>/avx512f_perm<mode>.
1672 * simplify-rtx.c (simplify_const_unary_operation,
1673 simplify_const_binary_operation): Punt for mode precision above
1674 MAX_BITSIZE_MODE_ANY_INT.
1676 2020-01-24 Andrew Pinski <apinski@marvell.com>
1678 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1681 2020-01-24 Jeff Law <law@redhat.com>
1684 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1685 for REGs. Call output_operand_lossage to get more reasonable
1688 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1690 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1691 gcn_fp_compare_operator.
1692 (vec_cmpu<mode>di): Use gcn_compare_operator.
1693 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1694 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1695 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1696 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1697 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1698 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1699 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1700 gcn_fp_compare_operator.
1701 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1702 gcn_fp_compare_operator.
1703 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1704 gcn_fp_compare_operator.
1705 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1706 gcn_fp_compare_operator.
1708 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1710 * doc/install.texi (Cross-Compiler-Specific Options): Document
1711 `--with-toolexeclibdir' option.
1713 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1715 * target.def (flags_regnum): Also mention effect on delay slot filling.
1716 * doc/tm.texi: Regenerate.
1718 2020-01-23 Jeff Law <law@redhat.com>
1720 PR translation/90162
1721 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1723 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1726 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1729 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1731 PR rtl-optimization/93402
1732 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1735 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1737 * config.in: Regenerated.
1738 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1739 for TARGET_LIBC_GNUSTACK.
1740 * configure: Regenerated.
1741 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1742 found to be 2.31 or greater.
1744 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1746 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1748 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1749 (mips_asm_file_end): New function. Delegate to
1750 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1751 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1753 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1756 * config/i386/i386-modes.def (POImode): New mode.
1757 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1758 * config/i386/i386.md (DPWI): New mode attribute.
1759 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1761 (QPWI): ... this. Use POI instead of OI for TImode.
1762 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1763 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1766 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1769 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1771 (speculation_tracker_rev): New pattern.
1772 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1773 Use speculation_tracker_rev to track the inverse condition.
1775 2020-01-23 Richard Biener <rguenther@suse.de>
1777 PR tree-optimization/93381
1778 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1779 alias-set of the def as argument and record the first one.
1780 (vn_walk_cb_data::first_set): New member.
1781 (vn_reference_lookup_3): Pass the alias-set of the current def
1782 to push_partial_def. Fix alias-set used in the aggregate copy
1784 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1785 * real.c (clear_significand_below): Fix out-of-bound access.
1787 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1790 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1791 New define_insn patterns.
1793 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1795 * doc/sourcebuild.texi (check-function-bodies): Add an
1796 optional target/xfail selector.
1798 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1800 PR rtl-optimization/93124
1801 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1802 bare USE and CLOBBER insns.
1804 2020-01-22 Andrew Pinski <apinski@marvell.com>
1806 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1808 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1811 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1812 diagnostic_manager into "ana" namespace.
1813 * selftest-run-tests.c (selftest::run_tests): Update for move of
1814 selftest::run_analyzer_selftests to
1815 ana::selftest::run_analyzer_selftests.
1817 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1819 * cfgexpand.c (union_stack_vars): Update the size.
1821 2020-01-22 Richard Biener <rguenther@suse.de>
1823 PR tree-optimization/93381
1824 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1825 throughout, handle all conversions the same.
1827 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1830 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1831 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1832 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1833 Call force_reg on high_in2 unconditionally.
1835 2020-01-22 Martin Liska <mliska@suse.cz>
1837 PR tree-optimization/92924
1838 * profile.c (compute_value_histograms): Divide
1841 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1844 * output.h (assemble_name_resolve): Declare.
1845 * varasm.c (assemble_name_resolve): New function.
1846 (assemble_name): Use it.
1847 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1849 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1851 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1852 update_web_docs_git instead of update_web_docs_svn.
1854 2020-01-21 Andrew Pinski <apinski@marvell.com>
1857 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1858 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1859 (*tlsgd_small_<mode>): Likewise.
1860 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1861 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1862 register. Convert that register back to dest using convert_mode.
1864 2020-01-21 Jim Wilson <jimw@sifive.com>
1866 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1869 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1870 Uros Bizjak <ubizjak@gmail.com>
1873 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1875 (legitimize_tls_address): Do GNU2 TLS address computation in
1876 ptr_mode and zero-extend result to Pmode.
1877 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1878 :P with :PTR and Pmode with ptr_mode.
1879 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1880 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1881 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1883 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1886 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1887 the last two operands are CONST_INT_P before using them as such.
1889 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1891 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1892 to get the integer element types.
1894 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1896 * config/aarch64/aarch64-sve-builtins.h
1897 (function_expander::convert_to_pmode): Declare.
1898 * config/aarch64/aarch64-sve-builtins.cc
1899 (function_expander::convert_to_pmode): New function.
1900 (function_expander::get_contiguous_base): Use it.
1901 (function_expander::prepare_gather_address_operands): Likewise.
1902 * config/aarch64/aarch64-sve-builtins-sve2.cc
1903 (svwhilerw_svwhilewr_impl::expand): Likewise.
1905 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1908 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1909 cfun->machine->label_is_assembled.
1910 (aarch64_print_patchable_function_entry): New.
1911 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1912 * config/aarch64/aarch64.h (struct machine_function): New field,
1915 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1918 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1921 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1924 * cgraph.c (cgraph_edge::resolve_speculation,
1925 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1926 call_stmt_site_hash.
1928 2020-01-21 Martin Liska <mliska@suse.cz>
1930 * config/rs6000/rs6000.c (common_mode_defined): Remove
1933 2020-01-21 Richard Biener <rguenther@suse.de>
1935 PR tree-optimization/92328
1936 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1937 type when value-numbering same-sized store by inserting a
1939 (eliminate_dom_walker::eliminate_stmt): When eliminating
1940 a redundant store handle bit-reinterpretation of the same value.
1942 2020-01-21 Andrew Pinski <apinski@marvel.com>
1945 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1947 (prepare_block_for_update): This. Use a worklist instead of
1950 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1952 * gcc/config/arm/arm.c (clear_operation_p):
1953 Initialise last_regno, skip first iteration
1954 based on the first_set value and use ints instead
1955 of the unnecessary HOST_WIDE_INTs.
1957 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1960 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1961 compare_mode other than SFmode or DFmode.
1963 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1966 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1967 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1968 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1970 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1972 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1974 2020-01-20 Andrew Pinski <apinski@marvell.com>
1977 * targhooks.c (default_print_patchable_function_entry): Use
1978 output_asm_insn to emit the nop instruction.
1980 2020-01-20 Fangrui Song <maskray@google.com>
1983 * targhooks.c (default_print_patchable_function_entry): Align to
1986 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
1989 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1990 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
1991 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1992 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
1993 (*tls_dynamic_gnu2_lea_64): Renamed to ...
1994 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
1995 Remove the {q} suffix from lea.
1996 (*tls_dynamic_gnu2_call_64): Renamed to ...
1997 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
1998 (*tls_dynamic_gnu2_combine_64): Renamed to ...
1999 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2000 Pass Pmode to gen_tls_dynamic_gnu2_64.
2002 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2004 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2006 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2008 * config/aarch64/aarch64-sve-builtins-base.cc
2009 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2011 2020-01-20 Richard Biener <rguenther@suse.de>
2014 * dwarf2out.c (prune_unused_types): Unconditionally mark
2015 called function DIEs.
2017 2020-01-20 Martin Liska <mliska@suse.cz>
2019 PR tree-optimization/93199
2020 * tree-eh.c (struct leh_state): Add
2021 new field outer_non_cleanup.
2022 (cleanup_is_dead_in): Pass leh_state instead
2023 of eh_region. Add a checking that state->outer_non_cleanup
2024 points to outer non-clean up region.
2025 (lower_try_finally): Record outer_non_cleanup
2027 (lower_catch): Likewise.
2028 (lower_eh_filter): Likewise.
2029 (lower_eh_must_not_throw): Likewise.
2030 (lower_cleanup): Likewise.
2032 2020-01-20 Richard Biener <rguenther@suse.de>
2034 PR tree-optimization/93094
2035 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2036 (vect_transform_loop): Likewise.
2037 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2038 loop_vectorized_call to vect_transform_loop.
2039 * tree-vect-loop.c (vect_transform_loop): Pass down
2040 loop_vectorized_call to vect_loop_versioning.
2041 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2042 the earlier discovered loop_vectorized_call.
2044 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2046 * doc/contribute.texi: Update for SVN -> Git transition.
2047 * doc/install.texi: Likewise.
2049 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2052 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2053 speculative targets.
2054 (verify_speculative_call): New function
2055 (cgraph_node::verify_node): Use it.
2056 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2059 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2062 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2063 (cgraph_edge::make_direct): Remove all indirect targets.
2064 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2065 (cgraph_node::verify_node): Verify that only one call_stmt or
2066 lto_stmt_uid is set.
2067 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2069 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2070 (lto_output_ref): Simplify streaming of stmt.
2071 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2073 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2075 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2076 Mark parameter unused.
2078 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2080 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2082 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2084 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2086 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2088 * Makefile.in: Add coroutine-passes.o.
2089 * builtin-types.def (BT_CONST_SIZE): New.
2090 (BT_FN_BOOL_PTR): New.
2091 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2092 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2093 * coroutine-builtins.def: New file.
2094 * coroutine-passes.cc: New file.
2095 * function.h (struct GTY function): Add a bit to indicate that the
2096 function is a coroutine component.
2097 * internal-fn.c (expand_CO_FRAME): New.
2098 (expand_CO_YIELD): New.
2099 (expand_CO_SUSPN): New.
2100 (expand_CO_ACTOR): New.
2101 * internal-fn.def (CO_ACTOR): New.
2105 * passes.def: Add pass_coroutine_lower_builtins,
2106 pass_coroutine_early_expand_ifns.
2107 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2108 (make_pass_coroutine_early_expand_ifns): New.
2109 * doc/invoke.texi: Document the fcoroutines command line
2112 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2114 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2117 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2118 after checking the argument is a REG. Don't use REGNO (reg)
2119 again to set last_regno, reuse regno variable instead.
2121 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2123 * doc/analyzer.texi (Limitations): Add note about NaN.
2125 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2126 Sudakshina Das <sudi.das@arm.com>
2128 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2129 and valid immediate.
2130 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2131 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2132 * config/arm/constraints.md (Pg): New.
2133 * config/arm/predicates.md (long_shift_imm): New.
2134 (arm_reg_or_long_shift_imm): Likewise.
2135 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2136 (thumb2_lsll): Likewise.
2139 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2140 Sudakshina Das <sudi.das@arm.com>
2142 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2143 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2144 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2145 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2146 * config/arm/thumb2.md (thumb2_asrl): New.
2147 (thumb2_lsll): Likewise.
2149 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2151 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2154 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2156 * gdbinit.in (help-gcc-hooks): New command.
2157 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2158 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2161 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2163 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2164 correct target macro.
2166 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2168 * config/aarch64/aarch64-protos.h
2169 (aarch64_sve_ld1ro_operand_p): New.
2170 * config/aarch64/aarch64-sve-builtins-base.cc
2171 (class load_replicate): New.
2172 (class svld1ro_impl): New.
2173 (class svld1rq_impl): Change to inherit from load_replicate.
2174 (svld1ro): New sve intrinsic function base.
2175 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2176 New DEF_SVE_FUNCTION.
2177 * config/aarch64/aarch64-sve-builtins-base.h
2178 (svld1ro): New decl.
2179 * config/aarch64/aarch64-sve-builtins.cc
2180 (function_expander::add_mem_operand): Modify assert to allow
2182 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2184 * config/aarch64/aarch64.c
2185 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2186 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2187 (aarch64_sve_ld1ro_operand_p): New.
2188 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2189 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2190 * config/aarch64/predicates.md
2191 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2193 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2195 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2196 Introduce this ACLE specified predefined macro.
2197 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2198 (fp): Disabling this disables f64mm.
2199 (simd): Disabling this disables f64mm.
2200 (fp16): Disabling this disables f64mm.
2201 (sve): Disabling this disables f64mm.
2202 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2203 (AARCH64_ISA_F64MM): New.
2204 (TARGET_F64MM): New.
2205 * doc/invoke.texi (f64mm): Document new option.
2207 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2209 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2210 (neoversen1_tunings): Likewise.
2212 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2215 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2216 Add assert to ensure prolog has been emitted.
2217 (aarch64_split_atomic_op): Likewise.
2218 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2219 Use epilogue_completed rather than reload_completed.
2220 (aarch64_atomic_exchange<mode>): Likewise.
2221 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2222 (atomic_nand<mode>): Likewise.
2223 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2224 (atomic_fetch_nand<mode>): Likewise.
2225 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2226 (atomic_nand_fetch<mode>): Likewise.
2228 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2231 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2233 (REVERSE_CONDITION): Delete.
2234 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2235 (CCFP_CCFPE): Likewise.
2236 (e): New mode attribute.
2237 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2238 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2239 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2240 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2241 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2242 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2243 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2244 name of generator from gen_ccmpdi to gen_ccmpccdi.
2245 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2246 the previous comparison but aren't able to, use the new ccmp_rev
2249 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2251 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2252 than testing directly for INTEGER_CST.
2253 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2255 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2257 PR tree-optimization/93292
2258 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2259 get_vectype_for_scalar_type returns NULL.
2261 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2263 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2264 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2266 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2268 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2270 * params.opt: (max-predicted-iterations): Set bounds.
2271 * predict.c (real_almost_one, real_br_prob_base,
2272 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2273 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2274 probabilities; do not truncate to reg_br_prob_bases.
2275 (estimate_loops_at_level): Pass max_cyclic_prob.
2276 (estimate_loops): Compute max_cyclic_prob.
2277 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2279 * profile-count.c (profile_probability::to_sreal): New.
2280 * profile-count.h (class sreal): Move up in file.
2281 (profile_probability::to_sreal): Declare.
2283 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2286 (arm_invalid_conversion): New function for target hook.
2287 (arm_invalid_unary_op): New function for target hook.
2288 (arm_invalid_binary_op): New function for target hook.
2290 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2292 * config.gcc: Add arm_bf16.h.
2293 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2294 (arm_simd_builtin_std_type): Add BFmode.
2295 (arm_init_simd_builtin_types): Define element types for vector types.
2296 (arm_init_bf16_types): New function.
2297 (arm_init_builtins): Add arm_init_bf16_types function call.
2298 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2299 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2300 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2301 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2302 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2303 (arm_mangle_type): Add __bf16.
2304 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2305 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2306 arm_bf16_ptr_type_node.
2307 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2308 define_split between ARM registers.
2309 * config/arm/arm_bf16.h: New file.
2310 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2311 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2313 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2314 * config/arm/vfp.md: Add BFmode to movhf patterns.
2316 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2317 Andre Vieira <andre.simoesdiasvieira@arm.com>
2319 * config/arm/arm-cpus.in (mve, mve_float): New features.
2320 (dsp, mve, mve.fp): New options.
2321 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2322 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2323 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2325 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2326 Thomas Preud'homme <thomas.preudhomme@arm.com>
2328 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2330 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2331 error for using -mcmse when targeting Armv8.1-M Mainline.
2333 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2334 Thomas Preud'homme <thomas.preudhomme@arm.com>
2336 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2337 address in r4 when targeting Armv8.1-M Mainline.
2338 (nonsecure_call_value_internal): Likewise.
2339 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2340 a register match_operand again. Emit BLXNS when targeting
2342 (nonsecure_call_value_reg_thumb2): Likewise.
2344 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2345 Thomas Preud'homme <thomas.preudhomme@arm.com>
2347 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2348 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2349 variable as true when floating-point ABI is not hard. Replace
2350 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2351 Generate VLSTM and VLLDM instruction respectively before and
2352 after a function call to cmse_nonsecure_call function.
2353 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2354 (VUNSPEC_VLLDM): Likewise.
2355 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2356 (lazy_load_multiple_insn): Likewise.
2358 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2359 Thomas Preud'homme <thomas.preudhomme@arm.com>
2361 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2362 (arm_emit_vfp_multi_reg_pop): Likewise.
2363 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2364 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2365 restore callee-saved VFP registers.
2367 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2368 Thomas Preud'homme <thomas.preudhomme@arm.com>
2370 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2371 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2372 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2373 callee-saved GPRs as well as clear ip register before doing a nonsecure
2374 call then restore callee-saved GPRs after it when targeting
2376 (arm_reorg): Adapt to function rename.
2378 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2379 Thomas Preud'homme <thomas.preudhomme@arm.com>
2381 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2382 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2383 clear_vfp_multiple pattern based on a new vfp parameter.
2384 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2385 targeting Armv8.1-M Mainline.
2386 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2387 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2388 whether VFP registers are available before looking call_used_regs for a
2390 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2391 of prototype of clear_operation_p.
2392 (clear_vfp_multiple_operation): New predicate.
2393 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2394 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2396 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2397 Thomas Preud'homme <thomas.preudhomme@arm.com>
2399 * config/arm/arm-protos.h (clear_operation_p): Declare.
2400 * config/arm/arm.c (clear_operation_p): New function.
2401 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2402 targeting Armv8.1-M Mainline or successor.
2403 (output_return_instruction): Only output APSR register clearing if
2404 Armv8.1-M Mainline instructions not available.
2405 (thumb_exit): Likewise.
2406 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2407 * config/arm/thumb2.md (clear_apsr): New define_insn.
2408 (clear_multiple): Likewise.
2409 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2411 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2412 Thomas Preud'homme <thomas.preudhomme@arm.com>
2414 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2415 (use_return_insn): Also return false for Armv8.1-M Mainline.
2416 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2417 Mainline instructions are available.
2418 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2419 when targeting Armv8.1-M Mainline Security Extensions.
2420 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2421 Mainline entry function.
2422 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2423 targeting Armv8.1-M Mainline or successor.
2424 (arm_expand_epilogue): Fix indentation of caller-saved register
2425 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2427 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2428 (FP_SYSREGS): Likewise.
2429 (enum vfp_sysregs_encoding): Define enum.
2430 (fp_sysreg_names): Declare.
2431 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2432 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2433 (pop_fpsysreg_insn): Likewise.
2435 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2436 Thomas Preud'homme <thomas.preudhomme@arm.com>
2438 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2439 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2440 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2441 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2442 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2443 (ARMv8_1m_main): New feature group.
2444 (armv8.1-m.main): New architecture.
2445 * config/arm/arm-tables.opt: Regenerate.
2446 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2447 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2448 (arm_options_perform_arch_sanity_checks): Error out when targeting
2449 Armv8.1-M Mainline Security Extensions.
2450 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2452 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2454 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2455 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2456 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2457 aarch64_bfdot_laneq): New.
2458 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2459 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2460 vbfdotq_laneq_f32): New.
2461 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2462 VBFMLA_W, VBF): New.
2463 (isquadop): Add V4BF, V8BF.
2465 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2467 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2468 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2469 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2470 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2471 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2472 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2473 usdot_laneq, sudot_lane,sudot_laneq): New.
2474 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2475 (aarch64_<sur>dot_lane): New.
2476 * config/aarch64/arm_neon.h (vusdot_s32): New.
2478 (vusdot_lane_s32): New.
2479 (vsudot_lane_s32): New.
2480 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2481 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2483 2020-01-16 Martin Liska <mliska@suse.cz>
2485 * value-prof.c (dump_histogram_value): Fix
2486 obvious spacing issue.
2488 2020-01-16 Andrew Pinski <apinski@marvell.com>
2490 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2491 !storage_order_barrier_p.
2493 2020-01-16 Andrew Pinski <apinski@marvell.com>
2495 * sched-int.h (_dep): Add unused bit-field field for the padding.
2496 * sched-deps.c (init_dep_1): Init unused field.
2498 2020-01-16 Andrew Pinski <apinski@marvell.com>
2500 * optabs.h (create_expand_operand): Initialize target field also.
2502 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2504 PR tree-optimization/92429
2505 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2506 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2508 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2511 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2513 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2514 aarch64_sve_int_mode to each mode.
2516 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2518 * doc/analyzer.texi (Overview): Add note about
2519 -fdump-ipa-analyzer.
2521 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2523 PR tree-optimization/93231
2524 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2525 input_type is unsigned. Use tree_to_shwi for shift constant.
2526 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2527 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2528 inputs more efficiently.
2530 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2532 * config/i386/i386.md (*movsf_internal): Do not require
2533 SSE2 ISA for alternatives 14 and 15.
2535 2020-01-15 Richard Biener <rguenther@suse.de>
2538 * tree-eh.c (sink_clobbers): If we already visited the destination
2539 block do not defer insertion.
2540 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2541 the purpose of defered insertion.
2543 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2545 * BASE-VER: Bump to 10.0.1.
2547 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2549 PR tree-optimization/93247
2550 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2551 type of the stmt that we're going to vectorize.
2553 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2555 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2556 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2559 2020-01-15 Martin Liska <mliska@suse.cz>
2561 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2562 2 calls of streamer_read_hwi in a function call.
2564 2020-01-15 Richard Biener <rguenther@suse.de>
2566 * alias.c (record_alias_subset): Avoid redundant work when
2567 subset is already recorded.
2569 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2571 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2572 the analyzer options provide CWE identifiers.
2574 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2576 * tree-diagnostic-path.cc (path_summary::event_range::print):
2577 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2578 using get_pure_location.
2580 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2582 PR tree-optimization/93262
2583 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2584 perform head trimming only if the last argument is constant,
2585 either all ones, or larger or equal to head trim, in the latter
2586 case decrease the last argument by head_trim.
2588 PR tree-optimization/93249
2589 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2590 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2591 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2592 perform head trim unless we can prove there are no '\0' chars
2593 from the source among the first head_trim chars.
2595 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2597 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2599 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2602 * config/i386/sse.md
2603 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2604 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2605 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2606 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2607 just a single alternative instead of two, make operands 1 and 2
2610 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2613 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2616 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2618 * Makefile.in (lang_opt_files): Add analyzer.opt.
2619 (ANALYZER_OBJS): New.
2620 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2621 tristate.o and ANALYZER_OBJS.
2622 (TEXI_GCCINT_FILES): Add analyzer.texi.
2623 * common.opt (-fanalyzer): New driver option.
2624 * config.in: Regenerate.
2625 * configure: Regenerate.
2626 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2627 (gccdepdir): Also create depdir for "analyzer" subdir.
2628 * digraph.cc: New file.
2629 * digraph.h: New file.
2630 * doc/analyzer.texi: New file.
2631 * doc/gccint.texi ("Static Analyzer") New menu item.
2632 (analyzer.texi): Include it.
2633 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2634 ("Warning Options"): Add static analysis warnings to the list.
2635 (-Wno-analyzer-double-fclose): New option.
2636 (-Wno-analyzer-double-free): New option.
2637 (-Wno-analyzer-exposure-through-output-file): New option.
2638 (-Wno-analyzer-file-leak): New option.
2639 (-Wno-analyzer-free-of-non-heap): New option.
2640 (-Wno-analyzer-malloc-leak): New option.
2641 (-Wno-analyzer-possible-null-argument): New option.
2642 (-Wno-analyzer-possible-null-dereference): New option.
2643 (-Wno-analyzer-null-argument): New option.
2644 (-Wno-analyzer-null-dereference): New option.
2645 (-Wno-analyzer-stale-setjmp-buffer): New option.
2646 (-Wno-analyzer-tainted-array-index): New option.
2647 (-Wno-analyzer-use-after-free): New option.
2648 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2649 (-Wno-analyzer-use-of-uninitialized-value): New option.
2650 (-Wanalyzer-too-complex): New option.
2651 (-fanalyzer-call-summaries): New warning.
2652 (-fanalyzer-checker=): New warning.
2653 (-fanalyzer-fine-grained): New warning.
2654 (-fno-analyzer-state-merge): New warning.
2655 (-fno-analyzer-state-purge): New warning.
2656 (-fanalyzer-transitivity): New warning.
2657 (-fanalyzer-verbose-edges): New warning.
2658 (-fanalyzer-verbose-state-changes): New warning.
2659 (-fanalyzer-verbosity=): New warning.
2660 (-fdump-analyzer): New warning.
2661 (-fdump-analyzer-callgraph): New warning.
2662 (-fdump-analyzer-exploded-graph): New warning.
2663 (-fdump-analyzer-exploded-nodes): New warning.
2664 (-fdump-analyzer-exploded-nodes-2): New warning.
2665 (-fdump-analyzer-exploded-nodes-3): New warning.
2666 (-fdump-analyzer-supergraph): New warning.
2667 * doc/sourcebuild.texi (dg-require-dot): New.
2668 (dg-check-dot): New.
2669 * gdbinit.in (break-on-saved-diagnostic): New command.
2670 * graphviz.cc: New file.
2671 * graphviz.h: New file.
2672 * ordered-hash-map-tests.cc: New file.
2673 * ordered-hash-map.h: New file.
2674 * passes.def (pass_analyzer): Add before
2675 pass_ipa_whole_program_visibility.
2676 * selftest-run-tests.c (selftest::run_tests): Call
2677 selftest::ordered_hash_map_tests_cc_tests.
2678 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2680 * shortest-paths.h: New file.
2681 * timevar.def (TV_ANALYZER): New timevar.
2682 (TV_ANALYZER_SUPERGRAPH): Likewise.
2683 (TV_ANALYZER_STATE_PURGE): Likewise.
2684 (TV_ANALYZER_PLAN): Likewise.
2685 (TV_ANALYZER_SCC): Likewise.
2686 (TV_ANALYZER_WORKLIST): Likewise.
2687 (TV_ANALYZER_DUMP): Likewise.
2688 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2689 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2690 * tree-pass.h (make_pass_analyzer): New decl.
2691 * tristate.cc: New file.
2692 * tristate.h: New file.
2694 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2697 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2698 alternatives 9 and 10.
2700 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2702 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2703 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2704 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2705 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2706 (selftest::hash_map_tests_c_tests): Call it.
2707 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2708 New static constant, using the value of = H::empty_zero_p.
2709 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2710 from default_hash_traits <Value>.
2711 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2713 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2714 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2715 calls to mark_empty with !Descriptor::empty_zero_p.
2716 (hash_table::empty_slow): Conditionalize the memset call with a
2717 check that Descriptor::empty_zero_p; otherwise, loop through the
2718 entries calling mark_empty on them.
2719 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2720 (pointer_hash::empty_zero_p): Likewise.
2721 (pair_hash::empty_zero_p): Likewise.
2722 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2724 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2725 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2726 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2727 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2728 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2729 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2730 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2731 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2733 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2736 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2738 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2739 fix typo on return value.
2741 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2744 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2746 (cgraph_edge::make_speculative): Add param for setting speculative_id
2748 (cgraph_edge::speculative_call_info): Update comments and find reference
2749 by speculative_id for multiple indirect targets.
2750 (cgraph_edge::resolve_speculation): Decrease the speculations
2751 for indirect edge, drop it's speculative if not direct target
2752 left. Update comments.
2753 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2754 (cgraph_node::dump): Print num_speculative_call_targets.
2755 (cgraph_node::verify_node): Don't report error if speculative
2756 edge not include statement.
2757 (cgraph_edge::num_speculative_call_targets_p): New function.
2758 * cgraph.h (int common_target_id): Remove.
2759 (int common_target_probability): Remove.
2760 (num_speculative_call_targets): New variable.
2761 (make_speculative): Add param for setting speculative_id.
2762 (cgraph_edge::num_speculative_call_targets_p): New declare.
2763 (target_prob): New variable.
2764 (speculative_id): New variable.
2765 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2766 call summaries for multiple speculative call targets.
2767 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2768 * ipa-profile.c (struct speculative_call_target): New struct.
2769 (class speculative_call_summary): New class.
2770 (class speculative_call_summaries): New class.
2771 (call_sums): New variable.
2772 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2773 (ipa_profile_write_edge_summary): New function.
2774 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2775 (ipa_profile_dump_all_summaries): New function.
2776 (ipa_profile_read_edge_summary): New function.
2777 (ipa_profile_read_summary_section): New function.
2778 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2779 (ipa_profile): Generate num_speculative_call_targets from
2781 * ipa-ref.h (speculative_id): New variable.
2782 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2783 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2784 common_target_probability. Stream out speculative_id and
2785 num_speculative_call_targets.
2786 (input_edge): Likewise.
2787 * predict.c (dump_prediction): Remove edges count assert to be
2789 * symtab.c (symtab_node::create_reference): Init speculative_id.
2790 (symtab_node::clone_references): Clone speculative_id.
2791 (symtab_node::clone_referring): Clone speculative_id.
2792 (symtab_node::clone_reference): Clone speculative_id.
2793 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2794 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2795 if indirect call contains multiple speculative targets.
2796 * value-prof.h (check_ic_target): Remove.
2797 * value-prof.c (gimple_value_profile_transformations):
2798 Use void function gimple_ic_transform.
2799 * value-prof.c (gimple_ic_transform): Handle topn case.
2800 Fix comment typos. Change it to a void function.
2802 2020-01-13 Andrew Pinski <apinski@marvell.com>
2804 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2805 (octeontx2t98): New define.
2806 (octeontx2t96): New define.
2807 (octeontx2t93): New define.
2808 (octeontx2f95): New define.
2809 (octeontx2f95n): New define.
2810 (octeontx2f95mm): New define.
2811 * config/aarch64/aarch64-tune.md: Regenerate.
2812 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2814 2020-01-13 Jason Merrill <jason@redhat.com>
2816 PR c++/33799 - destroy return value if local cleanup throws.
2817 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2819 2020-01-13 Martin Liska <mliska@suse.cz>
2821 * ipa-cp.c (get_max_overall_size): Use newly
2822 renamed param param_ipa_cp_unit_growth.
2823 * params.opt: Remove legacy param name.
2825 2020-01-13 Martin Sebor <msebor@redhat.com>
2827 PR tree-optimization/93213
2828 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2829 stores to be eliminated.
2831 2020-01-13 Martin Liska <mliska@suse.cz>
2833 * opts.c (print_help): Do not print CL_PARAM
2834 and CL_WARNING for CL_OPTIMIZATION.
2836 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2839 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2840 depending on optimization settings.
2842 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2844 PR tree-optimization/90838
2845 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2846 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2847 argument rather than to initialize temporary for targets that
2848 don't use the mode argument at all. Initialize ctzval to avoid
2851 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2853 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2854 * tree-core.h: Document it.
2855 * gimplify.c (gimplify_omp_workshare): Set it.
2856 * omp-low.c (lower_omp_target): Use it.
2857 * tree-pretty-print.c (dump_omp_clause): Print it.
2859 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2860 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2862 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2864 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2865 * common.opt (fdiagnostics-path-format=): New option.
2866 (diagnostic_path_format): New enum.
2867 (fdiagnostics-show-path-depths): New option.
2868 * coretypes.h (diagnostic_event_id_t): New forward decl.
2869 * diagnostic-color.c (color_dict): Add "path".
2870 * diagnostic-event-id.h: New file.
2871 * diagnostic-format-json.cc (json_from_expanded_location): Make
2873 (json_end_diagnostic): Call context->make_json_for_path if it
2874 exists and the diagnostic has a path.
2875 (diagnostic_output_format_init): Clear context->print_path.
2876 * diagnostic-path.h: New file.
2877 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2878 when printing a run of events in a diagnostic_path so that they
2879 all get the same color.
2880 (layout::m_diagnostic_path_p): New field.
2881 (layout::layout): Initialize it.
2882 (layout::print_any_labels): Don't colorize the label text for an
2883 event in a diagnostic_path.
2884 (gcc_rich_location::add_location_if_nearby): Add
2885 "restrict_to_current_line_spans" and "label" params. Pass the
2886 former to layout.maybe_add_location_range; pass the latter
2887 when calling add_range.
2888 * diagnostic.c: Include "diagnostic-path.h".
2889 (diagnostic_initialize): Initialize context->path_format and
2890 context->show_path_depths.
2891 (diagnostic_show_any_path): New function.
2892 (diagnostic_path::interprocedural_p): New function.
2893 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2894 (simple_diagnostic_path::num_events): New function.
2895 (simple_diagnostic_path::get_event): New function.
2896 (simple_diagnostic_path::add_event): New function.
2897 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2898 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2899 (debug): New overload taking a diagnostic_path *.
2900 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2901 * diagnostic.h (enum diagnostic_path_format): New enum.
2902 (json::value): New forward decl.
2903 (diagnostic_context::path_format): New field.
2904 (diagnostic_context::show_path_depths): New field.
2905 (diagnostic_context::print_path): New callback field.
2906 (diagnostic_context::make_json_for_path): New callback field.
2907 (diagnostic_show_any_path): New decl.
2908 (json_from_expanded_location): New decl.
2909 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2910 (-fdiagnostics-show-path-depths): New option.
2911 (-fdiagnostics-color): Add "path" to description of default
2912 GCC_COLORS; describe it.
2913 (-fdiagnostics-format=json): Document how diagnostic paths are
2914 represented in the JSON output format.
2915 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2916 Add optional params "restrict_to_current_line_spans" and "label".
2917 * opts.c (common_handle_option): Handle
2918 OPT_fdiagnostics_path_format_ and
2919 OPT_fdiagnostics_show_path_depths.
2920 * pretty-print.c: Include "diagnostic-event-id.h".
2921 (pp_format): Implement "%@" format code for printing
2922 diagnostic_event_id_t *.
2923 (selftest::test_pp_format): Add tests for "%@".
2924 * selftest-run-tests.c (selftest::run_tests): Call
2925 selftest::tree_diagnostic_path_cc_tests.
2926 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2927 * toplev.c (general_init): Initialize global_dc->path_format and
2928 global_dc->show_path_depths.
2929 * tree-diagnostic-path.cc: New file.
2930 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2931 non-static. Drop "diagnostic" param in favor of storing the
2932 original value of "where" and re-using it.
2933 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2934 maybe_unwind_expanded_macro_loc.
2935 (tree_diagnostics_defaults): Initialize context->print_path and
2936 context->make_json_for_path.
2937 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2939 (default_tree_make_json_for_path): New decl.
2940 (maybe_unwind_expanded_macro_loc): New decl.
2942 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2944 PR tree-optimization/93210
2945 * fold-const.h (native_encode_initializer,
2946 can_native_interpret_type_p): Declare.
2947 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2949 (native_encode_initializer): New function, moved from dwarf2out.c.
2950 Adjust to native_encode_expr compatible arguments, including dry-run
2951 and partial extraction modes. Don't handle STRING_CST.
2952 (can_native_interpret_type_p): No longer static.
2953 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2954 offset / BITS_PER_UNIT fits into int and don't call it if
2955 can_native_interpret_type_p fails. If suboff is NULL and for
2956 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2957 native_encode_initializer.
2958 (fold_const_aggregate_ref_1): Formatting fix.
2959 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2960 (tree_add_const_value_attribute): Adjust caller.
2962 PR tree-optimization/90838
2963 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2964 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2965 CTZ_DEFINED_VALUE_AT_ZERO.
2967 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2970 * lra-constraints.c (match_reload): Permit input operands have the
2971 same mode as output while other input operands have a different
2974 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
2976 PR tree-optimization/90838
2977 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2978 (check_ctz_string): Likewise.
2979 (optimize_count_trailing_zeroes): Likewise.
2980 (simplify_count_trailing_zeroes): Likewise.
2981 (pass_forwprop::execute): Try ctz simplification.
2982 * match.pd: Add matching for ctz idioms.
2984 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2986 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2988 (aarch64_invalid_unary_op): New function for target hook.
2989 (aarch64_invalid_binary_op): New function for target hook.
2991 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2993 * config.gcc: Add arm_bf16.h.
2994 * config/aarch64/aarch64-builtins.c
2995 (aarch64_simd_builtin_std_type): Add BFmode.
2996 (aarch64_init_simd_builtin_types): Define element types for vector
2998 (aarch64_init_bf16_types): New function.
2999 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3000 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3002 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3003 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3005 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3006 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3007 * config/aarch64/aarch64.c
3008 (aarch64_classify_vector_mode): Add support for BF types.
3009 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3010 (aarch64_vq_mode): Add support for BF types.
3011 (aarch64_simd_container_mode): Add support for BF types.
3012 (aarch64_mangle_type): Add support for BF scalar type.
3013 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3014 * config/aarch64/arm_bf16.h: New file.
3015 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3016 * config/aarch64/iterators.md: Add BF types to mode attributes.
3017 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3019 2020-01-10 Jason Merrill <jason@redhat.com>
3021 PR c++/93173 - incorrect tree sharing.
3022 * gimplify.c (copy_if_shared): No longer static.
3023 * gimplify.h: Declare it.
3025 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3027 * doc/invoke.texi (-msve-vector-bits=): Document that
3028 -msve-vector-bits=128 now generates VL-specific code for
3029 little-endian targets.
3030 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3031 build_vector_type_for_mode to construct the data vector types.
3032 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3033 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3034 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3035 for 128-bit vectors.
3037 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3039 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3042 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3044 * config/aarch64/aarch64-builtins.c
3045 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3046 rather than checking the number of elements and the element mode.
3048 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3050 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3051 get_related_vectype_for_scalar_type rather than build_vector_type
3052 to create the index type for a conditional reduction.
3054 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3056 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3057 for any type of gather or scatter, including strided accesses.
3059 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3061 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3064 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3066 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3068 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3069 parameter and its use to reset DR_OFFSET's.
3070 (vect_transform_loop): Remove orig_drs_init argument.
3071 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3072 member of dr_vec_info rather than the offset of the associated
3073 data_reference's innermost_loop_behavior.
3074 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3075 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3076 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3077 get_dr_vinfo_offset.
3078 (vectorizable_store): Likewise.
3079 (vectorizable_load): Likewise.
3081 2020-01-10 Richard Biener <rguenther@suse.de>
3083 * gimple-ssa-store-merging
3084 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3086 2020-01-10 Martin Liska <mliska@suse.cz>
3089 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3090 encapsulation that was there before r280040.
3092 2020-01-10 Richard Biener <rguenther@suse.de>
3095 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3096 sequences to avoid walking them again for secondary opportunities.
3097 (pass_lower_eh_dispatch::execute): Instead actually insert
3100 2020-01-10 Richard Biener <rguenther@suse.de>
3103 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3104 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3105 avoid quadraticness.
3107 2020-01-10 Martin Jambor <mjambor@suse.cz>
3109 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3110 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3111 to get param_ipa_sra_max_replacements.
3112 (param_splitting_across_edge): Pass the caller to
3113 pull_accesses_from_callee.
3115 2020-01-10 Martin Jambor <mjambor@suse.cz>
3117 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3118 * ipa-cp.c (max_new_size): Removed.
3119 (orig_overall_size): New variable.
3120 (get_max_overall_size): New function.
3121 (estimate_local_effects): Use it. Adjust dump.
3122 (decide_about_value): Likewise.
3123 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3124 orig_overall_size. Adjust dump.
3125 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3127 2020-01-10 Martin Jambor <mjambor@suse.cz>
3129 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3130 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3131 instead of param_ipa_max_agg_items.
3132 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3133 optimization info for the callee.
3135 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3137 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3138 markers if debug_inline_points is false.
3140 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3142 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3144 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3145 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3146 aarch64-sve-builtins-sve2.h.
3147 (aarch64-sve-builtins-sve2.o): New rule.
3148 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3149 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3150 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3151 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3152 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3153 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3155 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3156 instructions that are handled here.
3157 (@cond_asrd<mode>): Generalize to...
3158 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3159 (*cond_asrd<mode>_2): Generalize to...
3160 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3161 (*cond_asrd<mode>_z): Generalize to...
3162 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3163 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3164 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3165 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3166 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3168 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3169 (@aarch64_scatter_stnt<mode>): Likewise.
3170 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3171 (@aarch64_mul_lane_<mode>): Likewise.
3172 (@aarch64_sve_suqadd<mode>_const): Likewise.
3173 (*<sur>h<addsub><mode>): Generalize to...
3174 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3176 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3177 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3178 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3179 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3180 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3181 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3182 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3183 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3184 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3185 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3186 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3187 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3188 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3189 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3190 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3191 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3192 (@aarch64_sve2_xar<mode>): Likewise.
3193 (@aarch64_sve2_bcax<mode>): Likewise.
3194 (*aarch64_sve2_eor3<mode>): Rename to...
3195 (@aarch64_sve2_eor3<mode>): ...this.
3196 (@aarch64_sve2_bsl<mode>): New expander.
3197 (@aarch64_sve2_nbsl<mode>): Likewise.
3198 (@aarch64_sve2_bsl1n<mode>): Likewise.
3199 (@aarch64_sve2_bsl2n<mode>): Likewise.
3200 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3201 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3202 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3203 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3204 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3205 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3206 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3207 (<su>mull<bt><Vwide>): Generalize to...
3208 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3210 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3211 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3212 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3213 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3214 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3215 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3216 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3217 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3218 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3219 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3220 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3221 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3222 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3223 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3224 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3225 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3226 (<SHRNB:r>shrnb<mode>): Generalize to...
3227 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3229 (<SHRNT:r>shrnt<mode>): Generalize to...
3230 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3232 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3233 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3234 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3235 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3236 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3237 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3238 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3239 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3240 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3241 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3242 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3243 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3244 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3245 (@aarch64_sve2_cvtnt<mode>): Likewise.
3246 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3247 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3248 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3249 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3250 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3251 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3252 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3253 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3254 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3255 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3256 (@aarch64_sve2_pmul<mode>): Likewise.
3257 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3258 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3259 (@aarch64_sve2_tbl2<mode>): Likewise.
3260 (@aarch64_sve2_tbx<mode>): Likewise.
3261 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3262 (@aarch64_sve2_histcnt<mode>): Likewise.
3263 (@aarch64_sve2_histseg<mode>): Likewise.
3264 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3265 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3266 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3267 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3268 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3269 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3270 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3271 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3272 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3273 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3274 (SVE2_PMULL_PAIR_I): New mode iterators.
3275 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3276 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3277 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3278 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3279 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3280 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3281 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3282 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3283 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3284 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3285 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3286 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3287 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3288 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3289 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3290 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3291 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3292 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3293 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3294 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3295 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3296 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3297 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3298 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3299 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3300 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3301 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3302 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3303 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3304 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3305 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3307 (VNARROW, Ventype): New mode attributes.
3308 (Vewtype): Handle VNx2DI. Fix typo in comment.
3309 (VDOUBLE): New mode attribute.
3310 (sve_lane_con): Handle VNx8HI.
3311 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3312 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3313 (sve_int_op, sve_int_op_rev): Handle the above codes.
3314 (sve_pred_int_rhs2_operand): Likewise.
3315 (MULLBT, SHRNB, SHRNT): Delete.
3316 (SVE_INT_SHIFT_IMM): New int iterator.
3317 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3318 and UNSPEC_WHILEHS for TARGET_SVE2.
3319 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3320 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3321 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3322 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3323 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3324 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3325 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3326 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3327 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3328 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3329 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3330 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3331 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3332 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3333 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3334 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3335 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3336 (optab): Handle the new unspecs.
3337 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3339 (lr): Handle the new unspecs.
3341 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3342 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3343 (sve_int_qsub_op): New int attributes.
3344 (sve_fp_op, rot): Handle the new unspecs.
3345 * config/aarch64/aarch64-sve-builtins.h
3346 (function_resolver::require_matching_pointer_type): Declare.
3347 (function_resolver::resolve_unary): Add an optional boolean argument.
3348 (function_resolver::finish_opt_n_resolution): Add an optional
3349 type_suffix_index argument.
3350 (gimple_folder::redirect_call): Declare.
3351 (gimple_expander::prepare_gather_address_operands): Add an optional
3353 * config/aarch64/aarch64-sve-builtins.cc: Include
3354 aarch64-sve-builtins-sve2.h.
3355 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3356 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3357 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3358 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3359 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3360 (TYPES_s_unsigned): Likewise.
3361 (TYPES_s_integer): Use TYPES_s_unsigned.
3362 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3363 (TYPES_sd_integer): Use them.
3364 (TYPES_d_unsigned): New macro.
3365 (TYPES_d_integer): Use it.
3366 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3367 (TYPES_cvt_narrow): Likewise.
3368 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3369 (preds_mx): New variable.
3370 (function_builder::add_overloaded_function): Allow the new feature
3371 set to be more restrictive than the original one.
3372 (function_resolver::infer_pointer_type): Remove qualifiers from
3373 the pointer type before printing it.
3374 (function_resolver::require_matching_pointer_type): New function.
3375 (function_resolver::resolve_sv_displacement): Handle functions
3376 that don't support 32-bit vector indices or svint32_t vector offsets.
3377 (function_resolver::finish_opt_n_resolution): Take the inferred type
3378 as a separate argument.
3379 (function_resolver::resolve_unary): Optionally treat all forms in
3380 the same way as normal merging functions.
3381 (gimple_folder::redirect_call): New function.
3382 (function_expander::prepare_gather_address_operands): Add an argument
3383 that says whether scaled forms are available. If they aren't,
3384 handle scaling of vector indices and don't add the extension and
3386 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3387 fall back to using cond_* instead.
3388 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3389 Split out the member variables into...
3390 (rtx_code_function_base): ...this new base class.
3391 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3392 (unspec_based_function): Split out the member variables into...
3393 (unspec_based_function_base): ...this new base class.
3394 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3395 (unspec_based_function_exact_insn): New class.
3396 (unspec_based_add_function, unspec_based_add_lane_function)
3397 (unspec_based_lane_function, unspec_based_pred_function)
3398 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3399 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3400 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3402 (unspec_based_fused_function): New class.
3403 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3404 (unspec_based_fused_lane_function): New class.
3405 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3407 (CODE_FOR_MODE1): New macro.
3408 (fixed_insn_function): New class.
3409 (while_comparison): Likewise.
3410 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3411 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3412 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3413 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3414 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3415 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3416 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3417 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3418 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3419 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3420 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3421 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3422 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3423 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3424 Also add an initial argument for unary_convert_narrowt, regardless
3425 of the predication type.
3426 (build_32_64): Allow loads and stores to specify MODE_none.
3427 (build_sv_index64, build_sv_uint_offset): New functions.
3428 (long_type_suffix): New function.
3429 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3430 (binary_imm_long_base, load_gather_sv_base): Likewise.
3431 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3432 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3433 (unary_narrowb_base, unary_narrowt_base): Likewise.
3434 (binary_long_lane_def, binary_long_lane): New shape.
3435 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3436 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3437 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3438 (binary_to_uint_def, binary_to_uint): Likewise.
3439 (binary_wide_def, binary_wide): Likewise.
3440 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3441 (compare_def, compare): Likewise.
3442 (compare_ptr_def, compare_ptr): Likewise.
3443 (load_ext_gather_index_restricted_def,
3444 load_ext_gather_index_restricted): Likewise.
3445 (load_ext_gather_offset_restricted_def,
3446 load_ext_gather_offset_restricted): Likewise.
3447 (load_gather_sv_def): Inherit from load_gather_sv_base.
3448 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3449 (shift_left_imm_def, shift_left_imm): Likewise.
3450 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3451 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3452 (store_scatter_index_restricted_def,
3453 store_scatter_index_restricted): Likewise.
3454 (store_scatter_offset_restricted_def,
3455 store_scatter_offset_restricted): Likewise.
3456 (tbl_tuple_def, tbl_tuple): Likewise.
3457 (ternary_long_lane_def, ternary_long_lane): Likewise.
3458 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3459 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3460 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3461 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3462 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3463 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3464 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3465 (ternary_uint_def, ternary_uint): Likewise.
3466 (unary_convert): Fix typo in comment.
3467 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3468 (unary_long_def, unary_long): Likewise.
3469 (unary_narrowb_def, unary_narrowb): Likewise.
3470 (unary_narrowt_def, unary_narrowt): Likewise.
3471 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3472 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3473 (unary_to_int_def, unary_to_int): Likewise.
3474 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3475 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3476 (svasrd_impl): Delete.
3477 (svcadd_impl::expand): Handle integer operations too.
3478 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3479 new functions to derive the unspec numbers.
3480 (svmla_svmls_lane_impl): Replace with...
3481 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3482 integer operations too.
3483 (svwhile_impl): Rename to...
3484 (svwhilelx_impl): ...this and inherit from while_comparison.
3485 (svasrd): Use unspec_based_function.
3486 (svmla_lane): Use svmla_lane_impl.
3487 (svmls_lane): Use svmls_lane_impl.
3488 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3489 (svwhilele, svwhilelt): Use svwhilelx_impl.
3490 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3491 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3492 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3493 * config/aarch64/aarch64-sve-builtins.def: Include
3494 aarch64-sve-builtins-sve2.def.
3496 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3498 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3499 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3500 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3501 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3502 immediates as well as vector ones.
3503 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3504 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3505 (aarch64_sve_qsub_immediate): Update calls accordingly.
3507 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3509 * config/aarch64/aarch64-sve2.md: Add banner comments.
3510 (<su>mulh<r>s<mode>3): Move further up file.
3511 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3512 (*aarch64_sve2_sra<mode>): Move further down file.
3513 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3515 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3517 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3519 (while_optab_cmp): Handle them.
3520 * config/aarch64/aarch64-sve.md
3521 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3522 and add a "@" marker.
3523 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3524 instead of gen_aarch64_sve2_while_ptest.
3525 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3527 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3529 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3530 (UNSPEC_WHILELE): ...this.
3531 (UNSPEC_WHILE_LO): Rename to...
3532 (UNSPEC_WHILELO): ...this.
3533 (UNSPEC_WHILE_LS): Rename to...
3534 (UNSPEC_WHILELS): ...this.
3535 (UNSPEC_WHILE_LT): Rename to...
3536 (UNSPEC_WHILELT): ...this.
3537 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3538 (cmp_op, while_optab_cmp): Likewise.
3539 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3540 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3541 (svwhilelt): Likewise.
3543 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3545 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3546 (unary_to_uint): Define.
3547 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3548 (unary_count): Rename to...
3549 (unary_to_uint_def, unary_to_uint): ...this.
3550 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3552 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3554 * config/aarch64/aarch64-sve-builtins-functions.h
3555 (code_for_mode_function): New class.
3556 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3557 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3558 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3559 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3560 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3562 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3564 * config/aarch64/iterators.md (addsub): New code attribute.
3565 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3567 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3568 in the asm string and attributes. Fix indentation.
3569 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3571 (@aarch64_sve_<optab><mode>): ...this.
3572 * config/aarch64/aarch64-sve-builtins.h
3573 (function_expander::expand_signed_unpred_op): Delete.
3574 * config/aarch64/aarch64-sve-builtins.cc
3575 (function_expander::expand_signed_unpred_op): Likewise.
3576 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3577 try using code_for_aarch64_sve instead.
3578 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3579 (svqsub_impl): Likewise.
3580 (svqadd, svqsub): Use rtx_code_function instead.
3582 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3584 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3585 (HADDSUB, sur, addsub): Remove them.
3587 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3589 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3590 functions the same way, rather than singling out those that
3591 aren't mapped directly to optabs.
3593 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3595 * target.def (compatible_vector_types_p): New target hook.
3596 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3597 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3598 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3599 * doc/tm.texi: Regenerate.
3600 * gimple-expr.c: Include target.h.
3601 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3602 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3604 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3605 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3606 Use the original predicate if it already has a suitable type.
3608 2020-01-09 Martin Jambor <mjambor@suse.cz>
3610 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3611 resolve_speculation and redirect_call_stmt_to_callee static. Change
3612 return type of set_call_stmt to cgraph_edge *.
3613 * auto-profile.c (afdo_indirect_call): Adjust call to
3614 redirect_call_stmt_to_callee.
3615 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3616 make the this pointer explicit, adjust self-recursive calls and the
3617 call top make_direct. Return the resulting edge.
3618 (cgraph_edge::remove): Make this pointer explicit.
3619 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3620 (cgraph_edge::make_direct): Likewise, adjust call to
3621 resolve_speculation.
3622 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3623 call to set_call_stmt.
3624 (cgraph_update_edges_for_call_stmt_node): Update call to
3625 set_call_stmt and remove.
3626 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3627 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3628 (cgraph_node::create_edge_including_clones): Moved "first" definition
3629 of edge to the block where it was used. Adjusted calls to
3631 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3632 cgraph_edge::remove.
3633 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3634 make_direct and redirect_call_stmt_to_callee.
3635 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3636 resolve_speculation and make_direct.
3637 * ipa-inline-transform.c (inline_transform): Adjust call to
3638 redirect_call_stmt_to_callee.
3639 (check_speculations_1):: Adjust call to resolve_speculation.
3640 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3641 resolve-speculation.
3642 (inline_small_functions): Adjust call to resolve_speculation.
3643 (ipa_inline): Likewise.
3644 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3646 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3647 safe with regards to edge removal, adjust calls to
3648 redirect_call_stmt_to_callee.
3649 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3650 and redirect_call_stmt_to_callee.
3651 * multiple_target.c (create_dispatcher_calls): Adjust call to
3652 redirect_call_stmt_to_callee
3653 (redirect_to_specific_clone): Likewise.
3654 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3655 Adjust calls to cgraph_edge::remove.
3656 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3657 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3658 (expand_call_inline): Adjust call to cgraph_edge::remove.
3660 2020-01-09 Martin Liska <mliska@suse.cz>
3662 * params.opt: Set Optimization for
3663 param_max_speculative_devirt_maydefs.
3665 2020-01-09 Martin Sebor <msebor@redhat.com>
3669 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3671 2020-01-09 Martin Liska <mliska@suse.cz>
3673 * auto-profile.c (auto_profile): Use opt_for_fn
3675 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3676 (propagate_vals_across_arith_jfunc): Likewise.
3677 (hint_time_bonus): Likewise.
3678 (incorporate_penalties): Likewise.
3679 (good_cloning_opportunity_p): Likewise.
3680 (perform_estimation_of_a_value): Likewise.
3681 (estimate_local_effects): Likewise.
3682 (ipcp_propagate_stage): Likewise.
3683 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3684 (set_switch_stmt_execution_predicate): Likewise.
3685 (analyze_function_body): Likewise.
3686 * ipa-inline-analysis.c (offline_size): Likewise.
3687 * ipa-inline.c (early_inliner): Likewise.
3688 * ipa-prop.c (ipa_analyze_node): Likewise.
3689 (ipcp_transform_function): Likewise.
3690 * ipa-sra.c (process_scan_results): Likewise.
3691 (ipa_sra_summarize_function): Likewise.
3692 * params.opt: Rename ipcp-unit-growth to
3693 ipa-cp-unit-growth. Add Optimization for various
3694 IPA-related parameters.
3696 2020-01-09 Richard Biener <rguenther@suse.de>
3699 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3701 2020-01-09 Richard Biener <rguenther@suse.de>
3703 PR tree-optimization/93040
3704 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3706 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3708 * common/config/avr/avr-common.c (avr_option_optimization_table)
3709 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3711 2020-01-09 Martin Liska <mliska@suse.cz>
3713 * cgraphclones.c (symbol_table::materialize_all_clones):
3714 Use cgraph_node::dump_name.
3716 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3719 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3720 output_operand_lossage instead of gcc_unreachable.
3721 * doc/md.texi (riscv f constraint): Fix typo.
3724 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3725 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3726 CONST_SCALAR_INT_P instead of CONST_INT_P.
3727 (*subv<mode>4_1): Rename to ...
3728 (subv<mode>4_1): ... this.
3729 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3730 define_insn_and_split patterns.
3731 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3734 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3736 * vec.c (class selftest::count_dtor): New class.
3737 (selftest::test_auto_delete_vec): New test.
3738 (selftest::vec_c_tests): Call it.
3739 * vec.h (class auto_delete_vec): New class template.
3740 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3742 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3744 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3746 2020-01-08 Jim Wilson <jimw@sifive.com>
3748 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3749 use of TLS_MODEL_LOCAL_EXEC when not pic.
3751 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3753 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3756 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3759 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3760 *stack_protect_set_3 peephole2): Also check that the second
3761 insns source is general_operand.
3764 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3765 predicate for output operand instead of register_operand.
3766 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3767 memory destination and non-memory operands[2].
3769 2020-01-08 Martin Liska <mliska@suse.cz>
3771 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3772 ::dump_asm_name instead of (::name or ::asm_name).
3773 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3774 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3775 (analyze_functions): Likewise.
3776 (expand_all_functions): Likewise.
3777 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3778 (propagate_bits_across_jump_function): Likewise.
3779 (dump_profile_updates): Likewise.
3780 (ipcp_store_bits_results): Likewise.
3781 (ipcp_store_vr_results): Likewise.
3782 * ipa-devirt.c (dump_targets): Likewise.
3783 * ipa-fnsummary.c (analyze_function_body): Likewise.
3784 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3785 (process_hsa_functions): Likewise.
3786 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3787 (set_alias_uids): Likewise.
3788 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3789 * ipa-inline.c (recursive_inlining): Likewise.
3790 (inline_to_all_callers_1): Likewise.
3791 (ipa_inline): Likewise.
3792 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3793 (ipa_propagate_frequency): Likewise.
3794 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3795 (remove_described_reference): Likewise.
3796 * ipa-pure-const.c (worse_state): Likewise.
3797 (check_retval_uses): Likewise.
3798 (analyze_function): Likewise.
3799 (propagate_pure_const): Likewise.
3800 (propagate_nothrow): Likewise.
3801 (dump_malloc_lattice): Likewise.
3802 (propagate_malloc): Likewise.
3803 (pass_local_pure_const::execute): Likewise.
3804 * ipa-visibility.c (optimize_weakref): Likewise.
3805 (function_and_variable_visibility): Likewise.
3806 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3807 (ipa_discover_variable_flags): Likewise.
3808 * lto-streamer-out.c (output_function): Likewise.
3809 (output_constructor): Likewise.
3810 * tree-inline.c (copy_bb): Likewise.
3811 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3812 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3814 2020-01-08 Richard Biener <rguenther@suse.de>
3817 * tree-eh.c (sink_clobbers): Update virtual operands for
3818 the first and last stmt only. Add a dry-run capability.
3819 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3820 after CFG manipulations and in RPO order to catch all
3821 secondary opportunities reliably.
3823 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3826 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3828 2019-01-08 Richard Biener <rguenther@suse.de>
3831 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3832 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3833 virtual operand, also updating SSA use.
3834 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3835 Update stmt after resetting virtual operand.
3836 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3837 * gimple-iterator.c (gsi_remove): When not removing the stmt
3838 permanently do not delink immediate uses or mark the stmt modified.
3840 2020-01-08 Martin Liska <mliska@suse.cz>
3842 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3843 (ipa_call_context::estimate_size_and_time): Likewise.
3844 (inline_analyze_function): Likewise.
3846 2020-01-08 Martin Liska <mliska@suse.cz>
3848 * cgraph.c (cgraph_node::dump): Use systematically
3851 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3853 Add -nodevicespecs option for avr.
3856 * config/avr/avr.opt (-nodevicespecs): New driver option.
3857 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3858 "-specs=device-specs/..." if that option is not set.
3859 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3861 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3863 Implement 64-bit double functions for avr.
3866 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3867 --with-double-comparison.
3868 * doc/install.texi: Document them.
3869 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3870 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3871 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3872 * doc/invoke.texi (AVR Built-in Macros): Document them.
3873 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3874 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3875 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3877 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3880 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3881 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3882 when only building rm-profile multilibs.
3884 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3887 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3888 lattice for a value to check.
3889 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3890 finite propagation in self-recursive scc.
3892 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3894 * ipa-inline.c (caller_growth_limits): Restore the AND.
3896 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3898 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3899 (VEC_ALLREG_ALT): New iterator.
3900 (VEC_ALLREG_INT_MODE): New iterator.
3901 (VCMP_MODE): New iterator.
3902 (VCMP_MODE_INT): New iterator.
3903 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3904 (vec_cmp<u>v64qidi): New define_expand.
3905 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3906 (vec_cmpu<mode>di_exec): New define_expand.
3907 (vec_cmp<u>v64qidi_exec): New define_expand.
3908 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3909 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3910 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3911 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3912 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3913 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3914 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3915 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3916 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3917 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3919 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3920 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3922 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3924 * config/gcn/constraints.md (DA): Update description and match.
3926 (Db): New constraint.
3927 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3929 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3930 Implement 'Db' mixed immediate type.
3931 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3932 (addcv64si3_dup<exec_vcc>): Delete.
3933 (subcv64si3<exec_vcc>): Rework constraints.
3934 (addv64di3): Rework constraints.
3935 (addv64di3_exec): Rework constraints.
3936 (subv64di3): Rework constraints.
3937 (addv64di3_dup): Delete.
3938 (addv64di3_dup_exec): Delete.
3939 (addv64di3_zext): Rework constraints.
3940 (addv64di3_zext_exec): Rework constraints.
3941 (addv64di3_zext_dup): Rework constraints.
3942 (addv64di3_zext_dup_exec): Rework constraints.
3943 (addv64di3_zext_dup2): Rework constraints.
3944 (addv64di3_zext_dup2_exec): Rework constraints.
3945 (addv64di3_sext_dup2): Rework constraints.
3946 (addv64di3_sext_dup2_exec): Rework constraints.
3948 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3950 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3951 existing target checks.
3953 2020-01-07 Richard Biener <rguenther@suse.de>
3955 * doc/install.texi: Bump minimal supported MPC version.
3957 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3959 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3960 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3961 * langhooks.c: Include stor-layout.h.
3962 (lhd_simulate_enum_decl): New function.
3963 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3964 handle_arm_sve_h for the LTO frontend.
3965 (register_vector_type): Cope with null returns from pushdecl.
3967 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3969 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3970 (aarch64_sve::nvectors_if_data_type): Replace with...
3971 (aarch64_sve::builtin_type_p): ...this.
3972 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3973 (find_vector_type): Delete.
3974 (add_sve_type_attribute): New function.
3975 (lookup_sve_type_attribute): Likewise.
3976 (register_builtin_types): Add an "SVE type" attribute to each type.
3977 (register_tuple_type): Likewise.
3978 (svbool_type_p, nvectors_if_data_type): Delete.
3979 (mangle_builtin_type): Use lookup_sve_type_attribute.
3980 (builtin_type_p): Likewise. Add an overload that returns the
3981 number of constituent vector and predicate registers.
3982 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3983 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3984 instead of aarch64_sve_argument_p.
3985 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3986 (aarch64_pass_by_reference): Likewise.
3987 (aarch64_function_value_1): Likewise.
3988 (aarch64_return_in_memory): Likewise.
3989 (aarch64_layout_arg): Likewise.
3991 2020-01-07 Jakub Jelinek <jakub@redhat.com>
3993 PR tree-optimization/93156
3994 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3995 least significant bit is always clear.
3997 PR tree-optimization/93118
3998 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
3999 simplifier with two intermediate conversions.
4001 2020-01-07 Martin Liska <mliska@suse.cz>
4003 * params.opt: Add Optimization for various parameters.
4005 2020-01-07 Martin Liska <mliska@suse.cz>
4008 * doc/extend.texi: Explain cloning for target_clone
4011 2020-01-07 Martin Liska <mliska@suse.cz>
4013 PR tree-optimization/92860
4014 * common.opt: Make in Optimization option
4015 as it is affected by -O0, which is an Optimization
4017 * tree-inline.c (tree_inlinable_function_p):
4018 Use opt_for_fn for warn_inline.
4019 (expand_call_inline): Likewise.
4021 2020-01-07 Martin Liska <mliska@suse.cz>
4023 PR tree-optimization/92860
4024 * common.opt: Make flag_ree as optimization
4027 2020-01-07 Martin Liska <mliska@suse.cz>
4029 PR optimization/92860
4030 * params.opt: Mark param_min_crossjump_insns with Optimization
4033 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4035 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4036 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4038 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4040 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4041 helper function to return the valid addressing formats for a given
4042 hard register and mode.
4043 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4045 * config/rs6000/constraints.md (Q constraint): Update
4047 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4050 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4051 Use 'Q' for doing vector extract from memory.
4052 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4054 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4055 doing vector extract from memory.
4056 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4057 extract from memory.
4059 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4060 for the offset being 34-bits when -mcpu=future is used.
4062 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4064 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4065 instead of cmpib_comparison_operator in cmpib patterns.
4066 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4067 of cmpib_comparison_operator. Revise comment.
4069 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4071 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4072 in an IFN_DIV_POW2 node to be equal.
4074 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4076 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4077 (vect_check_scalar_mask): ...this.
4078 (vectorizable_store, vectorizable_load): Update call accordingly.
4079 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4080 argument in calls to conditional internal functions.
4082 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4084 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4085 '0' matching inputs.
4086 (subv64di3_exec): Likewise.
4088 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4090 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4091 * doc/md.texi (movstr): Likewise.
4093 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4095 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4098 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4100 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4102 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4103 to a temporary file and use move-if-change to update the real
4104 file where necessary.
4106 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4108 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4109 rather than Upa for CPY /M.
4111 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4113 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4116 2020-01-06 Martin Liska <mliska@suse.cz>
4118 PR tree-optimization/92860
4119 * params.opt: Mark param_max_combine_insns with Optimization
4122 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4125 * config/i386/i386.md (SWIDWI): New mode iterator.
4126 (DWI, dwi): Add TImode variants.
4127 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4128 <general_hilo_operand> instead of <general_operand>. Use
4129 CONST_SCALAR_INT_P instead of CONST_INT_P.
4130 (*addv<mode>4_1): Rename to ...
4131 (addv<mode>4_1): ... this.
4132 (QWI): New mode attribute.
4133 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4134 define_insn_and_split patterns.
4135 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4137 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4138 <general_hilo_operand> instead of <general_operand>.
4139 (*addcarry<mode>_1): New define_insn.
4140 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4142 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4144 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4145 Use "call" instead of "set".
4147 2020-01-03 Martin Jambor <mjambor@suse.cz>
4150 * ipa-cp.c (print_all_lattices): Skip functions without info.
4152 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4155 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4156 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4157 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4158 for 'e' simd clones.
4161 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4163 (mprefer-vector-width=): Add Save.
4164 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4165 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4166 (ix86_debug_options, ix86_function_specific_print): Adjust
4167 ix86_target_string callers.
4168 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4169 (ix86_valid_target_attribute_tree): Likewise.
4170 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4171 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4172 ix86_target_string caller.
4175 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4176 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4177 instead of gen_int_shift_amount + convert_modes.
4179 PR rtl-optimization/93088
4180 * loop-iv.c (find_single_def_src): Punt after looking through
4181 128 reg copies for regs with single definitions. Move definitions
4184 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4186 * config/arm/arm-c.c (arm_cpu_builtins): Define
4187 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4188 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4189 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4190 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4191 * config/arm/arm-tables.opt: Regenerated.
4192 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4193 arm_arch_i8mm and arm_arch_bf16 when enabled.
4194 * config/arm/arm.h (TARGET_I8MM): New macro.
4195 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4196 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4197 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4198 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4199 (v8_6_a_simd_variants): New.
4200 (v8_*_a_simd_variants): Add i8mm and bf16.
4201 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4203 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4206 * predict.c (compute_function_frequency): Don't call
4207 warn_function_cold on functions that already have cold attribute.
4209 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4212 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4213 COMDAT group function labels in .data.rel.ro.local section.
4214 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4217 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4218 comparison_operator in B and S integer comparisons. Likewise, use
4219 ordered_comparison_operator instead of cmpib_comparison_operator in
4221 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4223 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4225 Update copyright years.
4227 * gcc.c (process_command): Update copyright notice dates.
4228 * gcov-dump.c (print_version): Ditto.
4229 * gcov.c (print_version): Ditto.
4230 * gcov-tool.c (print_version): Ditto.
4231 * gengtype.c (create_file): Ditto.
4232 * doc/cpp.texi: Bump @copying's copyright year.
4233 * doc/cppinternals.texi: Ditto.
4234 * doc/gcc.texi: Ditto.
4235 * doc/gccint.texi: Ditto.
4236 * doc/gcov.texi: Ditto.
4237 * doc/install.texi: Ditto.
4238 * doc/invoke.texi: Ditto.
4240 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4242 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4245 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4247 PR tree-optimization/93098
4248 * match.pd (popcount): For shift amounts, use integer_onep
4249 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4250 tests. Make sure that precision is power of two larger than or equal
4251 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4252 instead of ULL suffixed constants. Formatting fixes.
4254 Copyright (C) 2020 Free Software Foundation, Inc.
4256 Copying and distribution of this file, with or without modification,
4257 are permitted in any medium without royalty provided the copyright
4258 notice and this notice are preserved.