1 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
3 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
4 support. Use aarch64_emit_mult instead of emitting multiplication
6 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
7 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
9 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
11 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
12 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
13 instead of emitting multiplication instructions directly.
14 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
15 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
16 (@aarch64_frecps<mode>): New expanders.
18 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
20 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
21 on and produce uint64_ts rather than ints.
22 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
23 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
25 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
27 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
28 an unused xmsk register when handling approximate rsqrt.
30 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
32 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
33 flag_finite_math_only condition.
35 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
38 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
39 to destination operand for shufps alternative.
40 (*vec_extractv2si_1): Ditto.
42 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
45 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
48 2020-02-20 Martin Liska <mliska@suse.cz>
51 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
53 2020-02-20 Martin Liska <mliska@suse.cz>
56 * common/config/avr/avr-common.c: Remote trailing "|".
58 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
60 * collect2.c (maybe_run_lto_and_relink): Fix typo in
63 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
65 PR tree-optimization/93767
66 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
67 access-size bias from the offset calculations for negative strides.
69 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
71 * collect2.c (c_file, o_file): Make const again.
72 (ldout,lderrout, dump_ld_file): Remove.
73 (tool_cleanup): Avoid calling not signal-safe functions.
74 (maybe_run_lto_and_relink): Avoid possible signal handler
75 access to unintialzed memory (lto_o_files).
76 (main): Avoid leaking temp files in $TMPDIR.
77 Initialize c_file/o_file with concat, which avoids exposing
78 uninitialized memory to signal handler, which calls unlink(!).
79 Avoid calling maybe_unlink when the main function returns,
80 since the atexit handler is already doing this.
81 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
83 2020-02-19 Martin Jambor <mjambor@suse.cz>
85 PR tree-optimization/93776
86 * tree-sra.c (create_access): Do not create zero size accesses.
87 (get_access_for_expr): Do not search for zero sized accesses.
89 2020-02-19 Martin Jambor <mjambor@suse.cz>
91 PR tree-optimization/93667
92 * tree-sra.c (scalarizable_type_p): Return false if record fields
93 do not follow wach other.
95 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
97 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
98 rather than fmv.x.s/fmv.s.x.
100 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
102 * config/aarch64/aarch64-simd-builtins.def
103 (intrinsic_vec_smult_lo_): New.
104 (intrinsic_vec_umult_lo_): Likewise.
105 (vec_widen_smult_hi_): Likewise.
106 (vec_widen_umult_hi_): Likewise.
107 * config/aarch64/aarch64-simd.md
108 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
109 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
110 (vmull_high_s16): Likewise.
111 (vmull_high_s32): Likewise.
112 (vmull_high_u8): Likewise.
113 (vmull_high_u16): Likewise.
114 (vmull_high_u32): Likewise.
115 (vmull_s8): Likewise.
116 (vmull_s16): Likewise.
117 (vmull_s32): Likewise.
118 (vmull_u8): Likewise.
119 (vmull_u16): Likewise.
120 (vmull_u32): Likewise.
122 2020-02-18 Martin Liska <mliska@suse.cz>
124 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
125 bootstrap by missing removal of invalid sanity check.
127 2020-02-18 Martin Liska <mliska@suse.cz>
130 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
131 Always compare LHS of gimple_assign.
133 2020-02-18 Martin Liska <mliska@suse.cz>
136 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
137 and return type of functions.
138 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
139 Drop MALLOC attribute for void functions.
140 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
141 malloc_state for a new VOID clone.
143 2020-02-18 Martin Liska <mliska@suse.cz>
146 * common.opt: Add -fprofile-reproducibility.
147 * doc/invoke.texi: Document it.
148 * value-prof.c (dump_histogram_value):
149 Document and support behavior for counters[0]
150 being a negative value.
151 (get_nth_most_common_value): Handle negative
152 counters[0] in respect to flag_profile_reproducible.
154 2020-02-18 Jakub Jelinek <jakub@redhat.com>
157 * cgraph.c (verify_speculative_call): Use speculative_id instead of
158 speculative_uid in messages. Remove trailing whitespace from error
159 message. Use num_speculative_call_targets instead of
160 num_speculative_targets in a message.
161 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
162 edge messages and stmt instead of cal_stmt in reference message.
164 PR tree-optimization/93780
165 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
166 before calling build_vector_type.
167 (execute_update_addresses_taken): Likewise.
170 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
171 typo, functoin -> function.
172 * tree.c (free_lang_data_in_decl): Fix comment typo,
173 functoin -> function.
174 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
176 2020-02-17 David Malcolm <dmalcolm@redhat.com>
178 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
180 (print_option_information): Don't call get_option_url if URLs
183 2020-02-17 Alexandre Oliva <oliva@adacore.com>
185 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
186 handling of register_common-less targets.
188 2020-02-17 Martin Liska <mliska@suse.cz>
191 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
193 2020-02-17 Martin Liska <mliska@suse.cz>
196 * config/rs6000/rs6000.c (rs6000_option_override_internal):
199 2020-02-17 Martin Liska <mliska@suse.cz>
202 * config/rx/elf.opt: Fix typo.
204 2020-02-17 Richard Biener <rguenther@suse.de>
207 * opts-global.c (print_ignored_options): Use inform and
210 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
213 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
215 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
218 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
219 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
221 2020-02-15 Jason Merrill <jason@redhat.com>
223 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
225 2020-02-15 Jakub Jelinek <jakub@redhat.com>
227 PR tree-optimization/93744
228 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
229 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
230 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
231 sure @2 in the first and @1 in the other patterns has no side-effects.
233 2020-02-15 David Malcolm <dmalcolm@redhat.com>
234 Bernd Edlinger <bernd.edlinger@hotmail.de>
238 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
239 * configure.ac (--with-diagnostics-urls): New configuration
240 option, based on --with-diagnostics-color.
241 (DIAGNOSTICS_URLS_DEFAULT): New define.
242 * config.h: Regenerate.
243 * configure: Regenerate.
244 * diagnostic.c (diagnostic_urls_init): Handle -1 for
245 DIAGNOSTICS_URLS_DEFAULT from configure-time
246 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
247 and TERM_URLS environment variable.
248 * diagnostic-url.h (diagnostic_url_format): New enum type.
249 (diagnostic_urls_enabled_p): rename to...
250 (determine_url_format): ... this, and change return type.
251 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
252 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
253 the linux console, and mingw.
254 (diagnostic_urls_enabled_p): rename to...
255 (determine_url_format): ... this, and adjust.
256 * pretty-print.h (pretty_printer::show_urls): rename to...
257 (pretty_printer::url_format): ... this, and change to enum.
258 * pretty-print.c (pretty_printer::pretty_printer,
259 pp_begin_url, pp_end_url, test_urls): Adjust.
260 * doc/install.texi (--with-diagnostics-urls): Document the new
261 configuration option.
262 (--with-diagnostics-color): Document the existing interaction
263 with GCC_COLORS better.
264 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
265 vindex reference. Update description of defaults based on the above.
266 (-fdiagnostics-color): Update description of how -fdiagnostics-color
267 interacts with GCC_COLORS.
269 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
272 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
273 conjunction with TARGET_GNU_TLS in early return.
275 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
277 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
278 the mode is not wider than UNITS_PER_WORD.
280 2020-02-14 Martin Jambor <mjambor@suse.cz>
282 PR tree-optimization/93516
283 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
284 access of the same type as the parent.
285 (propagate_subaccesses_from_lhs): Likewise.
287 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
290 * config/i386/avx512vbmi2intrin.h
291 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
292 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
293 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
294 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
295 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
296 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
297 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
298 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
299 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
300 of lacking a closing parenthesis.
301 * config/i386/avx512vbmi2vlintrin.h
302 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
303 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
304 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
305 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
306 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
307 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
308 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
309 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
310 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
311 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
312 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
313 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
314 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
315 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
316 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
317 _mm_shldi_epi32, _mm_mask_shldi_epi32,
318 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
319 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
321 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
324 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
325 the target function entry.
327 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
329 * common/config/arc/arc-common.c (arc_option_optimization_table):
330 Disable if-conversion step when optimized for size.
332 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
334 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
335 R12-R15 are always in ARCOMPACT16_REGS register class.
336 * config/arc/arc.opt (mq-class): Deprecate.
337 * config/arc/constraint.md ("q"): Remove dependency on mq-class
339 * doc/invoke.texi (mq-class): Update text.
340 * common/config/arc/arc-common.c (arc_option_optimization_table):
343 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
345 * config/arc/arc.c (arc_insn_cost): New function.
346 (TARGET_INSN_COST): Define.
347 * config/arc/arc.md (cost): New attribute.
348 (add_n): Use arc_nonmemory_operand.
349 (ashlsi3_insn): Likewise, also update constraints.
350 (ashrsi3_insn): Likewise.
352 (add_shift): Likewise.
353 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
355 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
357 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
359 (umulsidi_600): Likewise.
361 2020-02-13 Jakub Jelinek <jakub@redhat.com>
364 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
365 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
366 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
367 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
368 pass __A to the builtin followed by __W instead of __A followed by
370 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
371 _mm512_mask_popcnt_epi64): Likewise.
372 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
373 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
374 _mm256_mask_popcnt_epi64): Likewise.
376 PR tree-optimization/93582
377 * fold-const.h (shift_bytes_in_array_left,
378 shift_bytes_in_array_right): Declare.
379 * fold-const.c (shift_bytes_in_array_left,
380 shift_bytes_in_array_right): New function, moved from
381 gimple-ssa-store-merging.c, no longer static.
382 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
383 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
384 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
385 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
386 shift_bytes_in_array.
387 (verify_shift_bytes_in_array): Rename to ...
388 (verify_shift_bytes_in_array_left): ... this. Use
389 shift_bytes_in_array_left instead of shift_bytes_in_array.
390 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
391 instead of verify_shift_bytes_in_array.
392 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
393 / native_interpret_expr where the store covers all needed bits,
394 punt on PDP-endian, otherwise allow all involved offsets and sizes
395 not to be byte-aligned.
398 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
399 use const_0_to_255_operand predicate instead of immediate_operand.
400 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
401 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
402 vgf2p8affineinvqb_<mode><mask_name>,
403 vgf2p8affineqb_<mode><mask_name>): Drop mode from
404 const_0_to_255_operand predicated operands.
406 2020-02-12 Jeff Law <law@redhat.com>
408 * config/h8300/h8300.md (comparison shortening peepholes): Use
409 a mode iterator to merge the HImode and SImode peepholes.
411 2020-02-12 Jakub Jelinek <jakub@redhat.com>
414 * real.c (is_even): Make static. Function comment fix.
415 (is_halfway_below): Make static, don't assert R is not inf/nan,
416 instead return false for those. Small formatting fixes.
418 2020-02-12 Martin Sebor <msebor@redhat.com>
421 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
422 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
423 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
424 (strlen_check_and_optimize_call): Adjust callee name.
426 2020-02-12 Jeff Law <law@redhat.com>
428 * config/h8300/h8300.md (comparison shortening peepholes): Drop
429 (and (xor)) variant. Combine other two into single peephole.
431 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
433 PR rtl-optimization/93565
434 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
436 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
438 * config/aarch64/aarch64-simd.md
439 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
440 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
441 generating separate ADDV and zero_extend patterns.
442 * config/aarch64/iterators.md (VDQV_E): New iterator.
444 2020-02-12 Jeff Law <law@redhat.com>
446 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
447 expanders, splits, etc.
448 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
449 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
450 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
451 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
452 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
453 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
455 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
457 2020-02-12 Jakub Jelinek <jakub@redhat.com>
460 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
461 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
462 TARGET_AVX512DQ from condition.
463 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
464 instead of <mask_mode512bit_condition> in condition. If
465 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
467 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
470 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
473 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
475 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
477 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
478 where strlen is more legible.
479 (rs6000_builtin_vectorized_libmass): Ditto.
480 (rs6000_print_options_internal): Ditto.
482 2020-02-11 Martin Sebor <msebor@redhat.com>
484 PR tree-optimization/93683
485 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
487 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
489 * config/rs6000/predicates.md (cint34_operand): Rename the
490 -mprefixed-addr option to be -mprefixed.
491 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
492 the -mprefixed-addr option to be -mprefixed.
493 (OTHER_FUTURE_MASKS): Likewise.
494 (POWERPC_MASKS): Likewise.
495 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
496 the -mprefixed-addr option to be -mprefixed. Change error
497 messages to refer to -mprefixed.
498 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
500 (rs6000_legitimate_offset_address_p): Likewise.
501 (rs6000_mode_dependent_address): Likewise.
502 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
503 "-mprefixed" for target attributes and pragmas.
504 (address_to_insn_form): Rename the -mprefixed-addr option to be
506 (rs6000_adjust_insn_length): Likewise.
507 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
508 -mprefixed-addr option to be -mprefixed.
509 (ASM_OUTPUT_OPCODE): Likewise.
510 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
511 -mprefixed-addr option to be -mprefixed.
512 * config/rs6000/rs6000.opt (-mprefixed): Rename the
513 -mprefixed-addr option to be prefixed. Change the option from
514 being undocumented to being documented.
515 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
516 -mprefixed option. Update the -mpcrel documentation to mention
519 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
521 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
522 including FIRST_PSEUDO_REGISTER - 1.
523 * ira-color.c (print_hard_reg_set): Ditto.
525 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
527 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
528 (USTERNOP_QUALIFIERS): New define.
529 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
530 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
531 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
532 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
533 * config/arm/arm_neon.h (vusdot_s32): New.
534 (vusdot_lane_s32): New.
535 (vusdotq_lane_s32): New.
536 (vsudot_lane_s32): New.
537 (vsudotq_lane_s32): New.
538 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
539 * config/arm/iterators.md (DOTPROD_I8MM): New.
540 (sup, opsuffix): Add <us/su>.
541 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
542 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
544 2020-02-11 Richard Biener <rguenther@suse.de>
546 PR tree-optimization/93661
547 PR tree-optimization/93662
548 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
550 * tree-sra.c (get_access_for_expr): Likewise.
552 2020-02-10 Jakub Jelinek <jakub@redhat.com>
555 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
556 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
557 Change condition from TARGET_AVX2 to TARGET_AVX.
559 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
562 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
565 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
567 Try to generate zero-based comparisons.
568 * config/cris/cris.c (cris_reduce_compare): New function.
569 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
570 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
571 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
573 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
576 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
577 in Thumb state and also as a destination in Arm state. Add T16
580 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
582 * md.texi (Define Subst): Match closing paren in example.
584 2020-02-10 Jakub Jelinek <jakub@redhat.com>
588 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
589 arguments of strncmp.
591 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
594 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
595 but different source value.
596 (adjust_callers_for_value_intersection): New function.
597 (gather_edges_for_value): Adjust order of callers to let a
598 non-self-recursive caller be the first element.
599 (self_recursive_pass_through_p): Add a new parameter "simple", and
600 check generalized self-recursive pass-through jump function.
601 (self_recursive_agg_pass_through_p): Likewise.
602 (find_more_scalar_values_for_callers_subset): Compute value from
603 pass-through jump function for self-recursive.
604 (intersect_with_plats): Cleanup previous implementation code for value
605 itersection with self-recursive call edge.
606 (intersect_with_agg_replacements): Likewise.
607 (intersect_aggregates_with_edge): Deduce value from pass-through jump
608 function for self-recursive call edge. Cleanup previous implementation
609 code for value intersection with self-recursive call edge.
610 (decide_whether_version_node): Remove dead callers and adjust order
611 to let a non-self-recursive caller be the first element.
613 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
615 * recog.c: Move pass_split_before_sched2 code in front of
616 pass_split_before_regstack.
617 (pass_data_split_before_sched2): Rename pass to split3 from split4.
618 (pass_data_split_before_regstack): Rename pass to split4 from split3.
619 (rest_of_handle_split_before_sched2): Remove.
620 (pass_split_before_sched2::execute): Unconditionally call
622 (enable_split_before_sched2): New function.
623 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
624 (pass_split_before_regstack::gate): Ditto.
625 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
626 Update name check for renamed split4 pass.
627 * config/sh/sh.c (register_sh_passes): Update pass insertion
628 point for renamed split4 pass.
630 2020-02-09 Jakub Jelinek <jakub@redhat.com>
632 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
633 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
634 copying them around between host and target.
636 2020-02-08 Andrew Pinski <apinski@marvell.com>
639 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
640 STRICT_ALIGNMENT also.
642 2020-02-08 Jim Wilson <jimw@sifive.com>
645 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
647 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
648 Jakub Jelinek <jakub@redhat.com>
651 * config/i386/i386.h (CALL_USED_REGISTERS): Make
652 xmm16-xmm31 call-used even in 64-bit ms-abi.
654 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
656 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
657 (simd_ummla, simd_usmmla): Likewise.
658 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
659 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
662 2020-02-07 Richard Biener <rguenther@suse.de>
665 * tree-inline.c (fold_marked_statements): Do a PRE walk,
666 skipping unreachable regions.
667 (optimize_inline_calls): Skip folding stmts when we didn't
670 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
673 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
674 Don't return aggregates with only SFmode and DFmode in SSE
676 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
678 2020-02-07 Jakub Jelinek <jakub@redhat.com>
681 * config/rs6000/rs6000-logue.c
682 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
683 if it fails, move rs into end_addr and retry. Add
684 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
685 the insn pattern doesn't describe well what exactly happens to
689 * config/i386/predicates.md (avx_identity_operand): Remove.
690 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
691 (avx_<castmode><avxsizesuffix>_<castmode>,
692 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
693 a VEC_CONCAT of the operand and UNSPEC_CAST.
694 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
695 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
699 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
700 recog_data.insn if distance_non_agu_define changed it.
702 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
705 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
706 we only had X-FORM (reg+reg) addressing for vectors. Also before
707 ISA 3.0, we only had X-FORM addressing for scalars in the
708 traditional Altivec registers.
710 2020-02-06 <zhongyunde@huawei.com>
711 Vladimir Makarov <vmakarov@redhat.com>
713 PR rtl-optimization/93561
714 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
717 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
719 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
722 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
724 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
725 where the low and the high 32 bits are equal to each other specially,
726 with an rldimi instruction.
728 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
730 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
732 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
734 * config/arm/arm-tables.opt: Regenerate.
736 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
739 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
740 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
741 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
743 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
745 PR rtl-optimization/87763
746 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
748 2020-02-06 Delia Burduv <delia.burduv@arm.com>
750 * config/aarch64/aarch64-simd-builtins.def
751 (bfmlaq): New built-in function.
752 (bfmlalb): New built-in function.
753 (bfmlalt): New built-in function.
754 (bfmlalb_lane): New built-in function.
755 (bfmlalt_lane): New built-in function.
756 * config/aarch64/aarch64-simd.md
757 (aarch64_bfmmlaqv4sf): New pattern.
758 (aarch64_bfmlal<bt>v4sf): New pattern.
759 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
760 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
761 (vbfmlalbq_f32): New intrinsic.
762 (vbfmlaltq_f32): New intrinsic.
763 (vbfmlalbq_lane_f32): New intrinsic.
764 (vbfmlaltq_lane_f32): New intrinsic.
765 (vbfmlalbq_laneq_f32): New intrinsic.
766 (vbfmlaltq_laneq_f32): New intrinsic.
767 * config/aarch64/iterators.md (BF_MLA): New int iterator.
768 (bt): New int attribute.
770 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
772 * config/i386/i386.md (*pushtf): Emit "#" instead of
773 calling gcc_unreachable in insn output.
776 (*pushsf_rex64): Ditto for alternatives other than 1.
777 (*pushsf): Ditto for alternatives other than 1.
779 2020-02-06 Martin Liska <mliska@suse.cz>
781 PR gcov-profile/91971
782 PR gcov-profile/93466
783 * coverage.c (coverage_init): Revert mangling of
784 path into filename. It can lead to huge filename length.
785 Creation of subfolders seem more natural.
787 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
790 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
791 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
792 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
794 2020-02-06 Jakub Jelinek <jakub@redhat.com>
797 * config/i386/predicates.md (avx_identity_operand): New predicate.
798 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
799 define_insn_and_split.
802 * omp-low.c (use_pointer_for_field): For nested constructs, also
803 look for map clauses on target construct.
804 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
805 taskreg_nesting_level.
808 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
809 shared clause, call omp_notice_variable on outer context if any.
811 2020-02-05 Jason Merrill <jason@redhat.com>
814 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
815 non-zero address even if weak and not yet defined.
817 2020-02-05 Martin Sebor <msebor@redhat.com>
819 PR tree-optimization/92765
820 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
821 * tree-ssa-strlen.c (compute_string_length): Remove.
822 (determine_min_objsize): Remove.
823 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
824 Avoid using type size as the upper bound on string length.
825 (handle_builtin_string_cmp): Add an argument. Adjust.
826 (strlen_check_and_optimize_call): Pass additional argument to
827 handle_builtin_string_cmp.
829 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
831 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
832 (*pushdi2_rex64 peephole2): Unconditionally split after
834 (*ashl<mode>3_doubleword): Ditto.
835 (*<shift_insn><mode>3_doubleword): Ditto.
837 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
840 * config/rs6000/rs6000.c (get_vector_offset): Fix
842 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
844 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
846 2020-02-05 David Malcolm <dmalcolm@redhat.com>
849 (Special Functions for Debugging the Analyzer): Update description
850 of __analyzer_dump_exploded_nodes.
852 2020-02-05 Jakub Jelinek <jakub@redhat.com>
855 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
856 include sets and not clobbers in the vzeroupper pattern.
857 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
858 the parallel has 17 (64-bit) or 9 (32-bit) elts.
859 (*avx_vzeroupper_1): New define_insn_and_split.
862 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
863 don't run when !optimize.
864 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
867 2020-02-05 Richard Biener <rguenther@suse.de>
870 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
871 checks before matching calls.
873 2020-02-05 Jakub Jelinek <jakub@redhat.com>
875 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
876 function comment typo.
879 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
880 simd_clone_create failed when i == 0, adjust clone->nargs by
883 2020-02-05 Martin Liska <mliska@suse.cz>
886 * doc/invoke.texi: Document that one should
887 not combine ASLR and -fpch.
889 2020-02-04 Richard Biener <rguenther@suse.de>
891 PR tree-optimization/93538
892 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
894 2020-02-04 Richard Biener <rguenther@suse.de>
896 PR tree-optimization/91123
897 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
898 (vn_walk_cb_data::last_vuse): New member.
899 (vn_walk_cb_data::saved_operands): Likewsie.
900 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
901 (vn_walk_cb_data::push_partial_def): Use finish.
902 (vn_reference_lookup_2): Update last_vuse and use finish if
903 we've saved operands.
904 (vn_reference_lookup_3): Use finish and update calls to
905 push_partial_defs everywhere. When translating through
906 memcpy or aggregate copies save off operands and alias-set.
907 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
908 operation for redundant store removal.
910 2020-02-04 Richard Biener <rguenther@suse.de>
912 PR tree-optimization/92819
913 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
914 generating more stmts than before.
916 2020-02-04 Martin Liska <mliska@suse.cz>
918 * config/arm/arm.c (arm_gen_far_branch): Move the function
919 outside of selftests.
921 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
923 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
924 function to adjust PC-relative vector addresses.
925 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
926 handle vectors with PC-relative addresses.
928 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
930 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
932 (hard_reg_and_mode_to_addr_mask): Delete.
933 (rs6000_adjust_vec_address): If the original vector address
934 was REG+REG or REG+OFFSET and the element is not zero, do the add
935 of the elements in the original address before adding the offset
936 for the vector element. Use address_to_insn_form to validate the
937 address using the register being loaded, rather than guessing
938 whether the address is a DS-FORM or DQ-FORM address.
940 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
942 * config/rs6000/rs6000.c (get_vector_offset): New helper function
943 to calculate the offset in memory from the start of a vector of a
944 particular element. Add code to keep the element number in
945 bounds if the element number is variable.
946 (rs6000_adjust_vec_address): Move calculation of offset of the
947 vector element to get_vector_offset.
948 (rs6000_split_vec_extract_var): Do not do the initial AND of
949 element here, move the code to get_vector_offset.
951 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
953 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
956 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
958 * config/rs6000/constraints.md: Improve documentation.
960 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
963 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
964 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
966 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
968 * config.gcc: Remove "carrizo" support.
969 * config/gcn/gcn-opts.h (processor_type): Likewise.
970 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
971 * config/gcn/gcn.opt (gpu_type): Likewise.
972 * config/gcn/t-omp-device: Likewise.
974 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
977 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
978 * config/arm/arm.c (arm_gen_far_branch): New function
980 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
982 2020-02-03 Julian Brown <julian@codesourcery.com>
983 Tobias Burnus <tobias@codesourcery.com>
985 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
987 2020-02-03 Jakub Jelinek <jakub@redhat.com>
990 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
991 valid RTL to sum up the lowest and second lowest bytes of the popcnt
994 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
996 PR rtl-optimization/91333
997 * ira-color.c (struct allocno_color_data): Add member
999 (init_allocno_threads): Set the member up.
1000 (bucket_allocno_compare_func): Add compare hard reg
1003 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1005 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1007 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1008 * config.in: Regenerated.
1009 * configure: Regenerated.
1010 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1011 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1012 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1014 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1016 * configure: Regenerate.
1018 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1020 PR rtl-optimization/91333
1021 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1022 reg preferences comparison up.
1024 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1026 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1027 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1028 aarch64-sve-builtins-base.h.
1029 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1030 aarch64-sve-builtins-base.cc.
1031 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1032 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1034 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1035 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1036 (svcvtnt): New functions.
1037 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1038 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1039 (svcvtnt): New functions.
1040 (svcvt): Add a form that converts f32 to bf16.
1041 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1042 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1044 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1045 Treat B as bfloat16_t.
1046 (ternary_bfloat_lane_base): New class.
1047 (ternary_bfloat_def): Likewise.
1048 (ternary_bfloat): New shape.
1049 (ternary_bfloat_lane_def): New class.
1050 (ternary_bfloat_lane): New shape.
1051 (ternary_bfloat_lanex2_def): New class.
1052 (ternary_bfloat_lanex2): New shape.
1053 (ternary_bfloat_opt_n_def): New class.
1054 (ternary_bfloat_opt_n): New shape.
1055 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1056 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1057 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1058 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1059 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1060 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1061 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1062 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1063 the pattern off the narrow mode instead of the wider one.
1064 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1065 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1066 (sve_fp_op): Handle them.
1067 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1068 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1070 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1072 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1073 * config/aarch64/aarch64-modes.def (BF): Move definition before
1074 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1075 (SVE_MODES): Handle BF modes.
1076 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1078 (aarch64_full_sve_mode): Likewise.
1079 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1081 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1082 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1083 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1084 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1086 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1088 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1089 (TYPES_all_data): Add bf16.
1090 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1091 (register_tuple_type): Increase buffer size.
1092 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1093 (bf16): New type suffix.
1094 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1095 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1096 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1097 Change type from all_data to all_arith.
1098 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1101 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1102 Matthew Malcomson <matthew.malcomson@arm.com>
1103 Richard Sandiford <richard.sandiford@arm.com>
1105 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1106 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1107 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1108 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1109 __ARM_FEATURE_MATMUL_FP64.
1110 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1111 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1112 be disabled at the same time.
1113 (f32mm): New extension.
1114 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1115 (AARCH64_FL_F64MM): Bump to the next bit up.
1116 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1117 (TARGET_SVE_F64MM): New macros.
1118 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1119 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1120 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1121 (UNSPEC_ZIP2Q): New unspeccs.
1122 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1123 (optab, sur, perm_insn): Handle the new unspecs.
1124 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1125 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1126 TARGET_SVE_F64MM instead of separate tests.
1127 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1128 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1129 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1130 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1131 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1132 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1133 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1134 (TYPES_s_signed): New macro.
1135 (TYPES_s_integer): Use it.
1136 (TYPES_d_float): New macro.
1137 (TYPES_d_data): Use it.
1138 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1139 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1140 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1141 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1142 (svmmla): New shape.
1143 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1144 template parameters.
1145 (ternary_resize2_lane_base): Likewise.
1146 (ternary_resize2_base): New class.
1147 (ternary_qq_lane_base): Likewise.
1148 (ternary_intq_uintq_lane_def): Likewise.
1149 (ternary_intq_uintq_lane): New shape.
1150 (ternary_intq_uintq_opt_n_def): New class
1151 (ternary_intq_uintq_opt_n): New shape.
1152 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1153 (ternary_uintq_intq_def): New class.
1154 (ternary_uintq_intq): New shape.
1155 (ternary_uintq_intq_lane_def): New class.
1156 (ternary_uintq_intq_lane): New shape.
1157 (ternary_uintq_intq_opt_n_def): New class.
1158 (ternary_uintq_intq_opt_n): New shape.
1159 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1160 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1161 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1162 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1164 (svdotprod_lane_impl): ...this new class.
1165 (svmmla_impl, svusdot_impl): New classes.
1166 (svdot_lane): Update to use svdotprod_lane_impl.
1167 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1168 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1170 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1171 function, with no types defined.
1172 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1173 AARCH64_FL_I8MM functions.
1174 (svmmla): New AARCH64_FL_F32MM function.
1175 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1176 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1177 AARCH64_FL_F64MM function.
1178 (REQUIRED_EXTENSIONS):
1180 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1182 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1185 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1187 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1188 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1189 (*movti_internal): Do not check for
1190 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1191 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1192 just after check for TARGET_AVX.
1193 (*movdf_internal): Ditto.
1194 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1195 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1196 * config/i386/sse.md (mov<mode>_internal): Only check
1197 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1198 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1199 (<sse>_andnot<mode>3<mask_name>): Move check for
1200 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1201 (<code><mode>3<mask_name>): Ditto.
1202 (*andnot<mode>3): Ditto.
1203 (*andnottf3): Ditto.
1204 (*<code><mode>3): Ditto.
1205 (*<code>tf3): Ditto.
1206 (*andnot<VI:mode>3): Remove
1207 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1208 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1209 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1210 (sse4_1_blendv<ssemodesuffix>): Ditto.
1211 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1212 Explain that tune applies to 128bit instructions only.
1214 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1216 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1217 to definition of hsa_kernel_description. Parse assembly to find SGPR
1218 and VGPR count of kernel and store in hsa_kernel_description.
1220 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1222 PR rtl-optimization/91838
1223 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1224 to truncate if allowed or reject combination.
1226 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1228 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1229 (find_inv_vars_cb): Likewise.
1231 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1233 * calls.c (special_function_p): Split out the check for DECL_NAME
1234 being non-NULL and fndecl being extern at file scope into a
1235 new maybe_special_function_p and call it. Drop check for fndecl
1236 being non-NULL that was after a usage of DECL_NAME (fndecl).
1237 * tree.h (maybe_special_function_p): New inline function.
1239 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1241 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1242 (mask_gather_load<mode>): ... here, and zero-initialize the
1244 (maskload<mode>di): Zero-initialize the destination.
1247 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1250 * doc/analyzer.texi (Limitations): Note that constraints on
1251 floating-point values are currently ignored.
1253 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1256 * symtab.c (symtab_node::noninterposable_alias): If localalias
1257 already exists, but is not usable, append numbers after it until
1258 a unique name is found. Formatting fix.
1261 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1264 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1266 * config/gcn/gcn.c (print_operand): Handle LTGT.
1267 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1269 2020-01-30 Richard Biener <rguenther@suse.de>
1271 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1272 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1274 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1276 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1277 without a DECL in .data.rel.ro.local.
1279 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1282 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1286 * config/i386/sse.md
1287 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1288 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1289 any_extend code iterator instead of always zero_extend.
1290 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1291 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1292 Use any_extend code iterator instead of always zero_extend.
1293 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1294 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1295 Use any_extend code iterator instead of always zero_extend.
1296 (*sse2_pmovmskb_ext): New define_insn.
1297 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1300 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1301 (*popcountsi2_zext_falsedep): New define_insn.
1303 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1305 * config.in: Regenerated.
1306 * configure: Regenerated.
1308 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1311 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1312 LLVM's assembler changed the default in version 9.
1314 2020-01-24 Jeff Law <law@redhat.com>
1316 PR tree-optimization/89689
1317 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1319 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1323 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1325 PR rtl-optimization/87763
1326 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1327 simplification to handle subregs as well as bare regs.
1328 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1330 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1333 * ira.c (ira): Revert use of simplified LRA algorithm.
1335 2020-01-29 Martin Jambor <mjambor@suse.cz>
1337 PR tree-optimization/92706
1338 * tree-sra.c (struct access): Fields first_link, last_link,
1339 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1340 next_rhs_queued and grp_rhs_queued respectively, new fields
1341 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1342 (struct assign_link): Field next renamed to next_rhs, new field
1343 next_lhs. Updated comment.
1344 (work_queue_head): Renamed to rhs_work_queue_head.
1345 (lhs_work_queue_head): New variable.
1346 (add_link_to_lhs): New function.
1347 (relink_to_new_repr): Also relink LHS lists.
1348 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1349 (add_access_to_lhs_work_queue): New function.
1350 (pop_access_from_work_queue): Renamed to
1351 pop_access_from_rhs_work_queue.
1352 (pop_access_from_lhs_work_queue): New function.
1353 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1355 (child_would_conflict_in_lacc): Renamed to
1356 child_would_conflict_in_acc. Adjusted parameter names.
1357 (create_artificial_child_access): New parameter set_grp_read, use it.
1358 (subtree_mark_written_and_enqueue): Renamed to
1359 subtree_mark_written_and_rhs_enqueue.
1360 (propagate_subaccesses_across_link): Renamed to
1361 propagate_subaccesses_from_rhs.
1362 (propagate_subaccesses_from_lhs): New function.
1363 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1366 2020-01-29 Martin Jambor <mjambor@suse.cz>
1368 PR tree-optimization/92706
1369 * tree-sra.c (struct access): Adjust comment of
1370 grp_total_scalarization.
1371 (find_access_in_subtree): Look for single children spanning an entire
1373 (scalarizable_type_p): Allow register accesses, adjust callers.
1374 (completely_scalarize): Remove function.
1375 (scalarize_elem): Likewise.
1376 (create_total_scalarization_access): Likewise.
1377 (sort_and_splice_var_accesses): Do not track total scalarization
1379 (analyze_access_subtree): New parameter totally, adjust to new meaning
1380 of grp_total_scalarization.
1381 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1382 (can_totally_scalarize_forest_p): New function.
1383 (create_total_scalarization_access): Likewise.
1384 (create_total_access_and_reshape): Likewise.
1385 (total_should_skip_creating_access): Likewise.
1386 (totally_scalarize_subtree): Likewise.
1387 (analyze_all_variable_accesses): Perform total scalarization after
1388 subaccess propagation using the new functions above.
1389 (initialize_constant_pool_replacements): Output initializers by
1390 traversing the access tree.
1392 2020-01-29 Martin Jambor <mjambor@suse.cz>
1394 * tree-sra.c (verify_sra_access_forest): New function.
1395 (verify_all_sra_access_forests): Likewise.
1396 (create_artificial_child_access): Set parent.
1397 (analyze_all_variable_accesses): Call the verifier.
1399 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1401 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1402 if called on indirect edge.
1403 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1404 speculative call if needed.
1406 2020-01-29 Richard Biener <rguenther@suse.de>
1408 PR tree-optimization/93428
1409 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1410 permutation when the load node is created.
1411 (vect_analyze_slp_instance): Re-use it here.
1413 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1415 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1417 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1419 PR rtl-optimization/93272
1420 * ira-lives.c (process_out_of_region_eh_regs): New function.
1421 (process_bb_node_lives): Call it.
1423 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1425 * coverage.c (read_counts_file): Make error message lowercase.
1427 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1429 * profile-count.c (profile_quality_display_names): Fix ordering.
1431 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1434 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1435 hash only when edge is first within the sequence.
1436 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1437 (symbol_table::create_edge): Do not set target_prob.
1438 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1440 (cgraph_edge::make_speculative): Drop target_prob parameter.
1441 (cgraph_edge::speculative_call_info): Remove.
1442 (cgraph_edge::first_speculative_call_target): New member function.
1443 (update_call_stmt_hash_for_removing_direct_edge): New function.
1444 (cgraph_edge::resolve_speculation): Rewrite to new API.
1445 (cgraph_edge::speculative_call_for_target): New member function.
1446 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1447 multiple speculation targets.
1448 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1450 (verify_speculative_call): Verify that targets form an interval.
1451 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1452 (cgraph_edge::first_speculative_call_target): New member function.
1453 (cgraph_edge::next_speculative_call_target): New member function.
1454 (cgraph_edge::speculative_call_target_ref): New member function.
1455 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1456 (cgraph_edge): Remove target_prob.
1457 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1458 Fix handling of speculative calls.
1459 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1460 * ipa-fnsummary.c (analyze_function_body): Likewise.
1461 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1462 * ipa-profile.c (dump_histogram): Fix formating.
1463 (ipa_profile_generate_summary): Watch for overflows.
1464 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1465 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1466 (update_indirect_edges_after_inlining): Update to new API.
1467 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1469 * profile-count.h: (profile_probability::adjusted): New.
1470 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1471 updating of profile.
1472 * value-prof.c (gimple_ic_transform): Rename to ...
1473 (dump_ic_profile): ... this one; update dumping.
1474 (stream_in_histogram_value): Fix formating.
1475 (gimple_value_profile_transformations): Update.
1477 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1480 * config/i386/i386.md (*movoi_internal_avx): Remove
1481 TARGET_SSE_TYPELESS_STORES check.
1482 (*movti_internal): Prefer TARGET_AVX over
1483 TARGET_SSE_TYPELESS_STORES.
1484 (*movtf_internal): Likewise.
1485 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1486 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1487 from TARGET_SSE_TYPELESS_STORES.
1489 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1491 * diagnostic-core.h (warning_at): Rename overload to...
1492 (warning_meta): ...this.
1493 (emit_diagnostic_valist): Delete decl of overload taking
1494 diagnostic_metadata.
1495 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1496 (warning_at): Rename overload taking diagnostic_metadata to...
1497 (warning_meta): ...this.
1499 2020-01-28 Richard Biener <rguenther@suse.de>
1501 PR tree-optimization/93439
1502 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1503 * tree-cfg.c (move_sese_region_to_fn): ... here.
1504 (verify_types_in_gimple_reference): Verify used cliques are
1507 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1510 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1511 argument of a pointer to struct gcc_options and pass it to
1512 parse_mtune_ctrl_str.
1513 (ix86_function_specific_restore): Pass opts to
1514 set_ix86_tune_features.
1515 (ix86_option_override_internal): Likewise.
1516 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1517 gcc_options and use it for x_ix86_tune_ctrl_string.
1519 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1521 PR rtl-optimization/87763
1522 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1523 simplification to handle subregs as well as bare regs.
1524 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1526 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1528 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1529 for reduction chains that (now) include a call.
1531 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1533 PR tree-optimization/92822
1534 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1535 out the don't-care elements of a vector whose significant elements
1536 are duplicates, make the don't-care elements duplicates too.
1538 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1540 PR tree-optimization/93434
1541 * tree-predcom.c (split_data_refs_to_components): Record which
1542 components have had aliasing loads removed. Prevent store-store
1543 commoning for all such components.
1545 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1548 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1549 -1 or is_vshift is true, use new_vector with number of elts npatterns
1550 rather than new_unary_operation.
1552 PR tree-optimization/93454
1553 * gimple-fold.c (fold_array_ctor_reference): Perform
1554 elt_size.to_uhwi () just once, instead of calling it in every
1555 iteration. Punt if that value is above size of the temporary
1556 buffer. Decrease third native_encode_expr argument when
1557 bufoff + elt_sz is above size of buf.
1559 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1561 * config/mips/mips.c (mips_declare_object_name)
1562 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1564 2020-01-27 Martin Liska <mliska@suse.cz>
1566 PR gcov-profile/93403
1567 * tree-profile.c (gimple_init_gcov_profiler): Generate
1568 both __gcov_indirect_call_profiler_v4 and
1569 __gcov_indirect_call_profiler_v4_atomic.
1571 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1574 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1576 (@aarch64_split_simd_mov<mode>): Use it.
1577 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1578 Leave the vec_extract patterns to handle 2-element vectors.
1579 (aarch64_simd_mov_from_<mode>high): Likewise.
1580 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1581 (vec_extractv2dfv1df): Likewise.
1583 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1585 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1586 jump conditions for *compare_condjump<GPI:mode>.
1588 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1591 * digraph.cc (test_edge::test_edge): Specify template for base
1594 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1596 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1598 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1600 * config/arc/arc-protos.h (gen_mlo): Remove.
1601 (gen_mhi): Likewise.
1602 * config/arc/arc.c (AUX_MULHI): Define.
1603 (arc_must_save_reister): Special handling for r58/59.
1604 (arc_compute_frame_size): Consider mlo/mhi registers.
1605 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1607 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1608 mlo/mhi name selection.
1609 (arc_restore_callee_saves): Don't early restore blink when ISR.
1610 (arc_expand_prologue): Add mlo/mhi saving.
1611 (arc_expand_epilogue): Add mlo/mhi restoring.
1614 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1615 numbering when MUL64 option is used.
1616 (DWARF2_FRAME_REG_OUT): Define.
1617 * config/arc/arc.md (arc600_stall): New pattern.
1618 (VUNSPEC_ARC_ARC600_STALL): Define.
1619 (mulsi64): Use correct mlo/mhi registers.
1620 (mulsi_600): Clean it up.
1621 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1623 (mhi_operand): Likewise.
1625 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1626 Petro Karashchenko <petro.karashchenko@ring.com>
1628 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1629 attributes if needed.
1630 (prepare_move_operands): Generate special unspec instruction for
1632 (arc_isuncached_mem_p): Propagate uncached attribute to each
1634 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1635 (VUNSPEC_ARC_STDI): Likewise.
1636 (ALLI): New mode iterator.
1637 (mALLI): New mode attribute.
1638 (lddi): New instruction pattern.
1640 (stdidi_split): Split instruction for architectures which are not
1641 supporting ll64 option.
1642 (lddidi_split): Likewise.
1644 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1646 PR rtl-optimization/92989
1647 * lra-lives.c (process_bb_lives): Update the live-in set before
1648 processing additional clobbers.
1650 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1652 PR rtl-optimization/93170
1653 * cselib.c (cselib_invalidate_regno_val): New function, split out
1655 (cselib_invalidate_regno): ...here.
1656 (cselib_invalidated_by_call_p): New function.
1657 (cselib_process_insn): Iterate over all the hard-register entries in
1658 REG_VALUES and invalidate any that cross call-clobbered registers.
1660 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1662 * dojump.c (split_comparison): Use HONOR_NANS rather than
1663 HONOR_SNANS when splitting LTGT.
1665 2020-01-27 Martin Liska <mliska@suse.cz>
1668 * opts.c (print_filtered_help): Exclude language-specific
1669 options from --help=common unless enabled in all FEs.
1671 2020-01-27 Martin Liska <mliska@suse.cz>
1673 * opts.c (print_help): Exclude params from
1674 all except --help=param.
1676 2020-01-27 Martin Liska <mliska@suse.cz>
1679 * config/i386/i386-features.c (make_resolver_func):
1680 Align the code with ppc64 target implementation.
1681 Do not generate a unique name for resolver function.
1683 2020-01-27 Richard Biener <rguenther@suse.de>
1685 PR tree-optimization/93397
1686 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1687 converted reduction chain SLP graph adjustment.
1689 2020-01-26 Marek Polacek <polacek@redhat.com>
1692 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1695 2020-01-26 Jason Merrill <jason@redhat.com>
1698 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1701 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1703 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1704 (rx_setmem): Likewise.
1706 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1709 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1710 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1711 drop <di> from constraint of last operand.
1714 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1715 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1716 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1718 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1721 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1724 2020-01-24 Jeff Law <law@redhat.com>
1726 PR tree-optimization/92788
1727 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1730 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1733 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1734 *avx_vperm_broadcast_<mode>,
1735 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1736 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1737 Move before avx2_perm<mode>/avx512f_perm<mode>.
1740 * simplify-rtx.c (simplify_const_unary_operation,
1741 simplify_const_binary_operation): Punt for mode precision above
1742 MAX_BITSIZE_MODE_ANY_INT.
1744 2020-01-24 Andrew Pinski <apinski@marvell.com>
1746 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1749 2020-01-24 Jeff Law <law@redhat.com>
1752 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1753 for REGs. Call output_operand_lossage to get more reasonable
1756 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1758 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1759 gcn_fp_compare_operator.
1760 (vec_cmpu<mode>di): Use gcn_compare_operator.
1761 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1762 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1763 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1764 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1765 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1766 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1767 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1768 gcn_fp_compare_operator.
1769 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1770 gcn_fp_compare_operator.
1771 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1772 gcn_fp_compare_operator.
1773 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1774 gcn_fp_compare_operator.
1776 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1778 * doc/install.texi (Cross-Compiler-Specific Options): Document
1779 `--with-toolexeclibdir' option.
1781 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1783 * target.def (flags_regnum): Also mention effect on delay slot filling.
1784 * doc/tm.texi: Regenerate.
1786 2020-01-23 Jeff Law <law@redhat.com>
1788 PR translation/90162
1789 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1791 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1794 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1797 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1799 PR rtl-optimization/93402
1800 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1803 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1805 * config.in: Regenerated.
1806 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1807 for TARGET_LIBC_GNUSTACK.
1808 * configure: Regenerated.
1809 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1810 found to be 2.31 or greater.
1812 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1814 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1816 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1817 (mips_asm_file_end): New function. Delegate to
1818 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1819 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1821 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1824 * config/i386/i386-modes.def (POImode): New mode.
1825 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1826 * config/i386/i386.md (DPWI): New mode attribute.
1827 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1829 (QPWI): ... this. Use POI instead of OI for TImode.
1830 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1831 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1834 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1837 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1839 (speculation_tracker_rev): New pattern.
1840 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1841 Use speculation_tracker_rev to track the inverse condition.
1843 2020-01-23 Richard Biener <rguenther@suse.de>
1845 PR tree-optimization/93381
1846 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1847 alias-set of the def as argument and record the first one.
1848 (vn_walk_cb_data::first_set): New member.
1849 (vn_reference_lookup_3): Pass the alias-set of the current def
1850 to push_partial_def. Fix alias-set used in the aggregate copy
1852 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1853 * real.c (clear_significand_below): Fix out-of-bound access.
1855 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1858 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1859 New define_insn patterns.
1861 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1863 * doc/sourcebuild.texi (check-function-bodies): Add an
1864 optional target/xfail selector.
1866 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1868 PR rtl-optimization/93124
1869 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1870 bare USE and CLOBBER insns.
1872 2020-01-22 Andrew Pinski <apinski@marvell.com>
1874 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1876 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1879 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1880 diagnostic_manager into "ana" namespace.
1881 * selftest-run-tests.c (selftest::run_tests): Update for move of
1882 selftest::run_analyzer_selftests to
1883 ana::selftest::run_analyzer_selftests.
1885 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1887 * cfgexpand.c (union_stack_vars): Update the size.
1889 2020-01-22 Richard Biener <rguenther@suse.de>
1891 PR tree-optimization/93381
1892 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1893 throughout, handle all conversions the same.
1895 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1898 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1899 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1900 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1901 Call force_reg on high_in2 unconditionally.
1903 2020-01-22 Martin Liska <mliska@suse.cz>
1905 PR tree-optimization/92924
1906 * profile.c (compute_value_histograms): Divide
1909 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1912 * output.h (assemble_name_resolve): Declare.
1913 * varasm.c (assemble_name_resolve): New function.
1914 (assemble_name): Use it.
1915 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1917 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1919 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1920 update_web_docs_git instead of update_web_docs_svn.
1922 2020-01-21 Andrew Pinski <apinski@marvell.com>
1925 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1926 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1927 (*tlsgd_small_<mode>): Likewise.
1928 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1929 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1930 register. Convert that register back to dest using convert_mode.
1932 2020-01-21 Jim Wilson <jimw@sifive.com>
1934 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1937 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1938 Uros Bizjak <ubizjak@gmail.com>
1941 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1943 (legitimize_tls_address): Do GNU2 TLS address computation in
1944 ptr_mode and zero-extend result to Pmode.
1945 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1946 :P with :PTR and Pmode with ptr_mode.
1947 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1948 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1949 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1951 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1954 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1955 the last two operands are CONST_INT_P before using them as such.
1957 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1959 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1960 to get the integer element types.
1962 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1964 * config/aarch64/aarch64-sve-builtins.h
1965 (function_expander::convert_to_pmode): Declare.
1966 * config/aarch64/aarch64-sve-builtins.cc
1967 (function_expander::convert_to_pmode): New function.
1968 (function_expander::get_contiguous_base): Use it.
1969 (function_expander::prepare_gather_address_operands): Likewise.
1970 * config/aarch64/aarch64-sve-builtins-sve2.cc
1971 (svwhilerw_svwhilewr_impl::expand): Likewise.
1973 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1976 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1977 cfun->machine->label_is_assembled.
1978 (aarch64_print_patchable_function_entry): New.
1979 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1980 * config/aarch64/aarch64.h (struct machine_function): New field,
1983 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1986 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1989 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1992 * cgraph.c (cgraph_edge::resolve_speculation,
1993 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1994 call_stmt_site_hash.
1996 2020-01-21 Martin Liska <mliska@suse.cz>
1998 * config/rs6000/rs6000.c (common_mode_defined): Remove
2001 2020-01-21 Richard Biener <rguenther@suse.de>
2003 PR tree-optimization/92328
2004 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2005 type when value-numbering same-sized store by inserting a
2007 (eliminate_dom_walker::eliminate_stmt): When eliminating
2008 a redundant store handle bit-reinterpretation of the same value.
2010 2020-01-21 Andrew Pinski <apinski@marvel.com>
2013 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2015 (prepare_block_for_update): This. Use a worklist instead of
2018 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2020 * gcc/config/arm/arm.c (clear_operation_p):
2021 Initialise last_regno, skip first iteration
2022 based on the first_set value and use ints instead
2023 of the unnecessary HOST_WIDE_INTs.
2025 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2028 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2029 compare_mode other than SFmode or DFmode.
2031 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2034 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2035 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2036 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2038 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2040 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2042 2020-01-20 Andrew Pinski <apinski@marvell.com>
2045 * targhooks.c (default_print_patchable_function_entry): Use
2046 output_asm_insn to emit the nop instruction.
2048 2020-01-20 Fangrui Song <maskray@google.com>
2051 * targhooks.c (default_print_patchable_function_entry): Align to
2054 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2057 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2058 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2059 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2060 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2061 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2062 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2063 Remove the {q} suffix from lea.
2064 (*tls_dynamic_gnu2_call_64): Renamed to ...
2065 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2066 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2067 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2068 Pass Pmode to gen_tls_dynamic_gnu2_64.
2070 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2072 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2074 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2076 * config/aarch64/aarch64-sve-builtins-base.cc
2077 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2079 2020-01-20 Richard Biener <rguenther@suse.de>
2082 * dwarf2out.c (prune_unused_types): Unconditionally mark
2083 called function DIEs.
2085 2020-01-20 Martin Liska <mliska@suse.cz>
2087 PR tree-optimization/93199
2088 * tree-eh.c (struct leh_state): Add
2089 new field outer_non_cleanup.
2090 (cleanup_is_dead_in): Pass leh_state instead
2091 of eh_region. Add a checking that state->outer_non_cleanup
2092 points to outer non-clean up region.
2093 (lower_try_finally): Record outer_non_cleanup
2095 (lower_catch): Likewise.
2096 (lower_eh_filter): Likewise.
2097 (lower_eh_must_not_throw): Likewise.
2098 (lower_cleanup): Likewise.
2100 2020-01-20 Richard Biener <rguenther@suse.de>
2102 PR tree-optimization/93094
2103 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2104 (vect_transform_loop): Likewise.
2105 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2106 loop_vectorized_call to vect_transform_loop.
2107 * tree-vect-loop.c (vect_transform_loop): Pass down
2108 loop_vectorized_call to vect_loop_versioning.
2109 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2110 the earlier discovered loop_vectorized_call.
2112 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2114 * doc/contribute.texi: Update for SVN -> Git transition.
2115 * doc/install.texi: Likewise.
2117 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2120 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2121 speculative targets.
2122 (verify_speculative_call): New function
2123 (cgraph_node::verify_node): Use it.
2124 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2127 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2130 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2131 (cgraph_edge::make_direct): Remove all indirect targets.
2132 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2133 (cgraph_node::verify_node): Verify that only one call_stmt or
2134 lto_stmt_uid is set.
2135 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2137 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2138 (lto_output_ref): Simplify streaming of stmt.
2139 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2141 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2143 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2144 Mark parameter unused.
2146 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2148 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2150 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2152 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2154 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2156 * Makefile.in: Add coroutine-passes.o.
2157 * builtin-types.def (BT_CONST_SIZE): New.
2158 (BT_FN_BOOL_PTR): New.
2159 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2160 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2161 * coroutine-builtins.def: New file.
2162 * coroutine-passes.cc: New file.
2163 * function.h (struct GTY function): Add a bit to indicate that the
2164 function is a coroutine component.
2165 * internal-fn.c (expand_CO_FRAME): New.
2166 (expand_CO_YIELD): New.
2167 (expand_CO_SUSPN): New.
2168 (expand_CO_ACTOR): New.
2169 * internal-fn.def (CO_ACTOR): New.
2173 * passes.def: Add pass_coroutine_lower_builtins,
2174 pass_coroutine_early_expand_ifns.
2175 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2176 (make_pass_coroutine_early_expand_ifns): New.
2177 * doc/invoke.texi: Document the fcoroutines command line
2180 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2182 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2185 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2186 after checking the argument is a REG. Don't use REGNO (reg)
2187 again to set last_regno, reuse regno variable instead.
2189 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2191 * doc/analyzer.texi (Limitations): Add note about NaN.
2193 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2194 Sudakshina Das <sudi.das@arm.com>
2196 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2197 and valid immediate.
2198 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2199 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2200 * config/arm/constraints.md (Pg): New.
2201 * config/arm/predicates.md (long_shift_imm): New.
2202 (arm_reg_or_long_shift_imm): Likewise.
2203 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2204 (thumb2_lsll): Likewise.
2207 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2208 Sudakshina Das <sudi.das@arm.com>
2210 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2211 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2212 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2213 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2214 * config/arm/thumb2.md (thumb2_asrl): New.
2215 (thumb2_lsll): Likewise.
2217 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2219 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2222 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2224 * gdbinit.in (help-gcc-hooks): New command.
2225 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2226 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2229 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2231 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2232 correct target macro.
2234 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2236 * config/aarch64/aarch64-protos.h
2237 (aarch64_sve_ld1ro_operand_p): New.
2238 * config/aarch64/aarch64-sve-builtins-base.cc
2239 (class load_replicate): New.
2240 (class svld1ro_impl): New.
2241 (class svld1rq_impl): Change to inherit from load_replicate.
2242 (svld1ro): New sve intrinsic function base.
2243 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2244 New DEF_SVE_FUNCTION.
2245 * config/aarch64/aarch64-sve-builtins-base.h
2246 (svld1ro): New decl.
2247 * config/aarch64/aarch64-sve-builtins.cc
2248 (function_expander::add_mem_operand): Modify assert to allow
2250 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2252 * config/aarch64/aarch64.c
2253 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2254 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2255 (aarch64_sve_ld1ro_operand_p): New.
2256 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2257 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2258 * config/aarch64/predicates.md
2259 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2261 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2263 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2264 Introduce this ACLE specified predefined macro.
2265 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2266 (fp): Disabling this disables f64mm.
2267 (simd): Disabling this disables f64mm.
2268 (fp16): Disabling this disables f64mm.
2269 (sve): Disabling this disables f64mm.
2270 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2271 (AARCH64_ISA_F64MM): New.
2272 (TARGET_F64MM): New.
2273 * doc/invoke.texi (f64mm): Document new option.
2275 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2277 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2278 (neoversen1_tunings): Likewise.
2280 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2283 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2284 Add assert to ensure prolog has been emitted.
2285 (aarch64_split_atomic_op): Likewise.
2286 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2287 Use epilogue_completed rather than reload_completed.
2288 (aarch64_atomic_exchange<mode>): Likewise.
2289 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2290 (atomic_nand<mode>): Likewise.
2291 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2292 (atomic_fetch_nand<mode>): Likewise.
2293 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2294 (atomic_nand_fetch<mode>): Likewise.
2296 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2299 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2301 (REVERSE_CONDITION): Delete.
2302 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2303 (CCFP_CCFPE): Likewise.
2304 (e): New mode attribute.
2305 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2306 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2307 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2308 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2309 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2310 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2311 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2312 name of generator from gen_ccmpdi to gen_ccmpccdi.
2313 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2314 the previous comparison but aren't able to, use the new ccmp_rev
2317 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2319 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2320 than testing directly for INTEGER_CST.
2321 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2323 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2325 PR tree-optimization/93292
2326 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2327 get_vectype_for_scalar_type returns NULL.
2329 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2331 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2332 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2334 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2336 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2338 * params.opt: (max-predicted-iterations): Set bounds.
2339 * predict.c (real_almost_one, real_br_prob_base,
2340 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2341 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2342 probabilities; do not truncate to reg_br_prob_bases.
2343 (estimate_loops_at_level): Pass max_cyclic_prob.
2344 (estimate_loops): Compute max_cyclic_prob.
2345 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2347 * profile-count.c (profile_probability::to_sreal): New.
2348 * profile-count.h (class sreal): Move up in file.
2349 (profile_probability::to_sreal): Declare.
2351 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2354 (arm_invalid_conversion): New function for target hook.
2355 (arm_invalid_unary_op): New function for target hook.
2356 (arm_invalid_binary_op): New function for target hook.
2358 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2360 * config.gcc: Add arm_bf16.h.
2361 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2362 (arm_simd_builtin_std_type): Add BFmode.
2363 (arm_init_simd_builtin_types): Define element types for vector types.
2364 (arm_init_bf16_types): New function.
2365 (arm_init_builtins): Add arm_init_bf16_types function call.
2366 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2367 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2368 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2369 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2370 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2371 (arm_mangle_type): Add __bf16.
2372 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2373 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2374 arm_bf16_ptr_type_node.
2375 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2376 define_split between ARM registers.
2377 * config/arm/arm_bf16.h: New file.
2378 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2379 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2381 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2382 * config/arm/vfp.md: Add BFmode to movhf patterns.
2384 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2385 Andre Vieira <andre.simoesdiasvieira@arm.com>
2387 * config/arm/arm-cpus.in (mve, mve_float): New features.
2388 (dsp, mve, mve.fp): New options.
2389 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2390 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2391 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2393 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2394 Thomas Preud'homme <thomas.preudhomme@arm.com>
2396 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2398 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2399 error for using -mcmse when targeting Armv8.1-M Mainline.
2401 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2402 Thomas Preud'homme <thomas.preudhomme@arm.com>
2404 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2405 address in r4 when targeting Armv8.1-M Mainline.
2406 (nonsecure_call_value_internal): Likewise.
2407 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2408 a register match_operand again. Emit BLXNS when targeting
2410 (nonsecure_call_value_reg_thumb2): Likewise.
2412 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2413 Thomas Preud'homme <thomas.preudhomme@arm.com>
2415 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2416 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2417 variable as true when floating-point ABI is not hard. Replace
2418 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2419 Generate VLSTM and VLLDM instruction respectively before and
2420 after a function call to cmse_nonsecure_call function.
2421 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2422 (VUNSPEC_VLLDM): Likewise.
2423 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2424 (lazy_load_multiple_insn): Likewise.
2426 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2427 Thomas Preud'homme <thomas.preudhomme@arm.com>
2429 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2430 (arm_emit_vfp_multi_reg_pop): Likewise.
2431 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2432 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2433 restore callee-saved VFP registers.
2435 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2436 Thomas Preud'homme <thomas.preudhomme@arm.com>
2438 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2439 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2440 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2441 callee-saved GPRs as well as clear ip register before doing a nonsecure
2442 call then restore callee-saved GPRs after it when targeting
2444 (arm_reorg): Adapt to function rename.
2446 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2447 Thomas Preud'homme <thomas.preudhomme@arm.com>
2449 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2450 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2451 clear_vfp_multiple pattern based on a new vfp parameter.
2452 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2453 targeting Armv8.1-M Mainline.
2454 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2455 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2456 whether VFP registers are available before looking call_used_regs for a
2458 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2459 of prototype of clear_operation_p.
2460 (clear_vfp_multiple_operation): New predicate.
2461 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2462 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2464 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2465 Thomas Preud'homme <thomas.preudhomme@arm.com>
2467 * config/arm/arm-protos.h (clear_operation_p): Declare.
2468 * config/arm/arm.c (clear_operation_p): New function.
2469 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2470 targeting Armv8.1-M Mainline or successor.
2471 (output_return_instruction): Only output APSR register clearing if
2472 Armv8.1-M Mainline instructions not available.
2473 (thumb_exit): Likewise.
2474 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2475 * config/arm/thumb2.md (clear_apsr): New define_insn.
2476 (clear_multiple): Likewise.
2477 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2479 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2480 Thomas Preud'homme <thomas.preudhomme@arm.com>
2482 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2483 (use_return_insn): Also return false for Armv8.1-M Mainline.
2484 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2485 Mainline instructions are available.
2486 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2487 when targeting Armv8.1-M Mainline Security Extensions.
2488 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2489 Mainline entry function.
2490 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2491 targeting Armv8.1-M Mainline or successor.
2492 (arm_expand_epilogue): Fix indentation of caller-saved register
2493 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2495 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2496 (FP_SYSREGS): Likewise.
2497 (enum vfp_sysregs_encoding): Define enum.
2498 (fp_sysreg_names): Declare.
2499 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2500 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2501 (pop_fpsysreg_insn): Likewise.
2503 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2504 Thomas Preud'homme <thomas.preudhomme@arm.com>
2506 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2507 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2508 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2509 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2510 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2511 (ARMv8_1m_main): New feature group.
2512 (armv8.1-m.main): New architecture.
2513 * config/arm/arm-tables.opt: Regenerate.
2514 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2515 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2516 (arm_options_perform_arch_sanity_checks): Error out when targeting
2517 Armv8.1-M Mainline Security Extensions.
2518 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2520 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2522 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2523 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2524 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2525 aarch64_bfdot_laneq): New.
2526 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2527 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2528 vbfdotq_laneq_f32): New.
2529 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2530 VBFMLA_W, VBF): New.
2531 (isquadop): Add V4BF, V8BF.
2533 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2535 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2536 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2537 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2538 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2539 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2540 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2541 usdot_laneq, sudot_lane,sudot_laneq): New.
2542 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2543 (aarch64_<sur>dot_lane): New.
2544 * config/aarch64/arm_neon.h (vusdot_s32): New.
2546 (vusdot_lane_s32): New.
2547 (vsudot_lane_s32): New.
2548 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2549 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2551 2020-01-16 Martin Liska <mliska@suse.cz>
2553 * value-prof.c (dump_histogram_value): Fix
2554 obvious spacing issue.
2556 2020-01-16 Andrew Pinski <apinski@marvell.com>
2558 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2559 !storage_order_barrier_p.
2561 2020-01-16 Andrew Pinski <apinski@marvell.com>
2563 * sched-int.h (_dep): Add unused bit-field field for the padding.
2564 * sched-deps.c (init_dep_1): Init unused field.
2566 2020-01-16 Andrew Pinski <apinski@marvell.com>
2568 * optabs.h (create_expand_operand): Initialize target field also.
2570 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2572 PR tree-optimization/92429
2573 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2574 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2576 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2579 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2581 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2582 aarch64_sve_int_mode to each mode.
2584 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2586 * doc/analyzer.texi (Overview): Add note about
2587 -fdump-ipa-analyzer.
2589 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2591 PR tree-optimization/93231
2592 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2593 input_type is unsigned. Use tree_to_shwi for shift constant.
2594 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2595 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2596 inputs more efficiently.
2598 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2600 * config/i386/i386.md (*movsf_internal): Do not require
2601 SSE2 ISA for alternatives 14 and 15.
2603 2020-01-15 Richard Biener <rguenther@suse.de>
2606 * tree-eh.c (sink_clobbers): If we already visited the destination
2607 block do not defer insertion.
2608 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2609 the purpose of defered insertion.
2611 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2613 * BASE-VER: Bump to 10.0.1.
2615 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2617 PR tree-optimization/93247
2618 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2619 type of the stmt that we're going to vectorize.
2621 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2623 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2624 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2627 2020-01-15 Martin Liska <mliska@suse.cz>
2629 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2630 2 calls of streamer_read_hwi in a function call.
2632 2020-01-15 Richard Biener <rguenther@suse.de>
2634 * alias.c (record_alias_subset): Avoid redundant work when
2635 subset is already recorded.
2637 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2639 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2640 the analyzer options provide CWE identifiers.
2642 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2644 * tree-diagnostic-path.cc (path_summary::event_range::print):
2645 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2646 using get_pure_location.
2648 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2650 PR tree-optimization/93262
2651 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2652 perform head trimming only if the last argument is constant,
2653 either all ones, or larger or equal to head trim, in the latter
2654 case decrease the last argument by head_trim.
2656 PR tree-optimization/93249
2657 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2658 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2659 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2660 perform head trim unless we can prove there are no '\0' chars
2661 from the source among the first head_trim chars.
2663 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2665 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2667 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2670 * config/i386/sse.md
2671 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2672 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2673 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2674 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2675 just a single alternative instead of two, make operands 1 and 2
2678 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2681 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2684 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2686 * Makefile.in (lang_opt_files): Add analyzer.opt.
2687 (ANALYZER_OBJS): New.
2688 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2689 tristate.o and ANALYZER_OBJS.
2690 (TEXI_GCCINT_FILES): Add analyzer.texi.
2691 * common.opt (-fanalyzer): New driver option.
2692 * config.in: Regenerate.
2693 * configure: Regenerate.
2694 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2695 (gccdepdir): Also create depdir for "analyzer" subdir.
2696 * digraph.cc: New file.
2697 * digraph.h: New file.
2698 * doc/analyzer.texi: New file.
2699 * doc/gccint.texi ("Static Analyzer") New menu item.
2700 (analyzer.texi): Include it.
2701 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2702 ("Warning Options"): Add static analysis warnings to the list.
2703 (-Wno-analyzer-double-fclose): New option.
2704 (-Wno-analyzer-double-free): New option.
2705 (-Wno-analyzer-exposure-through-output-file): New option.
2706 (-Wno-analyzer-file-leak): New option.
2707 (-Wno-analyzer-free-of-non-heap): New option.
2708 (-Wno-analyzer-malloc-leak): New option.
2709 (-Wno-analyzer-possible-null-argument): New option.
2710 (-Wno-analyzer-possible-null-dereference): New option.
2711 (-Wno-analyzer-null-argument): New option.
2712 (-Wno-analyzer-null-dereference): New option.
2713 (-Wno-analyzer-stale-setjmp-buffer): New option.
2714 (-Wno-analyzer-tainted-array-index): New option.
2715 (-Wno-analyzer-use-after-free): New option.
2716 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2717 (-Wno-analyzer-use-of-uninitialized-value): New option.
2718 (-Wanalyzer-too-complex): New option.
2719 (-fanalyzer-call-summaries): New warning.
2720 (-fanalyzer-checker=): New warning.
2721 (-fanalyzer-fine-grained): New warning.
2722 (-fno-analyzer-state-merge): New warning.
2723 (-fno-analyzer-state-purge): New warning.
2724 (-fanalyzer-transitivity): New warning.
2725 (-fanalyzer-verbose-edges): New warning.
2726 (-fanalyzer-verbose-state-changes): New warning.
2727 (-fanalyzer-verbosity=): New warning.
2728 (-fdump-analyzer): New warning.
2729 (-fdump-analyzer-callgraph): New warning.
2730 (-fdump-analyzer-exploded-graph): New warning.
2731 (-fdump-analyzer-exploded-nodes): New warning.
2732 (-fdump-analyzer-exploded-nodes-2): New warning.
2733 (-fdump-analyzer-exploded-nodes-3): New warning.
2734 (-fdump-analyzer-supergraph): New warning.
2735 * doc/sourcebuild.texi (dg-require-dot): New.
2736 (dg-check-dot): New.
2737 * gdbinit.in (break-on-saved-diagnostic): New command.
2738 * graphviz.cc: New file.
2739 * graphviz.h: New file.
2740 * ordered-hash-map-tests.cc: New file.
2741 * ordered-hash-map.h: New file.
2742 * passes.def (pass_analyzer): Add before
2743 pass_ipa_whole_program_visibility.
2744 * selftest-run-tests.c (selftest::run_tests): Call
2745 selftest::ordered_hash_map_tests_cc_tests.
2746 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2748 * shortest-paths.h: New file.
2749 * timevar.def (TV_ANALYZER): New timevar.
2750 (TV_ANALYZER_SUPERGRAPH): Likewise.
2751 (TV_ANALYZER_STATE_PURGE): Likewise.
2752 (TV_ANALYZER_PLAN): Likewise.
2753 (TV_ANALYZER_SCC): Likewise.
2754 (TV_ANALYZER_WORKLIST): Likewise.
2755 (TV_ANALYZER_DUMP): Likewise.
2756 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2757 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2758 * tree-pass.h (make_pass_analyzer): New decl.
2759 * tristate.cc: New file.
2760 * tristate.h: New file.
2762 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2765 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2766 alternatives 9 and 10.
2768 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2770 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2771 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2772 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2773 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2774 (selftest::hash_map_tests_c_tests): Call it.
2775 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2776 New static constant, using the value of = H::empty_zero_p.
2777 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2778 from default_hash_traits <Value>.
2779 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2781 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2782 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2783 calls to mark_empty with !Descriptor::empty_zero_p.
2784 (hash_table::empty_slow): Conditionalize the memset call with a
2785 check that Descriptor::empty_zero_p; otherwise, loop through the
2786 entries calling mark_empty on them.
2787 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2788 (pointer_hash::empty_zero_p): Likewise.
2789 (pair_hash::empty_zero_p): Likewise.
2790 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2792 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2793 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2794 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2795 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2796 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2797 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2798 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2799 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2801 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2804 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2806 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2807 fix typo on return value.
2809 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2812 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2814 (cgraph_edge::make_speculative): Add param for setting speculative_id
2816 (cgraph_edge::speculative_call_info): Update comments and find reference
2817 by speculative_id for multiple indirect targets.
2818 (cgraph_edge::resolve_speculation): Decrease the speculations
2819 for indirect edge, drop it's speculative if not direct target
2820 left. Update comments.
2821 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2822 (cgraph_node::dump): Print num_speculative_call_targets.
2823 (cgraph_node::verify_node): Don't report error if speculative
2824 edge not include statement.
2825 (cgraph_edge::num_speculative_call_targets_p): New function.
2826 * cgraph.h (int common_target_id): Remove.
2827 (int common_target_probability): Remove.
2828 (num_speculative_call_targets): New variable.
2829 (make_speculative): Add param for setting speculative_id.
2830 (cgraph_edge::num_speculative_call_targets_p): New declare.
2831 (target_prob): New variable.
2832 (speculative_id): New variable.
2833 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2834 call summaries for multiple speculative call targets.
2835 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2836 * ipa-profile.c (struct speculative_call_target): New struct.
2837 (class speculative_call_summary): New class.
2838 (class speculative_call_summaries): New class.
2839 (call_sums): New variable.
2840 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2841 (ipa_profile_write_edge_summary): New function.
2842 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2843 (ipa_profile_dump_all_summaries): New function.
2844 (ipa_profile_read_edge_summary): New function.
2845 (ipa_profile_read_summary_section): New function.
2846 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2847 (ipa_profile): Generate num_speculative_call_targets from
2849 * ipa-ref.h (speculative_id): New variable.
2850 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2851 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2852 common_target_probability. Stream out speculative_id and
2853 num_speculative_call_targets.
2854 (input_edge): Likewise.
2855 * predict.c (dump_prediction): Remove edges count assert to be
2857 * symtab.c (symtab_node::create_reference): Init speculative_id.
2858 (symtab_node::clone_references): Clone speculative_id.
2859 (symtab_node::clone_referring): Clone speculative_id.
2860 (symtab_node::clone_reference): Clone speculative_id.
2861 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2862 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2863 if indirect call contains multiple speculative targets.
2864 * value-prof.h (check_ic_target): Remove.
2865 * value-prof.c (gimple_value_profile_transformations):
2866 Use void function gimple_ic_transform.
2867 * value-prof.c (gimple_ic_transform): Handle topn case.
2868 Fix comment typos. Change it to a void function.
2870 2020-01-13 Andrew Pinski <apinski@marvell.com>
2872 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2873 (octeontx2t98): New define.
2874 (octeontx2t96): New define.
2875 (octeontx2t93): New define.
2876 (octeontx2f95): New define.
2877 (octeontx2f95n): New define.
2878 (octeontx2f95mm): New define.
2879 * config/aarch64/aarch64-tune.md: Regenerate.
2880 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2882 2020-01-13 Jason Merrill <jason@redhat.com>
2884 PR c++/33799 - destroy return value if local cleanup throws.
2885 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2887 2020-01-13 Martin Liska <mliska@suse.cz>
2889 * ipa-cp.c (get_max_overall_size): Use newly
2890 renamed param param_ipa_cp_unit_growth.
2891 * params.opt: Remove legacy param name.
2893 2020-01-13 Martin Sebor <msebor@redhat.com>
2895 PR tree-optimization/93213
2896 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2897 stores to be eliminated.
2899 2020-01-13 Martin Liska <mliska@suse.cz>
2901 * opts.c (print_help): Do not print CL_PARAM
2902 and CL_WARNING for CL_OPTIMIZATION.
2904 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2907 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2908 depending on optimization settings.
2910 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2912 PR tree-optimization/90838
2913 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2914 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2915 argument rather than to initialize temporary for targets that
2916 don't use the mode argument at all. Initialize ctzval to avoid
2919 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2921 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2922 * tree-core.h: Document it.
2923 * gimplify.c (gimplify_omp_workshare): Set it.
2924 * omp-low.c (lower_omp_target): Use it.
2925 * tree-pretty-print.c (dump_omp_clause): Print it.
2927 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2928 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2930 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2932 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2933 * common.opt (fdiagnostics-path-format=): New option.
2934 (diagnostic_path_format): New enum.
2935 (fdiagnostics-show-path-depths): New option.
2936 * coretypes.h (diagnostic_event_id_t): New forward decl.
2937 * diagnostic-color.c (color_dict): Add "path".
2938 * diagnostic-event-id.h: New file.
2939 * diagnostic-format-json.cc (json_from_expanded_location): Make
2941 (json_end_diagnostic): Call context->make_json_for_path if it
2942 exists and the diagnostic has a path.
2943 (diagnostic_output_format_init): Clear context->print_path.
2944 * diagnostic-path.h: New file.
2945 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2946 when printing a run of events in a diagnostic_path so that they
2947 all get the same color.
2948 (layout::m_diagnostic_path_p): New field.
2949 (layout::layout): Initialize it.
2950 (layout::print_any_labels): Don't colorize the label text for an
2951 event in a diagnostic_path.
2952 (gcc_rich_location::add_location_if_nearby): Add
2953 "restrict_to_current_line_spans" and "label" params. Pass the
2954 former to layout.maybe_add_location_range; pass the latter
2955 when calling add_range.
2956 * diagnostic.c: Include "diagnostic-path.h".
2957 (diagnostic_initialize): Initialize context->path_format and
2958 context->show_path_depths.
2959 (diagnostic_show_any_path): New function.
2960 (diagnostic_path::interprocedural_p): New function.
2961 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2962 (simple_diagnostic_path::num_events): New function.
2963 (simple_diagnostic_path::get_event): New function.
2964 (simple_diagnostic_path::add_event): New function.
2965 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2966 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2967 (debug): New overload taking a diagnostic_path *.
2968 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2969 * diagnostic.h (enum diagnostic_path_format): New enum.
2970 (json::value): New forward decl.
2971 (diagnostic_context::path_format): New field.
2972 (diagnostic_context::show_path_depths): New field.
2973 (diagnostic_context::print_path): New callback field.
2974 (diagnostic_context::make_json_for_path): New callback field.
2975 (diagnostic_show_any_path): New decl.
2976 (json_from_expanded_location): New decl.
2977 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2978 (-fdiagnostics-show-path-depths): New option.
2979 (-fdiagnostics-color): Add "path" to description of default
2980 GCC_COLORS; describe it.
2981 (-fdiagnostics-format=json): Document how diagnostic paths are
2982 represented in the JSON output format.
2983 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2984 Add optional params "restrict_to_current_line_spans" and "label".
2985 * opts.c (common_handle_option): Handle
2986 OPT_fdiagnostics_path_format_ and
2987 OPT_fdiagnostics_show_path_depths.
2988 * pretty-print.c: Include "diagnostic-event-id.h".
2989 (pp_format): Implement "%@" format code for printing
2990 diagnostic_event_id_t *.
2991 (selftest::test_pp_format): Add tests for "%@".
2992 * selftest-run-tests.c (selftest::run_tests): Call
2993 selftest::tree_diagnostic_path_cc_tests.
2994 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2995 * toplev.c (general_init): Initialize global_dc->path_format and
2996 global_dc->show_path_depths.
2997 * tree-diagnostic-path.cc: New file.
2998 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2999 non-static. Drop "diagnostic" param in favor of storing the
3000 original value of "where" and re-using it.
3001 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3002 maybe_unwind_expanded_macro_loc.
3003 (tree_diagnostics_defaults): Initialize context->print_path and
3004 context->make_json_for_path.
3005 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3007 (default_tree_make_json_for_path): New decl.
3008 (maybe_unwind_expanded_macro_loc): New decl.
3010 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3012 PR tree-optimization/93210
3013 * fold-const.h (native_encode_initializer,
3014 can_native_interpret_type_p): Declare.
3015 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3017 (native_encode_initializer): New function, moved from dwarf2out.c.
3018 Adjust to native_encode_expr compatible arguments, including dry-run
3019 and partial extraction modes. Don't handle STRING_CST.
3020 (can_native_interpret_type_p): No longer static.
3021 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3022 offset / BITS_PER_UNIT fits into int and don't call it if
3023 can_native_interpret_type_p fails. If suboff is NULL and for
3024 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3025 native_encode_initializer.
3026 (fold_const_aggregate_ref_1): Formatting fix.
3027 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3028 (tree_add_const_value_attribute): Adjust caller.
3030 PR tree-optimization/90838
3031 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3032 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3033 CTZ_DEFINED_VALUE_AT_ZERO.
3035 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3038 * lra-constraints.c (match_reload): Permit input operands have the
3039 same mode as output while other input operands have a different
3042 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3044 PR tree-optimization/90838
3045 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3046 (check_ctz_string): Likewise.
3047 (optimize_count_trailing_zeroes): Likewise.
3048 (simplify_count_trailing_zeroes): Likewise.
3049 (pass_forwprop::execute): Try ctz simplification.
3050 * match.pd: Add matching for ctz idioms.
3052 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3054 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3056 (aarch64_invalid_unary_op): New function for target hook.
3057 (aarch64_invalid_binary_op): New function for target hook.
3059 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3061 * config.gcc: Add arm_bf16.h.
3062 * config/aarch64/aarch64-builtins.c
3063 (aarch64_simd_builtin_std_type): Add BFmode.
3064 (aarch64_init_simd_builtin_types): Define element types for vector
3066 (aarch64_init_bf16_types): New function.
3067 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3068 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3070 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3071 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3073 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3074 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3075 * config/aarch64/aarch64.c
3076 (aarch64_classify_vector_mode): Add support for BF types.
3077 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3078 (aarch64_vq_mode): Add support for BF types.
3079 (aarch64_simd_container_mode): Add support for BF types.
3080 (aarch64_mangle_type): Add support for BF scalar type.
3081 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3082 * config/aarch64/arm_bf16.h: New file.
3083 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3084 * config/aarch64/iterators.md: Add BF types to mode attributes.
3085 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3087 2020-01-10 Jason Merrill <jason@redhat.com>
3089 PR c++/93173 - incorrect tree sharing.
3090 * gimplify.c (copy_if_shared): No longer static.
3091 * gimplify.h: Declare it.
3093 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3095 * doc/invoke.texi (-msve-vector-bits=): Document that
3096 -msve-vector-bits=128 now generates VL-specific code for
3097 little-endian targets.
3098 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3099 build_vector_type_for_mode to construct the data vector types.
3100 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3101 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3102 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3103 for 128-bit vectors.
3105 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3107 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3110 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3112 * config/aarch64/aarch64-builtins.c
3113 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3114 rather than checking the number of elements and the element mode.
3116 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3118 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3119 get_related_vectype_for_scalar_type rather than build_vector_type
3120 to create the index type for a conditional reduction.
3122 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3124 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3125 for any type of gather or scatter, including strided accesses.
3127 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3129 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3132 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3134 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3136 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3137 parameter and its use to reset DR_OFFSET's.
3138 (vect_transform_loop): Remove orig_drs_init argument.
3139 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3140 member of dr_vec_info rather than the offset of the associated
3141 data_reference's innermost_loop_behavior.
3142 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3143 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3144 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3145 get_dr_vinfo_offset.
3146 (vectorizable_store): Likewise.
3147 (vectorizable_load): Likewise.
3149 2020-01-10 Richard Biener <rguenther@suse.de>
3151 * gimple-ssa-store-merging
3152 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3154 2020-01-10 Martin Liska <mliska@suse.cz>
3157 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3158 encapsulation that was there before r280040.
3160 2020-01-10 Richard Biener <rguenther@suse.de>
3163 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3164 sequences to avoid walking them again for secondary opportunities.
3165 (pass_lower_eh_dispatch::execute): Instead actually insert
3168 2020-01-10 Richard Biener <rguenther@suse.de>
3171 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3172 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3173 avoid quadraticness.
3175 2020-01-10 Martin Jambor <mjambor@suse.cz>
3177 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3178 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3179 to get param_ipa_sra_max_replacements.
3180 (param_splitting_across_edge): Pass the caller to
3181 pull_accesses_from_callee.
3183 2020-01-10 Martin Jambor <mjambor@suse.cz>
3185 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3186 * ipa-cp.c (max_new_size): Removed.
3187 (orig_overall_size): New variable.
3188 (get_max_overall_size): New function.
3189 (estimate_local_effects): Use it. Adjust dump.
3190 (decide_about_value): Likewise.
3191 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3192 orig_overall_size. Adjust dump.
3193 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3195 2020-01-10 Martin Jambor <mjambor@suse.cz>
3197 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3198 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3199 instead of param_ipa_max_agg_items.
3200 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3201 optimization info for the callee.
3203 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3205 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3206 markers if debug_inline_points is false.
3208 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3210 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3212 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3213 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3214 aarch64-sve-builtins-sve2.h.
3215 (aarch64-sve-builtins-sve2.o): New rule.
3216 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3217 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3218 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3219 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3220 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3221 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3223 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3224 instructions that are handled here.
3225 (@cond_asrd<mode>): Generalize to...
3226 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3227 (*cond_asrd<mode>_2): Generalize to...
3228 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3229 (*cond_asrd<mode>_z): Generalize to...
3230 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3231 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3232 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3233 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3234 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3236 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3237 (@aarch64_scatter_stnt<mode>): Likewise.
3238 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3239 (@aarch64_mul_lane_<mode>): Likewise.
3240 (@aarch64_sve_suqadd<mode>_const): Likewise.
3241 (*<sur>h<addsub><mode>): Generalize to...
3242 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3244 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3245 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3246 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3247 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3248 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3249 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3250 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3251 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3252 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3253 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3254 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3255 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3256 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3257 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3258 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3259 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3260 (@aarch64_sve2_xar<mode>): Likewise.
3261 (@aarch64_sve2_bcax<mode>): Likewise.
3262 (*aarch64_sve2_eor3<mode>): Rename to...
3263 (@aarch64_sve2_eor3<mode>): ...this.
3264 (@aarch64_sve2_bsl<mode>): New expander.
3265 (@aarch64_sve2_nbsl<mode>): Likewise.
3266 (@aarch64_sve2_bsl1n<mode>): Likewise.
3267 (@aarch64_sve2_bsl2n<mode>): Likewise.
3268 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3269 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3270 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3271 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3272 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3273 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3274 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3275 (<su>mull<bt><Vwide>): Generalize to...
3276 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3278 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3279 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3280 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3281 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3282 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3283 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3284 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3285 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3286 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3287 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3288 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3289 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3290 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3291 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3292 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3293 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3294 (<SHRNB:r>shrnb<mode>): Generalize to...
3295 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3297 (<SHRNT:r>shrnt<mode>): Generalize to...
3298 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3300 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3301 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3302 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3303 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3304 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3305 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3306 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3307 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3308 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3309 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3310 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3311 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3312 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3313 (@aarch64_sve2_cvtnt<mode>): Likewise.
3314 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3315 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3316 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3317 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3318 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3319 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3320 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3321 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3322 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3323 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3324 (@aarch64_sve2_pmul<mode>): Likewise.
3325 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3326 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3327 (@aarch64_sve2_tbl2<mode>): Likewise.
3328 (@aarch64_sve2_tbx<mode>): Likewise.
3329 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3330 (@aarch64_sve2_histcnt<mode>): Likewise.
3331 (@aarch64_sve2_histseg<mode>): Likewise.
3332 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3333 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3334 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3335 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3336 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3337 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3338 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3339 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3340 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3341 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3342 (SVE2_PMULL_PAIR_I): New mode iterators.
3343 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3344 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3345 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3346 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3347 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3348 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3349 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3350 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3351 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3352 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3353 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3354 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3355 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3356 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3357 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3358 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3359 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3360 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3361 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3362 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3363 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3364 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3365 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3366 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3367 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3368 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3369 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3370 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3371 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3372 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3373 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3375 (VNARROW, Ventype): New mode attributes.
3376 (Vewtype): Handle VNx2DI. Fix typo in comment.
3377 (VDOUBLE): New mode attribute.
3378 (sve_lane_con): Handle VNx8HI.
3379 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3380 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3381 (sve_int_op, sve_int_op_rev): Handle the above codes.
3382 (sve_pred_int_rhs2_operand): Likewise.
3383 (MULLBT, SHRNB, SHRNT): Delete.
3384 (SVE_INT_SHIFT_IMM): New int iterator.
3385 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3386 and UNSPEC_WHILEHS for TARGET_SVE2.
3387 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3388 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3389 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3390 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3391 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3392 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3393 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3394 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3395 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3396 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3397 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3398 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3399 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3400 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3401 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3402 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3403 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3404 (optab): Handle the new unspecs.
3405 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3407 (lr): Handle the new unspecs.
3409 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3410 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3411 (sve_int_qsub_op): New int attributes.
3412 (sve_fp_op, rot): Handle the new unspecs.
3413 * config/aarch64/aarch64-sve-builtins.h
3414 (function_resolver::require_matching_pointer_type): Declare.
3415 (function_resolver::resolve_unary): Add an optional boolean argument.
3416 (function_resolver::finish_opt_n_resolution): Add an optional
3417 type_suffix_index argument.
3418 (gimple_folder::redirect_call): Declare.
3419 (gimple_expander::prepare_gather_address_operands): Add an optional
3421 * config/aarch64/aarch64-sve-builtins.cc: Include
3422 aarch64-sve-builtins-sve2.h.
3423 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3424 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3425 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3426 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3427 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3428 (TYPES_s_unsigned): Likewise.
3429 (TYPES_s_integer): Use TYPES_s_unsigned.
3430 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3431 (TYPES_sd_integer): Use them.
3432 (TYPES_d_unsigned): New macro.
3433 (TYPES_d_integer): Use it.
3434 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3435 (TYPES_cvt_narrow): Likewise.
3436 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3437 (preds_mx): New variable.
3438 (function_builder::add_overloaded_function): Allow the new feature
3439 set to be more restrictive than the original one.
3440 (function_resolver::infer_pointer_type): Remove qualifiers from
3441 the pointer type before printing it.
3442 (function_resolver::require_matching_pointer_type): New function.
3443 (function_resolver::resolve_sv_displacement): Handle functions
3444 that don't support 32-bit vector indices or svint32_t vector offsets.
3445 (function_resolver::finish_opt_n_resolution): Take the inferred type
3446 as a separate argument.
3447 (function_resolver::resolve_unary): Optionally treat all forms in
3448 the same way as normal merging functions.
3449 (gimple_folder::redirect_call): New function.
3450 (function_expander::prepare_gather_address_operands): Add an argument
3451 that says whether scaled forms are available. If they aren't,
3452 handle scaling of vector indices and don't add the extension and
3454 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3455 fall back to using cond_* instead.
3456 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3457 Split out the member variables into...
3458 (rtx_code_function_base): ...this new base class.
3459 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3460 (unspec_based_function): Split out the member variables into...
3461 (unspec_based_function_base): ...this new base class.
3462 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3463 (unspec_based_function_exact_insn): New class.
3464 (unspec_based_add_function, unspec_based_add_lane_function)
3465 (unspec_based_lane_function, unspec_based_pred_function)
3466 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3467 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3468 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3470 (unspec_based_fused_function): New class.
3471 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3472 (unspec_based_fused_lane_function): New class.
3473 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3475 (CODE_FOR_MODE1): New macro.
3476 (fixed_insn_function): New class.
3477 (while_comparison): Likewise.
3478 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3479 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3480 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3481 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3482 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3483 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3484 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3485 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3486 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3487 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3488 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3489 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3490 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3491 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3492 Also add an initial argument for unary_convert_narrowt, regardless
3493 of the predication type.
3494 (build_32_64): Allow loads and stores to specify MODE_none.
3495 (build_sv_index64, build_sv_uint_offset): New functions.
3496 (long_type_suffix): New function.
3497 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3498 (binary_imm_long_base, load_gather_sv_base): Likewise.
3499 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3500 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3501 (unary_narrowb_base, unary_narrowt_base): Likewise.
3502 (binary_long_lane_def, binary_long_lane): New shape.
3503 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3504 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3505 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3506 (binary_to_uint_def, binary_to_uint): Likewise.
3507 (binary_wide_def, binary_wide): Likewise.
3508 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3509 (compare_def, compare): Likewise.
3510 (compare_ptr_def, compare_ptr): Likewise.
3511 (load_ext_gather_index_restricted_def,
3512 load_ext_gather_index_restricted): Likewise.
3513 (load_ext_gather_offset_restricted_def,
3514 load_ext_gather_offset_restricted): Likewise.
3515 (load_gather_sv_def): Inherit from load_gather_sv_base.
3516 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3517 (shift_left_imm_def, shift_left_imm): Likewise.
3518 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3519 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3520 (store_scatter_index_restricted_def,
3521 store_scatter_index_restricted): Likewise.
3522 (store_scatter_offset_restricted_def,
3523 store_scatter_offset_restricted): Likewise.
3524 (tbl_tuple_def, tbl_tuple): Likewise.
3525 (ternary_long_lane_def, ternary_long_lane): Likewise.
3526 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3527 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3528 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3529 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3530 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3531 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3532 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3533 (ternary_uint_def, ternary_uint): Likewise.
3534 (unary_convert): Fix typo in comment.
3535 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3536 (unary_long_def, unary_long): Likewise.
3537 (unary_narrowb_def, unary_narrowb): Likewise.
3538 (unary_narrowt_def, unary_narrowt): Likewise.
3539 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3540 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3541 (unary_to_int_def, unary_to_int): Likewise.
3542 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3543 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3544 (svasrd_impl): Delete.
3545 (svcadd_impl::expand): Handle integer operations too.
3546 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3547 new functions to derive the unspec numbers.
3548 (svmla_svmls_lane_impl): Replace with...
3549 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3550 integer operations too.
3551 (svwhile_impl): Rename to...
3552 (svwhilelx_impl): ...this and inherit from while_comparison.
3553 (svasrd): Use unspec_based_function.
3554 (svmla_lane): Use svmla_lane_impl.
3555 (svmls_lane): Use svmls_lane_impl.
3556 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3557 (svwhilele, svwhilelt): Use svwhilelx_impl.
3558 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3559 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3560 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3561 * config/aarch64/aarch64-sve-builtins.def: Include
3562 aarch64-sve-builtins-sve2.def.
3564 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3566 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3567 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3568 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3569 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3570 immediates as well as vector ones.
3571 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3572 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3573 (aarch64_sve_qsub_immediate): Update calls accordingly.
3575 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3577 * config/aarch64/aarch64-sve2.md: Add banner comments.
3578 (<su>mulh<r>s<mode>3): Move further up file.
3579 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3580 (*aarch64_sve2_sra<mode>): Move further down file.
3581 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3583 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3585 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3587 (while_optab_cmp): Handle them.
3588 * config/aarch64/aarch64-sve.md
3589 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3590 and add a "@" marker.
3591 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3592 instead of gen_aarch64_sve2_while_ptest.
3593 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3595 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3597 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3598 (UNSPEC_WHILELE): ...this.
3599 (UNSPEC_WHILE_LO): Rename to...
3600 (UNSPEC_WHILELO): ...this.
3601 (UNSPEC_WHILE_LS): Rename to...
3602 (UNSPEC_WHILELS): ...this.
3603 (UNSPEC_WHILE_LT): Rename to...
3604 (UNSPEC_WHILELT): ...this.
3605 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3606 (cmp_op, while_optab_cmp): Likewise.
3607 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3608 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3609 (svwhilelt): Likewise.
3611 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3613 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3614 (unary_to_uint): Define.
3615 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3616 (unary_count): Rename to...
3617 (unary_to_uint_def, unary_to_uint): ...this.
3618 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3620 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3622 * config/aarch64/aarch64-sve-builtins-functions.h
3623 (code_for_mode_function): New class.
3624 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3625 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3626 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3627 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3628 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3630 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3632 * config/aarch64/iterators.md (addsub): New code attribute.
3633 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3635 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3636 in the asm string and attributes. Fix indentation.
3637 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3639 (@aarch64_sve_<optab><mode>): ...this.
3640 * config/aarch64/aarch64-sve-builtins.h
3641 (function_expander::expand_signed_unpred_op): Delete.
3642 * config/aarch64/aarch64-sve-builtins.cc
3643 (function_expander::expand_signed_unpred_op): Likewise.
3644 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3645 try using code_for_aarch64_sve instead.
3646 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3647 (svqsub_impl): Likewise.
3648 (svqadd, svqsub): Use rtx_code_function instead.
3650 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3652 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3653 (HADDSUB, sur, addsub): Remove them.
3655 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3657 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3658 functions the same way, rather than singling out those that
3659 aren't mapped directly to optabs.
3661 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3663 * target.def (compatible_vector_types_p): New target hook.
3664 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3665 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3666 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3667 * doc/tm.texi: Regenerate.
3668 * gimple-expr.c: Include target.h.
3669 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3670 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3672 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3673 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3674 Use the original predicate if it already has a suitable type.
3676 2020-01-09 Martin Jambor <mjambor@suse.cz>
3678 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3679 resolve_speculation and redirect_call_stmt_to_callee static. Change
3680 return type of set_call_stmt to cgraph_edge *.
3681 * auto-profile.c (afdo_indirect_call): Adjust call to
3682 redirect_call_stmt_to_callee.
3683 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3684 make the this pointer explicit, adjust self-recursive calls and the
3685 call top make_direct. Return the resulting edge.
3686 (cgraph_edge::remove): Make this pointer explicit.
3687 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3688 (cgraph_edge::make_direct): Likewise, adjust call to
3689 resolve_speculation.
3690 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3691 call to set_call_stmt.
3692 (cgraph_update_edges_for_call_stmt_node): Update call to
3693 set_call_stmt and remove.
3694 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3695 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3696 (cgraph_node::create_edge_including_clones): Moved "first" definition
3697 of edge to the block where it was used. Adjusted calls to
3699 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3700 cgraph_edge::remove.
3701 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3702 make_direct and redirect_call_stmt_to_callee.
3703 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3704 resolve_speculation and make_direct.
3705 * ipa-inline-transform.c (inline_transform): Adjust call to
3706 redirect_call_stmt_to_callee.
3707 (check_speculations_1):: Adjust call to resolve_speculation.
3708 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3709 resolve-speculation.
3710 (inline_small_functions): Adjust call to resolve_speculation.
3711 (ipa_inline): Likewise.
3712 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3714 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3715 safe with regards to edge removal, adjust calls to
3716 redirect_call_stmt_to_callee.
3717 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3718 and redirect_call_stmt_to_callee.
3719 * multiple_target.c (create_dispatcher_calls): Adjust call to
3720 redirect_call_stmt_to_callee
3721 (redirect_to_specific_clone): Likewise.
3722 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3723 Adjust calls to cgraph_edge::remove.
3724 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3725 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3726 (expand_call_inline): Adjust call to cgraph_edge::remove.
3728 2020-01-09 Martin Liska <mliska@suse.cz>
3730 * params.opt: Set Optimization for
3731 param_max_speculative_devirt_maydefs.
3733 2020-01-09 Martin Sebor <msebor@redhat.com>
3737 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3739 2020-01-09 Martin Liska <mliska@suse.cz>
3741 * auto-profile.c (auto_profile): Use opt_for_fn
3743 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3744 (propagate_vals_across_arith_jfunc): Likewise.
3745 (hint_time_bonus): Likewise.
3746 (incorporate_penalties): Likewise.
3747 (good_cloning_opportunity_p): Likewise.
3748 (perform_estimation_of_a_value): Likewise.
3749 (estimate_local_effects): Likewise.
3750 (ipcp_propagate_stage): Likewise.
3751 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3752 (set_switch_stmt_execution_predicate): Likewise.
3753 (analyze_function_body): Likewise.
3754 * ipa-inline-analysis.c (offline_size): Likewise.
3755 * ipa-inline.c (early_inliner): Likewise.
3756 * ipa-prop.c (ipa_analyze_node): Likewise.
3757 (ipcp_transform_function): Likewise.
3758 * ipa-sra.c (process_scan_results): Likewise.
3759 (ipa_sra_summarize_function): Likewise.
3760 * params.opt: Rename ipcp-unit-growth to
3761 ipa-cp-unit-growth. Add Optimization for various
3762 IPA-related parameters.
3764 2020-01-09 Richard Biener <rguenther@suse.de>
3767 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3769 2020-01-09 Richard Biener <rguenther@suse.de>
3771 PR tree-optimization/93040
3772 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3774 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3776 * common/config/avr/avr-common.c (avr_option_optimization_table)
3777 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3779 2020-01-09 Martin Liska <mliska@suse.cz>
3781 * cgraphclones.c (symbol_table::materialize_all_clones):
3782 Use cgraph_node::dump_name.
3784 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3787 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3788 output_operand_lossage instead of gcc_unreachable.
3789 * doc/md.texi (riscv f constraint): Fix typo.
3792 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3793 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3794 CONST_SCALAR_INT_P instead of CONST_INT_P.
3795 (*subv<mode>4_1): Rename to ...
3796 (subv<mode>4_1): ... this.
3797 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3798 define_insn_and_split patterns.
3799 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3802 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3804 * vec.c (class selftest::count_dtor): New class.
3805 (selftest::test_auto_delete_vec): New test.
3806 (selftest::vec_c_tests): Call it.
3807 * vec.h (class auto_delete_vec): New class template.
3808 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3810 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3812 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3814 2020-01-08 Jim Wilson <jimw@sifive.com>
3816 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3817 use of TLS_MODEL_LOCAL_EXEC when not pic.
3819 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3821 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3824 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3827 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3828 *stack_protect_set_3 peephole2): Also check that the second
3829 insns source is general_operand.
3832 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3833 predicate for output operand instead of register_operand.
3834 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3835 memory destination and non-memory operands[2].
3837 2020-01-08 Martin Liska <mliska@suse.cz>
3839 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3840 ::dump_asm_name instead of (::name or ::asm_name).
3841 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3842 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3843 (analyze_functions): Likewise.
3844 (expand_all_functions): Likewise.
3845 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3846 (propagate_bits_across_jump_function): Likewise.
3847 (dump_profile_updates): Likewise.
3848 (ipcp_store_bits_results): Likewise.
3849 (ipcp_store_vr_results): Likewise.
3850 * ipa-devirt.c (dump_targets): Likewise.
3851 * ipa-fnsummary.c (analyze_function_body): Likewise.
3852 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3853 (process_hsa_functions): Likewise.
3854 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3855 (set_alias_uids): Likewise.
3856 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3857 * ipa-inline.c (recursive_inlining): Likewise.
3858 (inline_to_all_callers_1): Likewise.
3859 (ipa_inline): Likewise.
3860 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3861 (ipa_propagate_frequency): Likewise.
3862 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3863 (remove_described_reference): Likewise.
3864 * ipa-pure-const.c (worse_state): Likewise.
3865 (check_retval_uses): Likewise.
3866 (analyze_function): Likewise.
3867 (propagate_pure_const): Likewise.
3868 (propagate_nothrow): Likewise.
3869 (dump_malloc_lattice): Likewise.
3870 (propagate_malloc): Likewise.
3871 (pass_local_pure_const::execute): Likewise.
3872 * ipa-visibility.c (optimize_weakref): Likewise.
3873 (function_and_variable_visibility): Likewise.
3874 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3875 (ipa_discover_variable_flags): Likewise.
3876 * lto-streamer-out.c (output_function): Likewise.
3877 (output_constructor): Likewise.
3878 * tree-inline.c (copy_bb): Likewise.
3879 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3880 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3882 2020-01-08 Richard Biener <rguenther@suse.de>
3885 * tree-eh.c (sink_clobbers): Update virtual operands for
3886 the first and last stmt only. Add a dry-run capability.
3887 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3888 after CFG manipulations and in RPO order to catch all
3889 secondary opportunities reliably.
3891 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3894 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3896 2019-01-08 Richard Biener <rguenther@suse.de>
3899 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3900 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3901 virtual operand, also updating SSA use.
3902 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3903 Update stmt after resetting virtual operand.
3904 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3905 * gimple-iterator.c (gsi_remove): When not removing the stmt
3906 permanently do not delink immediate uses or mark the stmt modified.
3908 2020-01-08 Martin Liska <mliska@suse.cz>
3910 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3911 (ipa_call_context::estimate_size_and_time): Likewise.
3912 (inline_analyze_function): Likewise.
3914 2020-01-08 Martin Liska <mliska@suse.cz>
3916 * cgraph.c (cgraph_node::dump): Use systematically
3919 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3921 Add -nodevicespecs option for avr.
3924 * config/avr/avr.opt (-nodevicespecs): New driver option.
3925 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3926 "-specs=device-specs/..." if that option is not set.
3927 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3929 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3931 Implement 64-bit double functions for avr.
3934 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3935 --with-double-comparison.
3936 * doc/install.texi: Document them.
3937 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3938 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3939 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3940 * doc/invoke.texi (AVR Built-in Macros): Document them.
3941 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3942 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3943 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3945 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3948 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3949 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3950 when only building rm-profile multilibs.
3952 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3955 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3956 lattice for a value to check.
3957 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3958 finite propagation in self-recursive scc.
3960 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3962 * ipa-inline.c (caller_growth_limits): Restore the AND.
3964 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3966 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3967 (VEC_ALLREG_ALT): New iterator.
3968 (VEC_ALLREG_INT_MODE): New iterator.
3969 (VCMP_MODE): New iterator.
3970 (VCMP_MODE_INT): New iterator.
3971 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3972 (vec_cmp<u>v64qidi): New define_expand.
3973 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3974 (vec_cmpu<mode>di_exec): New define_expand.
3975 (vec_cmp<u>v64qidi_exec): New define_expand.
3976 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3977 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3978 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3979 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3980 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3981 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3982 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3983 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3984 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3985 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3987 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3988 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3990 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3992 * config/gcn/constraints.md (DA): Update description and match.
3994 (Db): New constraint.
3995 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3997 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3998 Implement 'Db' mixed immediate type.
3999 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4000 (addcv64si3_dup<exec_vcc>): Delete.
4001 (subcv64si3<exec_vcc>): Rework constraints.
4002 (addv64di3): Rework constraints.
4003 (addv64di3_exec): Rework constraints.
4004 (subv64di3): Rework constraints.
4005 (addv64di3_dup): Delete.
4006 (addv64di3_dup_exec): Delete.
4007 (addv64di3_zext): Rework constraints.
4008 (addv64di3_zext_exec): Rework constraints.
4009 (addv64di3_zext_dup): Rework constraints.
4010 (addv64di3_zext_dup_exec): Rework constraints.
4011 (addv64di3_zext_dup2): Rework constraints.
4012 (addv64di3_zext_dup2_exec): Rework constraints.
4013 (addv64di3_sext_dup2): Rework constraints.
4014 (addv64di3_sext_dup2_exec): Rework constraints.
4016 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4018 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4019 existing target checks.
4021 2020-01-07 Richard Biener <rguenther@suse.de>
4023 * doc/install.texi: Bump minimal supported MPC version.
4025 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4027 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4028 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4029 * langhooks.c: Include stor-layout.h.
4030 (lhd_simulate_enum_decl): New function.
4031 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4032 handle_arm_sve_h for the LTO frontend.
4033 (register_vector_type): Cope with null returns from pushdecl.
4035 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4037 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4038 (aarch64_sve::nvectors_if_data_type): Replace with...
4039 (aarch64_sve::builtin_type_p): ...this.
4040 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4041 (find_vector_type): Delete.
4042 (add_sve_type_attribute): New function.
4043 (lookup_sve_type_attribute): Likewise.
4044 (register_builtin_types): Add an "SVE type" attribute to each type.
4045 (register_tuple_type): Likewise.
4046 (svbool_type_p, nvectors_if_data_type): Delete.
4047 (mangle_builtin_type): Use lookup_sve_type_attribute.
4048 (builtin_type_p): Likewise. Add an overload that returns the
4049 number of constituent vector and predicate registers.
4050 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4051 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4052 instead of aarch64_sve_argument_p.
4053 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4054 (aarch64_pass_by_reference): Likewise.
4055 (aarch64_function_value_1): Likewise.
4056 (aarch64_return_in_memory): Likewise.
4057 (aarch64_layout_arg): Likewise.
4059 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4061 PR tree-optimization/93156
4062 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4063 least significant bit is always clear.
4065 PR tree-optimization/93118
4066 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4067 simplifier with two intermediate conversions.
4069 2020-01-07 Martin Liska <mliska@suse.cz>
4071 * params.opt: Add Optimization for various parameters.
4073 2020-01-07 Martin Liska <mliska@suse.cz>
4076 * doc/extend.texi: Explain cloning for target_clone
4079 2020-01-07 Martin Liska <mliska@suse.cz>
4081 PR tree-optimization/92860
4082 * common.opt: Make in Optimization option
4083 as it is affected by -O0, which is an Optimization
4085 * tree-inline.c (tree_inlinable_function_p):
4086 Use opt_for_fn for warn_inline.
4087 (expand_call_inline): Likewise.
4089 2020-01-07 Martin Liska <mliska@suse.cz>
4091 PR tree-optimization/92860
4092 * common.opt: Make flag_ree as optimization
4095 2020-01-07 Martin Liska <mliska@suse.cz>
4097 PR optimization/92860
4098 * params.opt: Mark param_min_crossjump_insns with Optimization
4101 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4103 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4104 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4106 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4108 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4109 helper function to return the valid addressing formats for a given
4110 hard register and mode.
4111 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4113 * config/rs6000/constraints.md (Q constraint): Update
4115 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4118 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4119 Use 'Q' for doing vector extract from memory.
4120 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4122 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4123 doing vector extract from memory.
4124 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4125 extract from memory.
4127 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4128 for the offset being 34-bits when -mcpu=future is used.
4130 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4132 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4133 instead of cmpib_comparison_operator in cmpib patterns.
4134 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4135 of cmpib_comparison_operator. Revise comment.
4137 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4139 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4140 in an IFN_DIV_POW2 node to be equal.
4142 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4144 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4145 (vect_check_scalar_mask): ...this.
4146 (vectorizable_store, vectorizable_load): Update call accordingly.
4147 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4148 argument in calls to conditional internal functions.
4150 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4152 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4153 '0' matching inputs.
4154 (subv64di3_exec): Likewise.
4156 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4158 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4159 * doc/md.texi (movstr): Likewise.
4161 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4163 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4166 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4168 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4170 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4171 to a temporary file and use move-if-change to update the real
4172 file where necessary.
4174 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4176 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4177 rather than Upa for CPY /M.
4179 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4181 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4184 2020-01-06 Martin Liska <mliska@suse.cz>
4186 PR tree-optimization/92860
4187 * params.opt: Mark param_max_combine_insns with Optimization
4190 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4193 * config/i386/i386.md (SWIDWI): New mode iterator.
4194 (DWI, dwi): Add TImode variants.
4195 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4196 <general_hilo_operand> instead of <general_operand>. Use
4197 CONST_SCALAR_INT_P instead of CONST_INT_P.
4198 (*addv<mode>4_1): Rename to ...
4199 (addv<mode>4_1): ... this.
4200 (QWI): New mode attribute.
4201 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4202 define_insn_and_split patterns.
4203 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4205 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4206 <general_hilo_operand> instead of <general_operand>.
4207 (*addcarry<mode>_1): New define_insn.
4208 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4210 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4212 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4213 Use "call" instead of "set".
4215 2020-01-03 Martin Jambor <mjambor@suse.cz>
4218 * ipa-cp.c (print_all_lattices): Skip functions without info.
4220 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4223 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4224 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4225 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4226 for 'e' simd clones.
4229 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4231 (mprefer-vector-width=): Add Save.
4232 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4233 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4234 (ix86_debug_options, ix86_function_specific_print): Adjust
4235 ix86_target_string callers.
4236 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4237 (ix86_valid_target_attribute_tree): Likewise.
4238 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4239 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4240 ix86_target_string caller.
4243 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4244 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4245 instead of gen_int_shift_amount + convert_modes.
4247 PR rtl-optimization/93088
4248 * loop-iv.c (find_single_def_src): Punt after looking through
4249 128 reg copies for regs with single definitions. Move definitions
4252 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4254 * config/arm/arm-c.c (arm_cpu_builtins): Define
4255 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4256 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4257 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4258 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4259 * config/arm/arm-tables.opt: Regenerated.
4260 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4261 arm_arch_i8mm and arm_arch_bf16 when enabled.
4262 * config/arm/arm.h (TARGET_I8MM): New macro.
4263 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4264 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4265 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4266 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4267 (v8_6_a_simd_variants): New.
4268 (v8_*_a_simd_variants): Add i8mm and bf16.
4269 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4271 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4274 * predict.c (compute_function_frequency): Don't call
4275 warn_function_cold on functions that already have cold attribute.
4277 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4280 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4281 COMDAT group function labels in .data.rel.ro.local section.
4282 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4285 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4286 comparison_operator in B and S integer comparisons. Likewise, use
4287 ordered_comparison_operator instead of cmpib_comparison_operator in
4289 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4291 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4293 Update copyright years.
4295 * gcc.c (process_command): Update copyright notice dates.
4296 * gcov-dump.c (print_version): Ditto.
4297 * gcov.c (print_version): Ditto.
4298 * gcov-tool.c (print_version): Ditto.
4299 * gengtype.c (create_file): Ditto.
4300 * doc/cpp.texi: Bump @copying's copyright year.
4301 * doc/cppinternals.texi: Ditto.
4302 * doc/gcc.texi: Ditto.
4303 * doc/gccint.texi: Ditto.
4304 * doc/gcov.texi: Ditto.
4305 * doc/install.texi: Ditto.
4306 * doc/invoke.texi: Ditto.
4308 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4310 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4313 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4315 PR tree-optimization/93098
4316 * match.pd (popcount): For shift amounts, use integer_onep
4317 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4318 tests. Make sure that precision is power of two larger than or equal
4319 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4320 instead of ULL suffixed constants. Formatting fixes.
4322 Copyright (C) 2020 Free Software Foundation, Inc.
4324 Copying and distribution of this file, with or without modification,
4325 are permitted in any medium without royalty provided the copyright
4326 notice and this notice are preserved.