1 2020-02-21 Martin Jambor <mjambor@suse.cz>
3 PR tree-optimization/93845
4 * tree-sra.c (verify_sra_access_forest): Only test access size of
7 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
9 * config/gcn/gcn.c (gcn_hard_regno_mode_ok): Align VGPR pairs.
10 * config/gcn/gcn-valu.md (addv64di3): Remove early-clobber.
11 (addv64di3_exec): Likewise.
12 (subv64di3): Likewise.
13 (subv64di3_exec): Likewise.
14 (addv64di3_zext): Likewise.
15 (addv64di3_zext_exec): Likewise.
16 (addv64di3_zext_dup): Likewise.
17 (addv64di3_zext_dup_exec): Likewise.
18 (addv64di3_zext_dup2): Likewise.
19 (addv64di3_zext_dup2_exec): Likewise.
20 (addv64di3_sext_dup2): Likewise.
21 (addv64di3_sext_dup2_exec): Likewise.
22 (<expander>v64di3): Likewise.
23 (<expander>v64di3_exec): Likewise.
24 (*<reduc_op>_dpp_shr_v64di): Likewise.
25 (*plus_carry_dpp_shr_v64di): Likewise.
26 * config/gcn/gcn.md (adddi3): Likewise.
27 (addptrdi3): Likewise.
28 (<expander>di3): Likewise.
30 2020-02-21 Andrew Stubbs <ams@codesourcery.com>
32 * config/gcn/gcn-valu.md (vec_seriesv64di): Use gen_vec_duplicatev64di.
34 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
36 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Add SVE
37 support. Use aarch64_emit_mult instead of emitting multiplication
38 instructions directly.
39 * config/aarch64/aarch64-sve.md (sqrt<mode>2, rsqrt<mode>2)
40 (@aarch64_rsqrte<mode>, @aarch64_rsqrts<mode>): New expanders.
42 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
44 * config/aarch64/aarch64.c (aarch64_emit_mult): New function.
45 (aarch64_emit_approx_div): Add SVE support. Use aarch64_emit_mult
46 instead of emitting multiplication instructions directly.
47 * config/aarch64/iterators.md (SVE_COND_FP_BINARY_OPTAB): New iterator.
48 * config/aarch64/aarch64-sve.md (div<mode>3, @aarch64_frecpe<mode>)
49 (@aarch64_frecps<mode>): New expanders.
51 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
53 * config/aarch64/aarch64-protos.h (AARCH64_APPROX_MODE): Operate
54 on and produce uint64_ts rather than ints.
55 (AARCH64_APPROX_NONE, AARCH64_APPROX_ALL): Change to uint64_ts.
56 (cpu_approx_modes): Change the fields from unsigned int to uint64_t.
58 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
60 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Don't create
61 an unused xmsk register when handling approximate rsqrt.
63 2020-02-21 Richard Sandiford <richard.sandiford@arm.com>
65 * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Fix inverted
66 flag_finite_math_only condition.
68 2020-02-20 Uroš Bizjak <ubizjak@gmail.com>
71 * config/i386/mmx.md (*vec_extractv2sf_1): Match source operand
72 to destination operand for shufps alternative.
73 (*vec_extractv2si_1): Ditto.
75 2020-02-20 Peter Bergner <bergner@linux.ibm.com>
78 * config/rs6000/rs6000.c (rs6000_legitimate_address_p): Handle VSX
81 2020-02-20 Martin Liska <mliska@suse.cz>
84 * config/darwin.c (darwin_override_options): Change 64b to 64-bit mode.
86 2020-02-20 Martin Liska <mliska@suse.cz>
89 * common/config/avr/avr-common.c: Remote trailing "|".
91 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
93 * collect2.c (maybe_run_lto_and_relink): Fix typo in
96 2020-02-19 Richard Sandiford <richard.sandiford@arm.com>
98 PR tree-optimization/93767
99 * tree-vect-data-refs.c (vect_compile_time_alias): Remove the
100 access-size bias from the offset calculations for negative strides.
102 2020-02-19 Bernd Edlinger <bernd.edlinger@hotmail.de>
104 * collect2.c (c_file, o_file): Make const again.
105 (ldout,lderrout, dump_ld_file): Remove.
106 (tool_cleanup): Avoid calling not signal-safe functions.
107 (maybe_run_lto_and_relink): Avoid possible signal handler
108 access to unintialzed memory (lto_o_files).
109 (main): Avoid leaking temp files in $TMPDIR.
110 Initialize c_file/o_file with concat, which avoids exposing
111 uninitialized memory to signal handler, which calls unlink(!).
112 Avoid calling maybe_unlink when the main function returns,
113 since the atexit handler is already doing this.
114 * collect2.h (dump_ld_file, ldout, lderrout): Remove.
116 2020-02-19 Martin Jambor <mjambor@suse.cz>
118 PR tree-optimization/93776
119 * tree-sra.c (create_access): Do not create zero size accesses.
120 (get_access_for_expr): Do not search for zero sized accesses.
122 2020-02-19 Martin Jambor <mjambor@suse.cz>
124 PR tree-optimization/93667
125 * tree-sra.c (scalarizable_type_p): Return false if record fields
126 do not follow wach other.
128 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
130 * config/riscv/riscv.c (riscv_output_move) Using fmv.x.w/fmv.w.x
131 rather than fmv.x.s/fmv.s.x.
133 2020-02-18 James Greenhalgh <james.greenhalgh@arm.com>
135 * config/aarch64/aarch64-simd-builtins.def
136 (intrinsic_vec_smult_lo_): New.
137 (intrinsic_vec_umult_lo_): Likewise.
138 (vec_widen_smult_hi_): Likewise.
139 (vec_widen_umult_hi_): Likewise.
140 * config/aarch64/aarch64-simd.md
141 (aarch64_intrinsic_vec_<su>mult_lo_<mode>): New.
142 * config/aarch64/arm_neon.h (vmull_high_s8): Use intrinsics.
143 (vmull_high_s16): Likewise.
144 (vmull_high_s32): Likewise.
145 (vmull_high_u8): Likewise.
146 (vmull_high_u16): Likewise.
147 (vmull_high_u32): Likewise.
148 (vmull_s8): Likewise.
149 (vmull_s16): Likewise.
150 (vmull_s32): Likewise.
151 (vmull_u8): Likewise.
152 (vmull_u16): Likewise.
153 (vmull_u32): Likewise.
155 2020-02-18 Martin Liska <mliska@suse.cz>
157 * value-prof.c (stream_out_histogram_value): Restore LTO PGO
158 bootstrap by missing removal of invalid sanity check.
160 2020-02-18 Martin Liska <mliska@suse.cz>
163 * ipa-icf-gimple.c (func_checker::compare_gimple_assign):
164 Always compare LHS of gimple_assign.
166 2020-02-18 Martin Liska <mliska@suse.cz>
169 * cgraph.c (cgraph_node::verify_node): Verify MALLOC attribute
170 and return type of functions.
171 * ipa-param-manipulation.c (ipa_param_adjustments::adjust_decl):
172 Drop MALLOC attribute for void functions.
173 * ipa-pure-const.c (funct_state_summary_t::duplicate): Drop
174 malloc_state for a new VOID clone.
176 2020-02-18 Martin Liska <mliska@suse.cz>
179 * common.opt: Add -fprofile-reproducibility.
180 * doc/invoke.texi: Document it.
181 * value-prof.c (dump_histogram_value):
182 Document and support behavior for counters[0]
183 being a negative value.
184 (get_nth_most_common_value): Handle negative
185 counters[0] in respect to flag_profile_reproducible.
187 2020-02-18 Jakub Jelinek <jakub@redhat.com>
190 * cgraph.c (verify_speculative_call): Use speculative_id instead of
191 speculative_uid in messages. Remove trailing whitespace from error
192 message. Use num_speculative_call_targets instead of
193 num_speculative_targets in a message.
194 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
195 edge messages and stmt instead of cal_stmt in reference message.
197 PR tree-optimization/93780
198 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
199 before calling build_vector_type.
200 (execute_update_addresses_taken): Likewise.
203 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
204 typo, functoin -> function.
205 * tree.c (free_lang_data_in_decl): Fix comment typo,
206 functoin -> function.
207 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
209 2020-02-17 David Malcolm <dmalcolm@redhat.com>
211 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
213 (print_option_information): Don't call get_option_url if URLs
216 2020-02-17 Alexandre Oliva <oliva@adacore.com>
218 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
219 handling of register_common-less targets.
221 2020-02-17 Martin Liska <mliska@suse.cz>
224 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
226 2020-02-17 Martin Liska <mliska@suse.cz>
229 * config/rs6000/rs6000.c (rs6000_option_override_internal):
232 2020-02-17 Martin Liska <mliska@suse.cz>
235 * config/rx/elf.opt: Fix typo.
237 2020-02-17 Richard Biener <rguenther@suse.de>
240 * opts-global.c (print_ignored_options): Use inform and
243 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
246 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
248 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
251 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
252 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
254 2020-02-15 Jason Merrill <jason@redhat.com>
256 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
258 2020-02-15 Jakub Jelinek <jakub@redhat.com>
260 PR tree-optimization/93744
261 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
262 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
263 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
264 sure @2 in the first and @1 in the other patterns has no side-effects.
266 2020-02-15 David Malcolm <dmalcolm@redhat.com>
267 Bernd Edlinger <bernd.edlinger@hotmail.de>
271 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
272 * configure.ac (--with-diagnostics-urls): New configuration
273 option, based on --with-diagnostics-color.
274 (DIAGNOSTICS_URLS_DEFAULT): New define.
275 * config.h: Regenerate.
276 * configure: Regenerate.
277 * diagnostic.c (diagnostic_urls_init): Handle -1 for
278 DIAGNOSTICS_URLS_DEFAULT from configure-time
279 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
280 and TERM_URLS environment variable.
281 * diagnostic-url.h (diagnostic_url_format): New enum type.
282 (diagnostic_urls_enabled_p): rename to...
283 (determine_url_format): ... this, and change return type.
284 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
285 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
286 the linux console, and mingw.
287 (diagnostic_urls_enabled_p): rename to...
288 (determine_url_format): ... this, and adjust.
289 * pretty-print.h (pretty_printer::show_urls): rename to...
290 (pretty_printer::url_format): ... this, and change to enum.
291 * pretty-print.c (pretty_printer::pretty_printer,
292 pp_begin_url, pp_end_url, test_urls): Adjust.
293 * doc/install.texi (--with-diagnostics-urls): Document the new
294 configuration option.
295 (--with-diagnostics-color): Document the existing interaction
296 with GCC_COLORS better.
297 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
298 vindex reference. Update description of defaults based on the above.
299 (-fdiagnostics-color): Update description of how -fdiagnostics-color
300 interacts with GCC_COLORS.
302 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
305 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
306 conjunction with TARGET_GNU_TLS in early return.
308 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
310 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
311 the mode is not wider than UNITS_PER_WORD.
313 2020-02-14 Martin Jambor <mjambor@suse.cz>
315 PR tree-optimization/93516
316 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
317 access of the same type as the parent.
318 (propagate_subaccesses_from_lhs): Likewise.
320 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
323 * config/i386/avx512vbmi2intrin.h
324 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
325 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
326 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
327 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
328 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
329 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
330 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
331 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
332 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
333 of lacking a closing parenthesis.
334 * config/i386/avx512vbmi2vlintrin.h
335 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
336 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
337 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
338 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
339 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
340 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
341 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
342 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
343 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
344 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
345 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
346 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
347 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
348 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
349 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
350 _mm_shldi_epi32, _mm_mask_shldi_epi32,
351 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
352 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
354 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
357 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
358 the target function entry.
360 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
362 * common/config/arc/arc-common.c (arc_option_optimization_table):
363 Disable if-conversion step when optimized for size.
365 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
367 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
368 R12-R15 are always in ARCOMPACT16_REGS register class.
369 * config/arc/arc.opt (mq-class): Deprecate.
370 * config/arc/constraint.md ("q"): Remove dependency on mq-class
372 * doc/invoke.texi (mq-class): Update text.
373 * common/config/arc/arc-common.c (arc_option_optimization_table):
376 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
378 * config/arc/arc.c (arc_insn_cost): New function.
379 (TARGET_INSN_COST): Define.
380 * config/arc/arc.md (cost): New attribute.
381 (add_n): Use arc_nonmemory_operand.
382 (ashlsi3_insn): Likewise, also update constraints.
383 (ashrsi3_insn): Likewise.
385 (add_shift): Likewise.
386 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
388 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
390 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
392 (umulsidi_600): Likewise.
394 2020-02-13 Jakub Jelinek <jakub@redhat.com>
397 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
398 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
399 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
400 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
401 pass __A to the builtin followed by __W instead of __A followed by
403 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
404 _mm512_mask_popcnt_epi64): Likewise.
405 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
406 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
407 _mm256_mask_popcnt_epi64): Likewise.
409 PR tree-optimization/93582
410 * fold-const.h (shift_bytes_in_array_left,
411 shift_bytes_in_array_right): Declare.
412 * fold-const.c (shift_bytes_in_array_left,
413 shift_bytes_in_array_right): New function, moved from
414 gimple-ssa-store-merging.c, no longer static.
415 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
416 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
417 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
418 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
419 shift_bytes_in_array.
420 (verify_shift_bytes_in_array): Rename to ...
421 (verify_shift_bytes_in_array_left): ... this. Use
422 shift_bytes_in_array_left instead of shift_bytes_in_array.
423 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
424 instead of verify_shift_bytes_in_array.
425 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
426 / native_interpret_expr where the store covers all needed bits,
427 punt on PDP-endian, otherwise allow all involved offsets and sizes
428 not to be byte-aligned.
431 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
432 use const_0_to_255_operand predicate instead of immediate_operand.
433 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
434 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
435 vgf2p8affineinvqb_<mode><mask_name>,
436 vgf2p8affineqb_<mode><mask_name>): Drop mode from
437 const_0_to_255_operand predicated operands.
439 2020-02-12 Jeff Law <law@redhat.com>
441 * config/h8300/h8300.md (comparison shortening peepholes): Use
442 a mode iterator to merge the HImode and SImode peepholes.
444 2020-02-12 Jakub Jelinek <jakub@redhat.com>
447 * real.c (is_even): Make static. Function comment fix.
448 (is_halfway_below): Make static, don't assert R is not inf/nan,
449 instead return false for those. Small formatting fixes.
451 2020-02-12 Martin Sebor <msebor@redhat.com>
454 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
455 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
456 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
457 (strlen_check_and_optimize_call): Adjust callee name.
459 2020-02-12 Jeff Law <law@redhat.com>
461 * config/h8300/h8300.md (comparison shortening peepholes): Drop
462 (and (xor)) variant. Combine other two into single peephole.
464 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
466 PR rtl-optimization/93565
467 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
469 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
471 * config/aarch64/aarch64-simd.md
472 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
473 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
474 generating separate ADDV and zero_extend patterns.
475 * config/aarch64/iterators.md (VDQV_E): New iterator.
477 2020-02-12 Jeff Law <law@redhat.com>
479 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
480 expanders, splits, etc.
481 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
482 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
483 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
484 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
485 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
486 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
488 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
490 2020-02-12 Jakub Jelinek <jakub@redhat.com>
493 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
494 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
495 TARGET_AVX512DQ from condition.
496 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
497 instead of <mask_mode512bit_condition> in condition. If
498 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
500 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
503 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
506 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
508 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
510 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
511 where strlen is more legible.
512 (rs6000_builtin_vectorized_libmass): Ditto.
513 (rs6000_print_options_internal): Ditto.
515 2020-02-11 Martin Sebor <msebor@redhat.com>
517 PR tree-optimization/93683
518 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
520 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
522 * config/rs6000/predicates.md (cint34_operand): Rename the
523 -mprefixed-addr option to be -mprefixed.
524 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
525 the -mprefixed-addr option to be -mprefixed.
526 (OTHER_FUTURE_MASKS): Likewise.
527 (POWERPC_MASKS): Likewise.
528 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
529 the -mprefixed-addr option to be -mprefixed. Change error
530 messages to refer to -mprefixed.
531 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
533 (rs6000_legitimate_offset_address_p): Likewise.
534 (rs6000_mode_dependent_address): Likewise.
535 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
536 "-mprefixed" for target attributes and pragmas.
537 (address_to_insn_form): Rename the -mprefixed-addr option to be
539 (rs6000_adjust_insn_length): Likewise.
540 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
541 -mprefixed-addr option to be -mprefixed.
542 (ASM_OUTPUT_OPCODE): Likewise.
543 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
544 -mprefixed-addr option to be -mprefixed.
545 * config/rs6000/rs6000.opt (-mprefixed): Rename the
546 -mprefixed-addr option to be prefixed. Change the option from
547 being undocumented to being documented.
548 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
549 -mprefixed option. Update the -mpcrel documentation to mention
552 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
554 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
555 including FIRST_PSEUDO_REGISTER - 1.
556 * ira-color.c (print_hard_reg_set): Ditto.
558 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
560 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
561 (USTERNOP_QUALIFIERS): New define.
562 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
563 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
564 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
565 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
566 * config/arm/arm_neon.h (vusdot_s32): New.
567 (vusdot_lane_s32): New.
568 (vusdotq_lane_s32): New.
569 (vsudot_lane_s32): New.
570 (vsudotq_lane_s32): New.
571 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
572 * config/arm/iterators.md (DOTPROD_I8MM): New.
573 (sup, opsuffix): Add <us/su>.
574 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
575 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
577 2020-02-11 Richard Biener <rguenther@suse.de>
579 PR tree-optimization/93661
580 PR tree-optimization/93662
581 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
583 * tree-sra.c (get_access_for_expr): Likewise.
585 2020-02-10 Jakub Jelinek <jakub@redhat.com>
588 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
589 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
590 Change condition from TARGET_AVX2 to TARGET_AVX.
592 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
595 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
598 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
600 Try to generate zero-based comparisons.
601 * config/cris/cris.c (cris_reduce_compare): New function.
602 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
603 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
604 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
606 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
609 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
610 in Thumb state and also as a destination in Arm state. Add T16
613 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
615 * md.texi (Define Subst): Match closing paren in example.
617 2020-02-10 Jakub Jelinek <jakub@redhat.com>
621 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
622 arguments of strncmp.
624 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
627 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
628 but different source value.
629 (adjust_callers_for_value_intersection): New function.
630 (gather_edges_for_value): Adjust order of callers to let a
631 non-self-recursive caller be the first element.
632 (self_recursive_pass_through_p): Add a new parameter "simple", and
633 check generalized self-recursive pass-through jump function.
634 (self_recursive_agg_pass_through_p): Likewise.
635 (find_more_scalar_values_for_callers_subset): Compute value from
636 pass-through jump function for self-recursive.
637 (intersect_with_plats): Cleanup previous implementation code for value
638 itersection with self-recursive call edge.
639 (intersect_with_agg_replacements): Likewise.
640 (intersect_aggregates_with_edge): Deduce value from pass-through jump
641 function for self-recursive call edge. Cleanup previous implementation
642 code for value intersection with self-recursive call edge.
643 (decide_whether_version_node): Remove dead callers and adjust order
644 to let a non-self-recursive caller be the first element.
646 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
648 * recog.c: Move pass_split_before_sched2 code in front of
649 pass_split_before_regstack.
650 (pass_data_split_before_sched2): Rename pass to split3 from split4.
651 (pass_data_split_before_regstack): Rename pass to split4 from split3.
652 (rest_of_handle_split_before_sched2): Remove.
653 (pass_split_before_sched2::execute): Unconditionally call
655 (enable_split_before_sched2): New function.
656 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
657 (pass_split_before_regstack::gate): Ditto.
658 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
659 Update name check for renamed split4 pass.
660 * config/sh/sh.c (register_sh_passes): Update pass insertion
661 point for renamed split4 pass.
663 2020-02-09 Jakub Jelinek <jakub@redhat.com>
665 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
666 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
667 copying them around between host and target.
669 2020-02-08 Andrew Pinski <apinski@marvell.com>
672 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
673 STRICT_ALIGNMENT also.
675 2020-02-08 Jim Wilson <jimw@sifive.com>
678 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
680 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
681 Jakub Jelinek <jakub@redhat.com>
684 * config/i386/i386.h (CALL_USED_REGISTERS): Make
685 xmm16-xmm31 call-used even in 64-bit ms-abi.
687 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
689 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
690 (simd_ummla, simd_usmmla): Likewise.
691 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
692 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
695 2020-02-07 Richard Biener <rguenther@suse.de>
698 * tree-inline.c (fold_marked_statements): Do a PRE walk,
699 skipping unreachable regions.
700 (optimize_inline_calls): Skip folding stmts when we didn't
703 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
706 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
707 Don't return aggregates with only SFmode and DFmode in SSE
709 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
711 2020-02-07 Jakub Jelinek <jakub@redhat.com>
714 * config/rs6000/rs6000-logue.c
715 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
716 if it fails, move rs into end_addr and retry. Add
717 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
718 the insn pattern doesn't describe well what exactly happens to
722 * config/i386/predicates.md (avx_identity_operand): Remove.
723 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
724 (avx_<castmode><avxsizesuffix>_<castmode>,
725 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
726 a VEC_CONCAT of the operand and UNSPEC_CAST.
727 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
728 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
732 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
733 recog_data.insn if distance_non_agu_define changed it.
735 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
738 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
739 we only had X-FORM (reg+reg) addressing for vectors. Also before
740 ISA 3.0, we only had X-FORM addressing for scalars in the
741 traditional Altivec registers.
743 2020-02-06 <zhongyunde@huawei.com>
744 Vladimir Makarov <vmakarov@redhat.com>
746 PR rtl-optimization/93561
747 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
750 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
752 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
755 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
757 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
758 where the low and the high 32 bits are equal to each other specially,
759 with an rldimi instruction.
761 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
763 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
765 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
767 * config/arm/arm-tables.opt: Regenerate.
769 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
772 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
773 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
774 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
776 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
778 PR rtl-optimization/87763
779 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
781 2020-02-06 Delia Burduv <delia.burduv@arm.com>
783 * config/aarch64/aarch64-simd-builtins.def
784 (bfmlaq): New built-in function.
785 (bfmlalb): New built-in function.
786 (bfmlalt): New built-in function.
787 (bfmlalb_lane): New built-in function.
788 (bfmlalt_lane): New built-in function.
789 * config/aarch64/aarch64-simd.md
790 (aarch64_bfmmlaqv4sf): New pattern.
791 (aarch64_bfmlal<bt>v4sf): New pattern.
792 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
793 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
794 (vbfmlalbq_f32): New intrinsic.
795 (vbfmlaltq_f32): New intrinsic.
796 (vbfmlalbq_lane_f32): New intrinsic.
797 (vbfmlaltq_lane_f32): New intrinsic.
798 (vbfmlalbq_laneq_f32): New intrinsic.
799 (vbfmlaltq_laneq_f32): New intrinsic.
800 * config/aarch64/iterators.md (BF_MLA): New int iterator.
801 (bt): New int attribute.
803 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
805 * config/i386/i386.md (*pushtf): Emit "#" instead of
806 calling gcc_unreachable in insn output.
809 (*pushsf_rex64): Ditto for alternatives other than 1.
810 (*pushsf): Ditto for alternatives other than 1.
812 2020-02-06 Martin Liska <mliska@suse.cz>
814 PR gcov-profile/91971
815 PR gcov-profile/93466
816 * coverage.c (coverage_init): Revert mangling of
817 path into filename. It can lead to huge filename length.
818 Creation of subfolders seem more natural.
820 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
823 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
824 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
825 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
827 2020-02-06 Jakub Jelinek <jakub@redhat.com>
830 * config/i386/predicates.md (avx_identity_operand): New predicate.
831 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
832 define_insn_and_split.
835 * omp-low.c (use_pointer_for_field): For nested constructs, also
836 look for map clauses on target construct.
837 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
838 taskreg_nesting_level.
841 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
842 shared clause, call omp_notice_variable on outer context if any.
844 2020-02-05 Jason Merrill <jason@redhat.com>
847 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
848 non-zero address even if weak and not yet defined.
850 2020-02-05 Martin Sebor <msebor@redhat.com>
852 PR tree-optimization/92765
853 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
854 * tree-ssa-strlen.c (compute_string_length): Remove.
855 (determine_min_objsize): Remove.
856 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
857 Avoid using type size as the upper bound on string length.
858 (handle_builtin_string_cmp): Add an argument. Adjust.
859 (strlen_check_and_optimize_call): Pass additional argument to
860 handle_builtin_string_cmp.
862 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
864 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
865 (*pushdi2_rex64 peephole2): Unconditionally split after
867 (*ashl<mode>3_doubleword): Ditto.
868 (*<shift_insn><mode>3_doubleword): Ditto.
870 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
873 * config/rs6000/rs6000.c (get_vector_offset): Fix
875 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
877 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
879 2020-02-05 David Malcolm <dmalcolm@redhat.com>
882 (Special Functions for Debugging the Analyzer): Update description
883 of __analyzer_dump_exploded_nodes.
885 2020-02-05 Jakub Jelinek <jakub@redhat.com>
888 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
889 include sets and not clobbers in the vzeroupper pattern.
890 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
891 the parallel has 17 (64-bit) or 9 (32-bit) elts.
892 (*avx_vzeroupper_1): New define_insn_and_split.
895 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
896 don't run when !optimize.
897 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
900 2020-02-05 Richard Biener <rguenther@suse.de>
903 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
904 checks before matching calls.
906 2020-02-05 Jakub Jelinek <jakub@redhat.com>
908 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
909 function comment typo.
912 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
913 simd_clone_create failed when i == 0, adjust clone->nargs by
916 2020-02-05 Martin Liska <mliska@suse.cz>
919 * doc/invoke.texi: Document that one should
920 not combine ASLR and -fpch.
922 2020-02-04 Richard Biener <rguenther@suse.de>
924 PR tree-optimization/93538
925 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
927 2020-02-04 Richard Biener <rguenther@suse.de>
929 PR tree-optimization/91123
930 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
931 (vn_walk_cb_data::last_vuse): New member.
932 (vn_walk_cb_data::saved_operands): Likewsie.
933 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
934 (vn_walk_cb_data::push_partial_def): Use finish.
935 (vn_reference_lookup_2): Update last_vuse and use finish if
936 we've saved operands.
937 (vn_reference_lookup_3): Use finish and update calls to
938 push_partial_defs everywhere. When translating through
939 memcpy or aggregate copies save off operands and alias-set.
940 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
941 operation for redundant store removal.
943 2020-02-04 Richard Biener <rguenther@suse.de>
945 PR tree-optimization/92819
946 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
947 generating more stmts than before.
949 2020-02-04 Martin Liska <mliska@suse.cz>
951 * config/arm/arm.c (arm_gen_far_branch): Move the function
952 outside of selftests.
954 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
956 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
957 function to adjust PC-relative vector addresses.
958 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
959 handle vectors with PC-relative addresses.
961 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
963 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
965 (hard_reg_and_mode_to_addr_mask): Delete.
966 (rs6000_adjust_vec_address): If the original vector address
967 was REG+REG or REG+OFFSET and the element is not zero, do the add
968 of the elements in the original address before adding the offset
969 for the vector element. Use address_to_insn_form to validate the
970 address using the register being loaded, rather than guessing
971 whether the address is a DS-FORM or DQ-FORM address.
973 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
975 * config/rs6000/rs6000.c (get_vector_offset): New helper function
976 to calculate the offset in memory from the start of a vector of a
977 particular element. Add code to keep the element number in
978 bounds if the element number is variable.
979 (rs6000_adjust_vec_address): Move calculation of offset of the
980 vector element to get_vector_offset.
981 (rs6000_split_vec_extract_var): Do not do the initial AND of
982 element here, move the code to get_vector_offset.
984 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
986 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
989 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
991 * config/rs6000/constraints.md: Improve documentation.
993 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
996 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
997 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
999 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
1001 * config.gcc: Remove "carrizo" support.
1002 * config/gcn/gcn-opts.h (processor_type): Likewise.
1003 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
1004 * config/gcn/gcn.opt (gpu_type): Likewise.
1005 * config/gcn/t-omp-device: Likewise.
1007 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
1010 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
1011 * config/arm/arm.c (arm_gen_far_branch): New function
1013 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
1015 2020-02-03 Julian Brown <julian@codesourcery.com>
1016 Tobias Burnus <tobias@codesourcery.com>
1018 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
1020 2020-02-03 Jakub Jelinek <jakub@redhat.com>
1023 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
1024 valid RTL to sum up the lowest and second lowest bytes of the popcnt
1027 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
1029 PR rtl-optimization/91333
1030 * ira-color.c (struct allocno_color_data): Add member
1032 (init_allocno_threads): Set the member up.
1033 (bucket_allocno_compare_func): Add compare hard reg
1036 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
1038 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
1040 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1041 * config.in: Regenerated.
1042 * configure: Regenerated.
1043 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
1044 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
1045 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
1047 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
1049 * configure: Regenerate.
1051 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
1053 PR rtl-optimization/91333
1054 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
1055 reg preferences comparison up.
1057 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1059 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
1060 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
1061 aarch64-sve-builtins-base.h.
1062 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
1063 aarch64-sve-builtins-base.cc.
1064 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
1065 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1067 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
1068 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1069 (svcvtnt): New functions.
1070 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
1071 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
1072 (svcvtnt): New functions.
1073 (svcvt): Add a form that converts f32 to bf16.
1074 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
1075 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
1077 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
1078 Treat B as bfloat16_t.
1079 (ternary_bfloat_lane_base): New class.
1080 (ternary_bfloat_def): Likewise.
1081 (ternary_bfloat): New shape.
1082 (ternary_bfloat_lane_def): New class.
1083 (ternary_bfloat_lane): New shape.
1084 (ternary_bfloat_lanex2_def): New class.
1085 (ternary_bfloat_lanex2): New shape.
1086 (ternary_bfloat_opt_n_def): New class.
1087 (ternary_bfloat_opt_n): New shape.
1088 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
1089 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
1090 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
1091 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
1092 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1093 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
1094 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
1095 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
1096 the pattern off the narrow mode instead of the wider one.
1097 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
1098 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
1099 (sve_fp_op): Handle them.
1100 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
1101 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
1103 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
1105 * config/aarch64/arm_sve.h: Include arm_bf16.h.
1106 * config/aarch64/aarch64-modes.def (BF): Move definition before
1107 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
1108 (SVE_MODES): Handle BF modes.
1109 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
1111 (aarch64_full_sve_mode): Likewise.
1112 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
1114 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
1115 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
1116 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
1117 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
1119 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
1121 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
1122 (TYPES_all_data): Add bf16.
1123 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
1124 (register_tuple_type): Increase buffer size.
1125 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
1126 (bf16): New type suffix.
1127 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
1128 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
1129 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
1130 Change type from all_data to all_arith.
1131 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
1134 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
1135 Matthew Malcomson <matthew.malcomson@arm.com>
1136 Richard Sandiford <richard.sandiford@arm.com>
1138 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
1139 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
1140 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
1141 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
1142 __ARM_FEATURE_MATMUL_FP64.
1143 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
1144 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
1145 be disabled at the same time.
1146 (f32mm): New extension.
1147 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
1148 (AARCH64_FL_F64MM): Bump to the next bit up.
1149 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
1150 (TARGET_SVE_F64MM): New macros.
1151 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
1152 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
1153 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
1154 (UNSPEC_ZIP2Q): New unspeccs.
1155 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
1156 (optab, sur, perm_insn): Handle the new unspecs.
1157 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
1158 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
1159 TARGET_SVE_F64MM instead of separate tests.
1160 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
1161 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
1162 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
1163 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
1164 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
1165 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
1166 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
1167 (TYPES_s_signed): New macro.
1168 (TYPES_s_integer): Use it.
1169 (TYPES_d_float): New macro.
1170 (TYPES_d_data): Use it.
1171 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
1172 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
1173 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
1174 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
1175 (svmmla): New shape.
1176 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
1177 template parameters.
1178 (ternary_resize2_lane_base): Likewise.
1179 (ternary_resize2_base): New class.
1180 (ternary_qq_lane_base): Likewise.
1181 (ternary_intq_uintq_lane_def): Likewise.
1182 (ternary_intq_uintq_lane): New shape.
1183 (ternary_intq_uintq_opt_n_def): New class
1184 (ternary_intq_uintq_opt_n): New shape.
1185 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1186 (ternary_uintq_intq_def): New class.
1187 (ternary_uintq_intq): New shape.
1188 (ternary_uintq_intq_lane_def): New class.
1189 (ternary_uintq_intq_lane): New shape.
1190 (ternary_uintq_intq_opt_n_def): New class.
1191 (ternary_uintq_intq_opt_n): New shape.
1192 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1193 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1194 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1195 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1197 (svdotprod_lane_impl): ...this new class.
1198 (svmmla_impl, svusdot_impl): New classes.
1199 (svdot_lane): Update to use svdotprod_lane_impl.
1200 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1201 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1203 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1204 function, with no types defined.
1205 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1206 AARCH64_FL_I8MM functions.
1207 (svmmla): New AARCH64_FL_F32MM function.
1208 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1209 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1210 AARCH64_FL_F64MM function.
1211 (REQUIRED_EXTENSIONS):
1213 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1215 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1218 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1220 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1221 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1222 (*movti_internal): Do not check for
1223 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1224 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1225 just after check for TARGET_AVX.
1226 (*movdf_internal): Ditto.
1227 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1228 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1229 * config/i386/sse.md (mov<mode>_internal): Only check
1230 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1231 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1232 (<sse>_andnot<mode>3<mask_name>): Move check for
1233 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1234 (<code><mode>3<mask_name>): Ditto.
1235 (*andnot<mode>3): Ditto.
1236 (*andnottf3): Ditto.
1237 (*<code><mode>3): Ditto.
1238 (*<code>tf3): Ditto.
1239 (*andnot<VI:mode>3): Remove
1240 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1241 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1242 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1243 (sse4_1_blendv<ssemodesuffix>): Ditto.
1244 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1245 Explain that tune applies to 128bit instructions only.
1247 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1249 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1250 to definition of hsa_kernel_description. Parse assembly to find SGPR
1251 and VGPR count of kernel and store in hsa_kernel_description.
1253 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1255 PR rtl-optimization/91838
1256 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1257 to truncate if allowed or reject combination.
1259 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1261 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1262 (find_inv_vars_cb): Likewise.
1264 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1266 * calls.c (special_function_p): Split out the check for DECL_NAME
1267 being non-NULL and fndecl being extern at file scope into a
1268 new maybe_special_function_p and call it. Drop check for fndecl
1269 being non-NULL that was after a usage of DECL_NAME (fndecl).
1270 * tree.h (maybe_special_function_p): New inline function.
1272 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1274 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1275 (mask_gather_load<mode>): ... here, and zero-initialize the
1277 (maskload<mode>di): Zero-initialize the destination.
1280 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1283 * doc/analyzer.texi (Limitations): Note that constraints on
1284 floating-point values are currently ignored.
1286 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1289 * symtab.c (symtab_node::noninterposable_alias): If localalias
1290 already exists, but is not usable, append numbers after it until
1291 a unique name is found. Formatting fix.
1294 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1297 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1299 * config/gcn/gcn.c (print_operand): Handle LTGT.
1300 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1302 2020-01-30 Richard Biener <rguenther@suse.de>
1304 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1305 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1307 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1309 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1310 without a DECL in .data.rel.ro.local.
1312 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1315 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1319 * config/i386/sse.md
1320 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1321 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1322 any_extend code iterator instead of always zero_extend.
1323 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1324 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1325 Use any_extend code iterator instead of always zero_extend.
1326 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1327 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1328 Use any_extend code iterator instead of always zero_extend.
1329 (*sse2_pmovmskb_ext): New define_insn.
1330 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1333 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1334 (*popcountsi2_zext_falsedep): New define_insn.
1336 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1338 * config.in: Regenerated.
1339 * configure: Regenerated.
1341 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1344 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1345 LLVM's assembler changed the default in version 9.
1347 2020-01-24 Jeff Law <law@redhat.com>
1349 PR tree-optimization/89689
1350 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1352 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1356 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1358 PR rtl-optimization/87763
1359 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1360 simplification to handle subregs as well as bare regs.
1361 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1363 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1366 * ira.c (ira): Revert use of simplified LRA algorithm.
1368 2020-01-29 Martin Jambor <mjambor@suse.cz>
1370 PR tree-optimization/92706
1371 * tree-sra.c (struct access): Fields first_link, last_link,
1372 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1373 next_rhs_queued and grp_rhs_queued respectively, new fields
1374 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1375 (struct assign_link): Field next renamed to next_rhs, new field
1376 next_lhs. Updated comment.
1377 (work_queue_head): Renamed to rhs_work_queue_head.
1378 (lhs_work_queue_head): New variable.
1379 (add_link_to_lhs): New function.
1380 (relink_to_new_repr): Also relink LHS lists.
1381 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1382 (add_access_to_lhs_work_queue): New function.
1383 (pop_access_from_work_queue): Renamed to
1384 pop_access_from_rhs_work_queue.
1385 (pop_access_from_lhs_work_queue): New function.
1386 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1388 (child_would_conflict_in_lacc): Renamed to
1389 child_would_conflict_in_acc. Adjusted parameter names.
1390 (create_artificial_child_access): New parameter set_grp_read, use it.
1391 (subtree_mark_written_and_enqueue): Renamed to
1392 subtree_mark_written_and_rhs_enqueue.
1393 (propagate_subaccesses_across_link): Renamed to
1394 propagate_subaccesses_from_rhs.
1395 (propagate_subaccesses_from_lhs): New function.
1396 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1399 2020-01-29 Martin Jambor <mjambor@suse.cz>
1401 PR tree-optimization/92706
1402 * tree-sra.c (struct access): Adjust comment of
1403 grp_total_scalarization.
1404 (find_access_in_subtree): Look for single children spanning an entire
1406 (scalarizable_type_p): Allow register accesses, adjust callers.
1407 (completely_scalarize): Remove function.
1408 (scalarize_elem): Likewise.
1409 (create_total_scalarization_access): Likewise.
1410 (sort_and_splice_var_accesses): Do not track total scalarization
1412 (analyze_access_subtree): New parameter totally, adjust to new meaning
1413 of grp_total_scalarization.
1414 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1415 (can_totally_scalarize_forest_p): New function.
1416 (create_total_scalarization_access): Likewise.
1417 (create_total_access_and_reshape): Likewise.
1418 (total_should_skip_creating_access): Likewise.
1419 (totally_scalarize_subtree): Likewise.
1420 (analyze_all_variable_accesses): Perform total scalarization after
1421 subaccess propagation using the new functions above.
1422 (initialize_constant_pool_replacements): Output initializers by
1423 traversing the access tree.
1425 2020-01-29 Martin Jambor <mjambor@suse.cz>
1427 * tree-sra.c (verify_sra_access_forest): New function.
1428 (verify_all_sra_access_forests): Likewise.
1429 (create_artificial_child_access): Set parent.
1430 (analyze_all_variable_accesses): Call the verifier.
1432 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1434 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1435 if called on indirect edge.
1436 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1437 speculative call if needed.
1439 2020-01-29 Richard Biener <rguenther@suse.de>
1441 PR tree-optimization/93428
1442 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1443 permutation when the load node is created.
1444 (vect_analyze_slp_instance): Re-use it here.
1446 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1448 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1450 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1452 PR rtl-optimization/93272
1453 * ira-lives.c (process_out_of_region_eh_regs): New function.
1454 (process_bb_node_lives): Call it.
1456 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1458 * coverage.c (read_counts_file): Make error message lowercase.
1460 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1462 * profile-count.c (profile_quality_display_names): Fix ordering.
1464 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1467 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1468 hash only when edge is first within the sequence.
1469 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1470 (symbol_table::create_edge): Do not set target_prob.
1471 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1473 (cgraph_edge::make_speculative): Drop target_prob parameter.
1474 (cgraph_edge::speculative_call_info): Remove.
1475 (cgraph_edge::first_speculative_call_target): New member function.
1476 (update_call_stmt_hash_for_removing_direct_edge): New function.
1477 (cgraph_edge::resolve_speculation): Rewrite to new API.
1478 (cgraph_edge::speculative_call_for_target): New member function.
1479 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1480 multiple speculation targets.
1481 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1483 (verify_speculative_call): Verify that targets form an interval.
1484 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1485 (cgraph_edge::first_speculative_call_target): New member function.
1486 (cgraph_edge::next_speculative_call_target): New member function.
1487 (cgraph_edge::speculative_call_target_ref): New member function.
1488 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1489 (cgraph_edge): Remove target_prob.
1490 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1491 Fix handling of speculative calls.
1492 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1493 * ipa-fnsummary.c (analyze_function_body): Likewise.
1494 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1495 * ipa-profile.c (dump_histogram): Fix formating.
1496 (ipa_profile_generate_summary): Watch for overflows.
1497 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1498 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1499 (update_indirect_edges_after_inlining): Update to new API.
1500 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1502 * profile-count.h: (profile_probability::adjusted): New.
1503 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1504 updating of profile.
1505 * value-prof.c (gimple_ic_transform): Rename to ...
1506 (dump_ic_profile): ... this one; update dumping.
1507 (stream_in_histogram_value): Fix formating.
1508 (gimple_value_profile_transformations): Update.
1510 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1513 * config/i386/i386.md (*movoi_internal_avx): Remove
1514 TARGET_SSE_TYPELESS_STORES check.
1515 (*movti_internal): Prefer TARGET_AVX over
1516 TARGET_SSE_TYPELESS_STORES.
1517 (*movtf_internal): Likewise.
1518 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1519 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1520 from TARGET_SSE_TYPELESS_STORES.
1522 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1524 * diagnostic-core.h (warning_at): Rename overload to...
1525 (warning_meta): ...this.
1526 (emit_diagnostic_valist): Delete decl of overload taking
1527 diagnostic_metadata.
1528 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1529 (warning_at): Rename overload taking diagnostic_metadata to...
1530 (warning_meta): ...this.
1532 2020-01-28 Richard Biener <rguenther@suse.de>
1534 PR tree-optimization/93439
1535 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1536 * tree-cfg.c (move_sese_region_to_fn): ... here.
1537 (verify_types_in_gimple_reference): Verify used cliques are
1540 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1543 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1544 argument of a pointer to struct gcc_options and pass it to
1545 parse_mtune_ctrl_str.
1546 (ix86_function_specific_restore): Pass opts to
1547 set_ix86_tune_features.
1548 (ix86_option_override_internal): Likewise.
1549 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1550 gcc_options and use it for x_ix86_tune_ctrl_string.
1552 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1554 PR rtl-optimization/87763
1555 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1556 simplification to handle subregs as well as bare regs.
1557 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1559 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1561 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1562 for reduction chains that (now) include a call.
1564 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1566 PR tree-optimization/92822
1567 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1568 out the don't-care elements of a vector whose significant elements
1569 are duplicates, make the don't-care elements duplicates too.
1571 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1573 PR tree-optimization/93434
1574 * tree-predcom.c (split_data_refs_to_components): Record which
1575 components have had aliasing loads removed. Prevent store-store
1576 commoning for all such components.
1578 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1581 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1582 -1 or is_vshift is true, use new_vector with number of elts npatterns
1583 rather than new_unary_operation.
1585 PR tree-optimization/93454
1586 * gimple-fold.c (fold_array_ctor_reference): Perform
1587 elt_size.to_uhwi () just once, instead of calling it in every
1588 iteration. Punt if that value is above size of the temporary
1589 buffer. Decrease third native_encode_expr argument when
1590 bufoff + elt_sz is above size of buf.
1592 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1594 * config/mips/mips.c (mips_declare_object_name)
1595 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1597 2020-01-27 Martin Liska <mliska@suse.cz>
1599 PR gcov-profile/93403
1600 * tree-profile.c (gimple_init_gcov_profiler): Generate
1601 both __gcov_indirect_call_profiler_v4 and
1602 __gcov_indirect_call_profiler_v4_atomic.
1604 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1607 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1609 (@aarch64_split_simd_mov<mode>): Use it.
1610 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1611 Leave the vec_extract patterns to handle 2-element vectors.
1612 (aarch64_simd_mov_from_<mode>high): Likewise.
1613 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1614 (vec_extractv2dfv1df): Likewise.
1616 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1618 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1619 jump conditions for *compare_condjump<GPI:mode>.
1621 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1624 * digraph.cc (test_edge::test_edge): Specify template for base
1627 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1629 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1631 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1633 * config/arc/arc-protos.h (gen_mlo): Remove.
1634 (gen_mhi): Likewise.
1635 * config/arc/arc.c (AUX_MULHI): Define.
1636 (arc_must_save_reister): Special handling for r58/59.
1637 (arc_compute_frame_size): Consider mlo/mhi registers.
1638 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1640 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1641 mlo/mhi name selection.
1642 (arc_restore_callee_saves): Don't early restore blink when ISR.
1643 (arc_expand_prologue): Add mlo/mhi saving.
1644 (arc_expand_epilogue): Add mlo/mhi restoring.
1647 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1648 numbering when MUL64 option is used.
1649 (DWARF2_FRAME_REG_OUT): Define.
1650 * config/arc/arc.md (arc600_stall): New pattern.
1651 (VUNSPEC_ARC_ARC600_STALL): Define.
1652 (mulsi64): Use correct mlo/mhi registers.
1653 (mulsi_600): Clean it up.
1654 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1656 (mhi_operand): Likewise.
1658 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1659 Petro Karashchenko <petro.karashchenko@ring.com>
1661 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1662 attributes if needed.
1663 (prepare_move_operands): Generate special unspec instruction for
1665 (arc_isuncached_mem_p): Propagate uncached attribute to each
1667 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1668 (VUNSPEC_ARC_STDI): Likewise.
1669 (ALLI): New mode iterator.
1670 (mALLI): New mode attribute.
1671 (lddi): New instruction pattern.
1673 (stdidi_split): Split instruction for architectures which are not
1674 supporting ll64 option.
1675 (lddidi_split): Likewise.
1677 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1679 PR rtl-optimization/92989
1680 * lra-lives.c (process_bb_lives): Update the live-in set before
1681 processing additional clobbers.
1683 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1685 PR rtl-optimization/93170
1686 * cselib.c (cselib_invalidate_regno_val): New function, split out
1688 (cselib_invalidate_regno): ...here.
1689 (cselib_invalidated_by_call_p): New function.
1690 (cselib_process_insn): Iterate over all the hard-register entries in
1691 REG_VALUES and invalidate any that cross call-clobbered registers.
1693 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1695 * dojump.c (split_comparison): Use HONOR_NANS rather than
1696 HONOR_SNANS when splitting LTGT.
1698 2020-01-27 Martin Liska <mliska@suse.cz>
1701 * opts.c (print_filtered_help): Exclude language-specific
1702 options from --help=common unless enabled in all FEs.
1704 2020-01-27 Martin Liska <mliska@suse.cz>
1706 * opts.c (print_help): Exclude params from
1707 all except --help=param.
1709 2020-01-27 Martin Liska <mliska@suse.cz>
1712 * config/i386/i386-features.c (make_resolver_func):
1713 Align the code with ppc64 target implementation.
1714 Do not generate a unique name for resolver function.
1716 2020-01-27 Richard Biener <rguenther@suse.de>
1718 PR tree-optimization/93397
1719 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1720 converted reduction chain SLP graph adjustment.
1722 2020-01-26 Marek Polacek <polacek@redhat.com>
1725 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1728 2020-01-26 Jason Merrill <jason@redhat.com>
1731 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1734 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1736 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1737 (rx_setmem): Likewise.
1739 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1742 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1743 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1744 drop <di> from constraint of last operand.
1747 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1748 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1749 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1751 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1754 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1757 2020-01-24 Jeff Law <law@redhat.com>
1759 PR tree-optimization/92788
1760 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1763 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1766 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1767 *avx_vperm_broadcast_<mode>,
1768 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1769 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1770 Move before avx2_perm<mode>/avx512f_perm<mode>.
1773 * simplify-rtx.c (simplify_const_unary_operation,
1774 simplify_const_binary_operation): Punt for mode precision above
1775 MAX_BITSIZE_MODE_ANY_INT.
1777 2020-01-24 Andrew Pinski <apinski@marvell.com>
1779 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1782 2020-01-24 Jeff Law <law@redhat.com>
1785 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1786 for REGs. Call output_operand_lossage to get more reasonable
1789 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1791 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1792 gcn_fp_compare_operator.
1793 (vec_cmpu<mode>di): Use gcn_compare_operator.
1794 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1795 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1796 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1797 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1798 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1799 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1800 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1801 gcn_fp_compare_operator.
1802 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1803 gcn_fp_compare_operator.
1804 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1805 gcn_fp_compare_operator.
1806 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1807 gcn_fp_compare_operator.
1809 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1811 * doc/install.texi (Cross-Compiler-Specific Options): Document
1812 `--with-toolexeclibdir' option.
1814 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1816 * target.def (flags_regnum): Also mention effect on delay slot filling.
1817 * doc/tm.texi: Regenerate.
1819 2020-01-23 Jeff Law <law@redhat.com>
1821 PR translation/90162
1822 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1824 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1827 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1830 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1832 PR rtl-optimization/93402
1833 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1836 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1838 * config.in: Regenerated.
1839 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1840 for TARGET_LIBC_GNUSTACK.
1841 * configure: Regenerated.
1842 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1843 found to be 2.31 or greater.
1845 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1847 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1849 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1850 (mips_asm_file_end): New function. Delegate to
1851 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1852 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1854 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1857 * config/i386/i386-modes.def (POImode): New mode.
1858 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1859 * config/i386/i386.md (DPWI): New mode attribute.
1860 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1862 (QPWI): ... this. Use POI instead of OI for TImode.
1863 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1864 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1867 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1870 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1872 (speculation_tracker_rev): New pattern.
1873 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1874 Use speculation_tracker_rev to track the inverse condition.
1876 2020-01-23 Richard Biener <rguenther@suse.de>
1878 PR tree-optimization/93381
1879 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1880 alias-set of the def as argument and record the first one.
1881 (vn_walk_cb_data::first_set): New member.
1882 (vn_reference_lookup_3): Pass the alias-set of the current def
1883 to push_partial_def. Fix alias-set used in the aggregate copy
1885 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1886 * real.c (clear_significand_below): Fix out-of-bound access.
1888 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1891 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1892 New define_insn patterns.
1894 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1896 * doc/sourcebuild.texi (check-function-bodies): Add an
1897 optional target/xfail selector.
1899 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1901 PR rtl-optimization/93124
1902 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1903 bare USE and CLOBBER insns.
1905 2020-01-22 Andrew Pinski <apinski@marvell.com>
1907 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1909 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1912 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1913 diagnostic_manager into "ana" namespace.
1914 * selftest-run-tests.c (selftest::run_tests): Update for move of
1915 selftest::run_analyzer_selftests to
1916 ana::selftest::run_analyzer_selftests.
1918 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1920 * cfgexpand.c (union_stack_vars): Update the size.
1922 2020-01-22 Richard Biener <rguenther@suse.de>
1924 PR tree-optimization/93381
1925 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1926 throughout, handle all conversions the same.
1928 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1931 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1932 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1933 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1934 Call force_reg on high_in2 unconditionally.
1936 2020-01-22 Martin Liska <mliska@suse.cz>
1938 PR tree-optimization/92924
1939 * profile.c (compute_value_histograms): Divide
1942 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1945 * output.h (assemble_name_resolve): Declare.
1946 * varasm.c (assemble_name_resolve): New function.
1947 (assemble_name): Use it.
1948 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1950 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1952 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1953 update_web_docs_git instead of update_web_docs_svn.
1955 2020-01-21 Andrew Pinski <apinski@marvell.com>
1958 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1959 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1960 (*tlsgd_small_<mode>): Likewise.
1961 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1962 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1963 register. Convert that register back to dest using convert_mode.
1965 2020-01-21 Jim Wilson <jimw@sifive.com>
1967 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1970 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1971 Uros Bizjak <ubizjak@gmail.com>
1974 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1976 (legitimize_tls_address): Do GNU2 TLS address computation in
1977 ptr_mode and zero-extend result to Pmode.
1978 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1979 :P with :PTR and Pmode with ptr_mode.
1980 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1981 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1982 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1984 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1987 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1988 the last two operands are CONST_INT_P before using them as such.
1990 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1992 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1993 to get the integer element types.
1995 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1997 * config/aarch64/aarch64-sve-builtins.h
1998 (function_expander::convert_to_pmode): Declare.
1999 * config/aarch64/aarch64-sve-builtins.cc
2000 (function_expander::convert_to_pmode): New function.
2001 (function_expander::get_contiguous_base): Use it.
2002 (function_expander::prepare_gather_address_operands): Likewise.
2003 * config/aarch64/aarch64-sve-builtins-sve2.cc
2004 (svwhilerw_svwhilewr_impl::expand): Likewise.
2006 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
2009 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
2010 cfun->machine->label_is_assembled.
2011 (aarch64_print_patchable_function_entry): New.
2012 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
2013 * config/aarch64/aarch64.h (struct machine_function): New field,
2016 2020-01-21 David Malcolm <dmalcolm@redhat.com>
2019 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
2022 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2025 * cgraph.c (cgraph_edge::resolve_speculation,
2026 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
2027 call_stmt_site_hash.
2029 2020-01-21 Martin Liska <mliska@suse.cz>
2031 * config/rs6000/rs6000.c (common_mode_defined): Remove
2034 2020-01-21 Richard Biener <rguenther@suse.de>
2036 PR tree-optimization/92328
2037 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
2038 type when value-numbering same-sized store by inserting a
2040 (eliminate_dom_walker::eliminate_stmt): When eliminating
2041 a redundant store handle bit-reinterpretation of the same value.
2043 2020-01-21 Andrew Pinski <apinski@marvel.com>
2046 * tree-into-ssa.c (prepare_block_for_update_1): Split out
2048 (prepare_block_for_update): This. Use a worklist instead of
2051 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2053 * gcc/config/arm/arm.c (clear_operation_p):
2054 Initialise last_regno, skip first iteration
2055 based on the first_set value and use ints instead
2056 of the unnecessary HOST_WIDE_INTs.
2058 2020-01-21 Jakub Jelinek <jakub@redhat.com>
2061 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
2062 compare_mode other than SFmode or DFmode.
2064 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
2067 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
2068 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
2069 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
2071 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2073 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
2075 2020-01-20 Andrew Pinski <apinski@marvell.com>
2078 * targhooks.c (default_print_patchable_function_entry): Use
2079 output_asm_insn to emit the nop instruction.
2081 2020-01-20 Fangrui Song <maskray@google.com>
2084 * targhooks.c (default_print_patchable_function_entry): Align to
2087 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
2090 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
2091 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
2092 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
2093 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
2094 (*tls_dynamic_gnu2_lea_64): Renamed to ...
2095 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
2096 Remove the {q} suffix from lea.
2097 (*tls_dynamic_gnu2_call_64): Renamed to ...
2098 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
2099 (*tls_dynamic_gnu2_combine_64): Renamed to ...
2100 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
2101 Pass Pmode to gen_tls_dynamic_gnu2_64.
2103 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
2105 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
2107 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
2109 * config/aarch64/aarch64-sve-builtins-base.cc
2110 (svld1ro_impl::memory_vector_mode): Remove parameter name.
2112 2020-01-20 Richard Biener <rguenther@suse.de>
2115 * dwarf2out.c (prune_unused_types): Unconditionally mark
2116 called function DIEs.
2118 2020-01-20 Martin Liska <mliska@suse.cz>
2120 PR tree-optimization/93199
2121 * tree-eh.c (struct leh_state): Add
2122 new field outer_non_cleanup.
2123 (cleanup_is_dead_in): Pass leh_state instead
2124 of eh_region. Add a checking that state->outer_non_cleanup
2125 points to outer non-clean up region.
2126 (lower_try_finally): Record outer_non_cleanup
2128 (lower_catch): Likewise.
2129 (lower_eh_filter): Likewise.
2130 (lower_eh_must_not_throw): Likewise.
2131 (lower_cleanup): Likewise.
2133 2020-01-20 Richard Biener <rguenther@suse.de>
2135 PR tree-optimization/93094
2136 * tree-vectorizer.h (vect_loop_versioning): Adjust.
2137 (vect_transform_loop): Likewise.
2138 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
2139 loop_vectorized_call to vect_transform_loop.
2140 * tree-vect-loop.c (vect_transform_loop): Pass down
2141 loop_vectorized_call to vect_loop_versioning.
2142 * tree-vect-loop-manip.c (vect_loop_versioning): Use
2143 the earlier discovered loop_vectorized_call.
2145 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
2147 * doc/contribute.texi: Update for SVN -> Git transition.
2148 * doc/install.texi: Likewise.
2150 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2153 * cgraph.c (cgraph_edge::make_speculative): Increase number of
2154 speculative targets.
2155 (verify_speculative_call): New function
2156 (cgraph_node::verify_node): Use it.
2157 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
2160 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
2163 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
2164 (cgraph_edge::make_direct): Remove all indirect targets.
2165 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
2166 (cgraph_node::verify_node): Verify that only one call_stmt or
2167 lto_stmt_uid is set.
2168 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
2170 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
2171 (lto_output_ref): Simplify streaming of stmt.
2172 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
2174 2020-01-18 Tamar Christina <tamar.christina@arm.com>
2176 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
2177 Mark parameter unused.
2179 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
2181 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
2183 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
2185 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2187 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2189 * Makefile.in: Add coroutine-passes.o.
2190 * builtin-types.def (BT_CONST_SIZE): New.
2191 (BT_FN_BOOL_PTR): New.
2192 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2193 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2194 * coroutine-builtins.def: New file.
2195 * coroutine-passes.cc: New file.
2196 * function.h (struct GTY function): Add a bit to indicate that the
2197 function is a coroutine component.
2198 * internal-fn.c (expand_CO_FRAME): New.
2199 (expand_CO_YIELD): New.
2200 (expand_CO_SUSPN): New.
2201 (expand_CO_ACTOR): New.
2202 * internal-fn.def (CO_ACTOR): New.
2206 * passes.def: Add pass_coroutine_lower_builtins,
2207 pass_coroutine_early_expand_ifns.
2208 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2209 (make_pass_coroutine_early_expand_ifns): New.
2210 * doc/invoke.texi: Document the fcoroutines command line
2213 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2215 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2218 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2219 after checking the argument is a REG. Don't use REGNO (reg)
2220 again to set last_regno, reuse regno variable instead.
2222 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2224 * doc/analyzer.texi (Limitations): Add note about NaN.
2226 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2227 Sudakshina Das <sudi.das@arm.com>
2229 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2230 and valid immediate.
2231 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2232 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2233 * config/arm/constraints.md (Pg): New.
2234 * config/arm/predicates.md (long_shift_imm): New.
2235 (arm_reg_or_long_shift_imm): Likewise.
2236 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2237 (thumb2_lsll): Likewise.
2240 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2241 Sudakshina Das <sudi.das@arm.com>
2243 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2244 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2245 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2246 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2247 * config/arm/thumb2.md (thumb2_asrl): New.
2248 (thumb2_lsll): Likewise.
2250 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2252 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2255 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2257 * gdbinit.in (help-gcc-hooks): New command.
2258 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2259 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2262 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2264 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2265 correct target macro.
2267 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2269 * config/aarch64/aarch64-protos.h
2270 (aarch64_sve_ld1ro_operand_p): New.
2271 * config/aarch64/aarch64-sve-builtins-base.cc
2272 (class load_replicate): New.
2273 (class svld1ro_impl): New.
2274 (class svld1rq_impl): Change to inherit from load_replicate.
2275 (svld1ro): New sve intrinsic function base.
2276 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2277 New DEF_SVE_FUNCTION.
2278 * config/aarch64/aarch64-sve-builtins-base.h
2279 (svld1ro): New decl.
2280 * config/aarch64/aarch64-sve-builtins.cc
2281 (function_expander::add_mem_operand): Modify assert to allow
2283 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2285 * config/aarch64/aarch64.c
2286 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2287 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2288 (aarch64_sve_ld1ro_operand_p): New.
2289 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2290 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2291 * config/aarch64/predicates.md
2292 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2294 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2296 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2297 Introduce this ACLE specified predefined macro.
2298 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2299 (fp): Disabling this disables f64mm.
2300 (simd): Disabling this disables f64mm.
2301 (fp16): Disabling this disables f64mm.
2302 (sve): Disabling this disables f64mm.
2303 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2304 (AARCH64_ISA_F64MM): New.
2305 (TARGET_F64MM): New.
2306 * doc/invoke.texi (f64mm): Document new option.
2308 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2310 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2311 (neoversen1_tunings): Likewise.
2313 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2316 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2317 Add assert to ensure prolog has been emitted.
2318 (aarch64_split_atomic_op): Likewise.
2319 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2320 Use epilogue_completed rather than reload_completed.
2321 (aarch64_atomic_exchange<mode>): Likewise.
2322 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2323 (atomic_nand<mode>): Likewise.
2324 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2325 (atomic_fetch_nand<mode>): Likewise.
2326 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2327 (atomic_nand_fetch<mode>): Likewise.
2329 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2332 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2334 (REVERSE_CONDITION): Delete.
2335 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2336 (CCFP_CCFPE): Likewise.
2337 (e): New mode attribute.
2338 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2339 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2340 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2341 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2342 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2343 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2344 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2345 name of generator from gen_ccmpdi to gen_ccmpccdi.
2346 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2347 the previous comparison but aren't able to, use the new ccmp_rev
2350 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2352 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2353 than testing directly for INTEGER_CST.
2354 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2356 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2358 PR tree-optimization/93292
2359 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2360 get_vectype_for_scalar_type returns NULL.
2362 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2364 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2365 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2367 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2369 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2371 * params.opt: (max-predicted-iterations): Set bounds.
2372 * predict.c (real_almost_one, real_br_prob_base,
2373 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2374 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2375 probabilities; do not truncate to reg_br_prob_bases.
2376 (estimate_loops_at_level): Pass max_cyclic_prob.
2377 (estimate_loops): Compute max_cyclic_prob.
2378 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2380 * profile-count.c (profile_probability::to_sreal): New.
2381 * profile-count.h (class sreal): Move up in file.
2382 (profile_probability::to_sreal): Declare.
2384 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2387 (arm_invalid_conversion): New function for target hook.
2388 (arm_invalid_unary_op): New function for target hook.
2389 (arm_invalid_binary_op): New function for target hook.
2391 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2393 * config.gcc: Add arm_bf16.h.
2394 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2395 (arm_simd_builtin_std_type): Add BFmode.
2396 (arm_init_simd_builtin_types): Define element types for vector types.
2397 (arm_init_bf16_types): New function.
2398 (arm_init_builtins): Add arm_init_bf16_types function call.
2399 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2400 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2401 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2402 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2403 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2404 (arm_mangle_type): Add __bf16.
2405 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2406 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2407 arm_bf16_ptr_type_node.
2408 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2409 define_split between ARM registers.
2410 * config/arm/arm_bf16.h: New file.
2411 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2412 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2414 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2415 * config/arm/vfp.md: Add BFmode to movhf patterns.
2417 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2418 Andre Vieira <andre.simoesdiasvieira@arm.com>
2420 * config/arm/arm-cpus.in (mve, mve_float): New features.
2421 (dsp, mve, mve.fp): New options.
2422 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2423 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2424 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2426 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2427 Thomas Preud'homme <thomas.preudhomme@arm.com>
2429 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2431 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2432 error for using -mcmse when targeting Armv8.1-M Mainline.
2434 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2435 Thomas Preud'homme <thomas.preudhomme@arm.com>
2437 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2438 address in r4 when targeting Armv8.1-M Mainline.
2439 (nonsecure_call_value_internal): Likewise.
2440 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2441 a register match_operand again. Emit BLXNS when targeting
2443 (nonsecure_call_value_reg_thumb2): Likewise.
2445 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2446 Thomas Preud'homme <thomas.preudhomme@arm.com>
2448 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2449 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2450 variable as true when floating-point ABI is not hard. Replace
2451 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2452 Generate VLSTM and VLLDM instruction respectively before and
2453 after a function call to cmse_nonsecure_call function.
2454 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2455 (VUNSPEC_VLLDM): Likewise.
2456 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2457 (lazy_load_multiple_insn): Likewise.
2459 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2460 Thomas Preud'homme <thomas.preudhomme@arm.com>
2462 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2463 (arm_emit_vfp_multi_reg_pop): Likewise.
2464 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2465 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2466 restore callee-saved VFP registers.
2468 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2469 Thomas Preud'homme <thomas.preudhomme@arm.com>
2471 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2472 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2473 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2474 callee-saved GPRs as well as clear ip register before doing a nonsecure
2475 call then restore callee-saved GPRs after it when targeting
2477 (arm_reorg): Adapt to function rename.
2479 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2480 Thomas Preud'homme <thomas.preudhomme@arm.com>
2482 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2483 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2484 clear_vfp_multiple pattern based on a new vfp parameter.
2485 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2486 targeting Armv8.1-M Mainline.
2487 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2488 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2489 whether VFP registers are available before looking call_used_regs for a
2491 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2492 of prototype of clear_operation_p.
2493 (clear_vfp_multiple_operation): New predicate.
2494 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2495 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2497 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2498 Thomas Preud'homme <thomas.preudhomme@arm.com>
2500 * config/arm/arm-protos.h (clear_operation_p): Declare.
2501 * config/arm/arm.c (clear_operation_p): New function.
2502 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2503 targeting Armv8.1-M Mainline or successor.
2504 (output_return_instruction): Only output APSR register clearing if
2505 Armv8.1-M Mainline instructions not available.
2506 (thumb_exit): Likewise.
2507 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2508 * config/arm/thumb2.md (clear_apsr): New define_insn.
2509 (clear_multiple): Likewise.
2510 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2512 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2513 Thomas Preud'homme <thomas.preudhomme@arm.com>
2515 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2516 (use_return_insn): Also return false for Armv8.1-M Mainline.
2517 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2518 Mainline instructions are available.
2519 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2520 when targeting Armv8.1-M Mainline Security Extensions.
2521 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2522 Mainline entry function.
2523 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2524 targeting Armv8.1-M Mainline or successor.
2525 (arm_expand_epilogue): Fix indentation of caller-saved register
2526 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2528 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2529 (FP_SYSREGS): Likewise.
2530 (enum vfp_sysregs_encoding): Define enum.
2531 (fp_sysreg_names): Declare.
2532 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2533 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2534 (pop_fpsysreg_insn): Likewise.
2536 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2537 Thomas Preud'homme <thomas.preudhomme@arm.com>
2539 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2540 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2541 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2542 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2543 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2544 (ARMv8_1m_main): New feature group.
2545 (armv8.1-m.main): New architecture.
2546 * config/arm/arm-tables.opt: Regenerate.
2547 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2548 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2549 (arm_options_perform_arch_sanity_checks): Error out when targeting
2550 Armv8.1-M Mainline Security Extensions.
2551 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2553 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2555 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2556 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2557 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2558 aarch64_bfdot_laneq): New.
2559 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2560 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2561 vbfdotq_laneq_f32): New.
2562 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2563 VBFMLA_W, VBF): New.
2564 (isquadop): Add V4BF, V8BF.
2566 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2568 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2569 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2570 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2571 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2572 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2573 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2574 usdot_laneq, sudot_lane,sudot_laneq): New.
2575 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2576 (aarch64_<sur>dot_lane): New.
2577 * config/aarch64/arm_neon.h (vusdot_s32): New.
2579 (vusdot_lane_s32): New.
2580 (vsudot_lane_s32): New.
2581 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2582 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2584 2020-01-16 Martin Liska <mliska@suse.cz>
2586 * value-prof.c (dump_histogram_value): Fix
2587 obvious spacing issue.
2589 2020-01-16 Andrew Pinski <apinski@marvell.com>
2591 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2592 !storage_order_barrier_p.
2594 2020-01-16 Andrew Pinski <apinski@marvell.com>
2596 * sched-int.h (_dep): Add unused bit-field field for the padding.
2597 * sched-deps.c (init_dep_1): Init unused field.
2599 2020-01-16 Andrew Pinski <apinski@marvell.com>
2601 * optabs.h (create_expand_operand): Initialize target field also.
2603 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2605 PR tree-optimization/92429
2606 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2607 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2609 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2612 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2614 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2615 aarch64_sve_int_mode to each mode.
2617 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2619 * doc/analyzer.texi (Overview): Add note about
2620 -fdump-ipa-analyzer.
2622 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2624 PR tree-optimization/93231
2625 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2626 input_type is unsigned. Use tree_to_shwi for shift constant.
2627 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2628 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2629 inputs more efficiently.
2631 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2633 * config/i386/i386.md (*movsf_internal): Do not require
2634 SSE2 ISA for alternatives 14 and 15.
2636 2020-01-15 Richard Biener <rguenther@suse.de>
2639 * tree-eh.c (sink_clobbers): If we already visited the destination
2640 block do not defer insertion.
2641 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2642 the purpose of defered insertion.
2644 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2646 * BASE-VER: Bump to 10.0.1.
2648 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2650 PR tree-optimization/93247
2651 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2652 type of the stmt that we're going to vectorize.
2654 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2656 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2657 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2660 2020-01-15 Martin Liska <mliska@suse.cz>
2662 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2663 2 calls of streamer_read_hwi in a function call.
2665 2020-01-15 Richard Biener <rguenther@suse.de>
2667 * alias.c (record_alias_subset): Avoid redundant work when
2668 subset is already recorded.
2670 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2672 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2673 the analyzer options provide CWE identifiers.
2675 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2677 * tree-diagnostic-path.cc (path_summary::event_range::print):
2678 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2679 using get_pure_location.
2681 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2683 PR tree-optimization/93262
2684 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2685 perform head trimming only if the last argument is constant,
2686 either all ones, or larger or equal to head trim, in the latter
2687 case decrease the last argument by head_trim.
2689 PR tree-optimization/93249
2690 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2691 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2692 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2693 perform head trim unless we can prove there are no '\0' chars
2694 from the source among the first head_trim chars.
2696 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2698 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2700 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2703 * config/i386/sse.md
2704 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2705 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2706 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2707 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2708 just a single alternative instead of two, make operands 1 and 2
2711 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2714 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2717 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2719 * Makefile.in (lang_opt_files): Add analyzer.opt.
2720 (ANALYZER_OBJS): New.
2721 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2722 tristate.o and ANALYZER_OBJS.
2723 (TEXI_GCCINT_FILES): Add analyzer.texi.
2724 * common.opt (-fanalyzer): New driver option.
2725 * config.in: Regenerate.
2726 * configure: Regenerate.
2727 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2728 (gccdepdir): Also create depdir for "analyzer" subdir.
2729 * digraph.cc: New file.
2730 * digraph.h: New file.
2731 * doc/analyzer.texi: New file.
2732 * doc/gccint.texi ("Static Analyzer") New menu item.
2733 (analyzer.texi): Include it.
2734 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2735 ("Warning Options"): Add static analysis warnings to the list.
2736 (-Wno-analyzer-double-fclose): New option.
2737 (-Wno-analyzer-double-free): New option.
2738 (-Wno-analyzer-exposure-through-output-file): New option.
2739 (-Wno-analyzer-file-leak): New option.
2740 (-Wno-analyzer-free-of-non-heap): New option.
2741 (-Wno-analyzer-malloc-leak): New option.
2742 (-Wno-analyzer-possible-null-argument): New option.
2743 (-Wno-analyzer-possible-null-dereference): New option.
2744 (-Wno-analyzer-null-argument): New option.
2745 (-Wno-analyzer-null-dereference): New option.
2746 (-Wno-analyzer-stale-setjmp-buffer): New option.
2747 (-Wno-analyzer-tainted-array-index): New option.
2748 (-Wno-analyzer-use-after-free): New option.
2749 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2750 (-Wno-analyzer-use-of-uninitialized-value): New option.
2751 (-Wanalyzer-too-complex): New option.
2752 (-fanalyzer-call-summaries): New warning.
2753 (-fanalyzer-checker=): New warning.
2754 (-fanalyzer-fine-grained): New warning.
2755 (-fno-analyzer-state-merge): New warning.
2756 (-fno-analyzer-state-purge): New warning.
2757 (-fanalyzer-transitivity): New warning.
2758 (-fanalyzer-verbose-edges): New warning.
2759 (-fanalyzer-verbose-state-changes): New warning.
2760 (-fanalyzer-verbosity=): New warning.
2761 (-fdump-analyzer): New warning.
2762 (-fdump-analyzer-callgraph): New warning.
2763 (-fdump-analyzer-exploded-graph): New warning.
2764 (-fdump-analyzer-exploded-nodes): New warning.
2765 (-fdump-analyzer-exploded-nodes-2): New warning.
2766 (-fdump-analyzer-exploded-nodes-3): New warning.
2767 (-fdump-analyzer-supergraph): New warning.
2768 * doc/sourcebuild.texi (dg-require-dot): New.
2769 (dg-check-dot): New.
2770 * gdbinit.in (break-on-saved-diagnostic): New command.
2771 * graphviz.cc: New file.
2772 * graphviz.h: New file.
2773 * ordered-hash-map-tests.cc: New file.
2774 * ordered-hash-map.h: New file.
2775 * passes.def (pass_analyzer): Add before
2776 pass_ipa_whole_program_visibility.
2777 * selftest-run-tests.c (selftest::run_tests): Call
2778 selftest::ordered_hash_map_tests_cc_tests.
2779 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2781 * shortest-paths.h: New file.
2782 * timevar.def (TV_ANALYZER): New timevar.
2783 (TV_ANALYZER_SUPERGRAPH): Likewise.
2784 (TV_ANALYZER_STATE_PURGE): Likewise.
2785 (TV_ANALYZER_PLAN): Likewise.
2786 (TV_ANALYZER_SCC): Likewise.
2787 (TV_ANALYZER_WORKLIST): Likewise.
2788 (TV_ANALYZER_DUMP): Likewise.
2789 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2790 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2791 * tree-pass.h (make_pass_analyzer): New decl.
2792 * tristate.cc: New file.
2793 * tristate.h: New file.
2795 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2798 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2799 alternatives 9 and 10.
2801 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2803 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2804 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2805 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2806 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2807 (selftest::hash_map_tests_c_tests): Call it.
2808 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2809 New static constant, using the value of = H::empty_zero_p.
2810 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2811 from default_hash_traits <Value>.
2812 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2814 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2815 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2816 calls to mark_empty with !Descriptor::empty_zero_p.
2817 (hash_table::empty_slow): Conditionalize the memset call with a
2818 check that Descriptor::empty_zero_p; otherwise, loop through the
2819 entries calling mark_empty on them.
2820 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2821 (pointer_hash::empty_zero_p): Likewise.
2822 (pair_hash::empty_zero_p): Likewise.
2823 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2825 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2826 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2827 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2828 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2829 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2830 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2831 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2832 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2834 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2837 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2839 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2840 fix typo on return value.
2842 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2845 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2847 (cgraph_edge::make_speculative): Add param for setting speculative_id
2849 (cgraph_edge::speculative_call_info): Update comments and find reference
2850 by speculative_id for multiple indirect targets.
2851 (cgraph_edge::resolve_speculation): Decrease the speculations
2852 for indirect edge, drop it's speculative if not direct target
2853 left. Update comments.
2854 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2855 (cgraph_node::dump): Print num_speculative_call_targets.
2856 (cgraph_node::verify_node): Don't report error if speculative
2857 edge not include statement.
2858 (cgraph_edge::num_speculative_call_targets_p): New function.
2859 * cgraph.h (int common_target_id): Remove.
2860 (int common_target_probability): Remove.
2861 (num_speculative_call_targets): New variable.
2862 (make_speculative): Add param for setting speculative_id.
2863 (cgraph_edge::num_speculative_call_targets_p): New declare.
2864 (target_prob): New variable.
2865 (speculative_id): New variable.
2866 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2867 call summaries for multiple speculative call targets.
2868 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2869 * ipa-profile.c (struct speculative_call_target): New struct.
2870 (class speculative_call_summary): New class.
2871 (class speculative_call_summaries): New class.
2872 (call_sums): New variable.
2873 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2874 (ipa_profile_write_edge_summary): New function.
2875 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2876 (ipa_profile_dump_all_summaries): New function.
2877 (ipa_profile_read_edge_summary): New function.
2878 (ipa_profile_read_summary_section): New function.
2879 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2880 (ipa_profile): Generate num_speculative_call_targets from
2882 * ipa-ref.h (speculative_id): New variable.
2883 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2884 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2885 common_target_probability. Stream out speculative_id and
2886 num_speculative_call_targets.
2887 (input_edge): Likewise.
2888 * predict.c (dump_prediction): Remove edges count assert to be
2890 * symtab.c (symtab_node::create_reference): Init speculative_id.
2891 (symtab_node::clone_references): Clone speculative_id.
2892 (symtab_node::clone_referring): Clone speculative_id.
2893 (symtab_node::clone_reference): Clone speculative_id.
2894 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2895 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2896 if indirect call contains multiple speculative targets.
2897 * value-prof.h (check_ic_target): Remove.
2898 * value-prof.c (gimple_value_profile_transformations):
2899 Use void function gimple_ic_transform.
2900 * value-prof.c (gimple_ic_transform): Handle topn case.
2901 Fix comment typos. Change it to a void function.
2903 2020-01-13 Andrew Pinski <apinski@marvell.com>
2905 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2906 (octeontx2t98): New define.
2907 (octeontx2t96): New define.
2908 (octeontx2t93): New define.
2909 (octeontx2f95): New define.
2910 (octeontx2f95n): New define.
2911 (octeontx2f95mm): New define.
2912 * config/aarch64/aarch64-tune.md: Regenerate.
2913 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2915 2020-01-13 Jason Merrill <jason@redhat.com>
2917 PR c++/33799 - destroy return value if local cleanup throws.
2918 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2920 2020-01-13 Martin Liska <mliska@suse.cz>
2922 * ipa-cp.c (get_max_overall_size): Use newly
2923 renamed param param_ipa_cp_unit_growth.
2924 * params.opt: Remove legacy param name.
2926 2020-01-13 Martin Sebor <msebor@redhat.com>
2928 PR tree-optimization/93213
2929 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2930 stores to be eliminated.
2932 2020-01-13 Martin Liska <mliska@suse.cz>
2934 * opts.c (print_help): Do not print CL_PARAM
2935 and CL_WARNING for CL_OPTIMIZATION.
2937 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2940 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2941 depending on optimization settings.
2943 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2945 PR tree-optimization/90838
2946 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2947 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2948 argument rather than to initialize temporary for targets that
2949 don't use the mode argument at all. Initialize ctzval to avoid
2952 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2954 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2955 * tree-core.h: Document it.
2956 * gimplify.c (gimplify_omp_workshare): Set it.
2957 * omp-low.c (lower_omp_target): Use it.
2958 * tree-pretty-print.c (dump_omp_clause): Print it.
2960 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2961 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2963 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2965 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2966 * common.opt (fdiagnostics-path-format=): New option.
2967 (diagnostic_path_format): New enum.
2968 (fdiagnostics-show-path-depths): New option.
2969 * coretypes.h (diagnostic_event_id_t): New forward decl.
2970 * diagnostic-color.c (color_dict): Add "path".
2971 * diagnostic-event-id.h: New file.
2972 * diagnostic-format-json.cc (json_from_expanded_location): Make
2974 (json_end_diagnostic): Call context->make_json_for_path if it
2975 exists and the diagnostic has a path.
2976 (diagnostic_output_format_init): Clear context->print_path.
2977 * diagnostic-path.h: New file.
2978 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2979 when printing a run of events in a diagnostic_path so that they
2980 all get the same color.
2981 (layout::m_diagnostic_path_p): New field.
2982 (layout::layout): Initialize it.
2983 (layout::print_any_labels): Don't colorize the label text for an
2984 event in a diagnostic_path.
2985 (gcc_rich_location::add_location_if_nearby): Add
2986 "restrict_to_current_line_spans" and "label" params. Pass the
2987 former to layout.maybe_add_location_range; pass the latter
2988 when calling add_range.
2989 * diagnostic.c: Include "diagnostic-path.h".
2990 (diagnostic_initialize): Initialize context->path_format and
2991 context->show_path_depths.
2992 (diagnostic_show_any_path): New function.
2993 (diagnostic_path::interprocedural_p): New function.
2994 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2995 (simple_diagnostic_path::num_events): New function.
2996 (simple_diagnostic_path::get_event): New function.
2997 (simple_diagnostic_path::add_event): New function.
2998 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2999 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
3000 (debug): New overload taking a diagnostic_path *.
3001 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
3002 * diagnostic.h (enum diagnostic_path_format): New enum.
3003 (json::value): New forward decl.
3004 (diagnostic_context::path_format): New field.
3005 (diagnostic_context::show_path_depths): New field.
3006 (diagnostic_context::print_path): New callback field.
3007 (diagnostic_context::make_json_for_path): New callback field.
3008 (diagnostic_show_any_path): New decl.
3009 (json_from_expanded_location): New decl.
3010 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
3011 (-fdiagnostics-show-path-depths): New option.
3012 (-fdiagnostics-color): Add "path" to description of default
3013 GCC_COLORS; describe it.
3014 (-fdiagnostics-format=json): Document how diagnostic paths are
3015 represented in the JSON output format.
3016 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
3017 Add optional params "restrict_to_current_line_spans" and "label".
3018 * opts.c (common_handle_option): Handle
3019 OPT_fdiagnostics_path_format_ and
3020 OPT_fdiagnostics_show_path_depths.
3021 * pretty-print.c: Include "diagnostic-event-id.h".
3022 (pp_format): Implement "%@" format code for printing
3023 diagnostic_event_id_t *.
3024 (selftest::test_pp_format): Add tests for "%@".
3025 * selftest-run-tests.c (selftest::run_tests): Call
3026 selftest::tree_diagnostic_path_cc_tests.
3027 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
3028 * toplev.c (general_init): Initialize global_dc->path_format and
3029 global_dc->show_path_depths.
3030 * tree-diagnostic-path.cc: New file.
3031 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
3032 non-static. Drop "diagnostic" param in favor of storing the
3033 original value of "where" and re-using it.
3034 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
3035 maybe_unwind_expanded_macro_loc.
3036 (tree_diagnostics_defaults): Initialize context->print_path and
3037 context->make_json_for_path.
3038 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
3040 (default_tree_make_json_for_path): New decl.
3041 (maybe_unwind_expanded_macro_loc): New decl.
3043 2020-01-10 Jakub Jelinek <jakub@redhat.com>
3045 PR tree-optimization/93210
3046 * fold-const.h (native_encode_initializer,
3047 can_native_interpret_type_p): Declare.
3048 * fold-const.c (native_encode_string): Fix up handling with off != -1,
3050 (native_encode_initializer): New function, moved from dwarf2out.c.
3051 Adjust to native_encode_expr compatible arguments, including dry-run
3052 and partial extraction modes. Don't handle STRING_CST.
3053 (can_native_interpret_type_p): No longer static.
3054 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
3055 offset / BITS_PER_UNIT fits into int and don't call it if
3056 can_native_interpret_type_p fails. If suboff is NULL and for
3057 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
3058 native_encode_initializer.
3059 (fold_const_aggregate_ref_1): Formatting fix.
3060 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
3061 (tree_add_const_value_attribute): Adjust caller.
3063 PR tree-optimization/90838
3064 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
3065 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
3066 CTZ_DEFINED_VALUE_AT_ZERO.
3068 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
3071 * lra-constraints.c (match_reload): Permit input operands have the
3072 same mode as output while other input operands have a different
3075 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
3077 PR tree-optimization/90838
3078 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
3079 (check_ctz_string): Likewise.
3080 (optimize_count_trailing_zeroes): Likewise.
3081 (simplify_count_trailing_zeroes): Likewise.
3082 (pass_forwprop::execute): Try ctz simplification.
3083 * match.pd: Add matching for ctz idioms.
3085 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3087 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
3089 (aarch64_invalid_unary_op): New function for target hook.
3090 (aarch64_invalid_binary_op): New function for target hook.
3092 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
3094 * config.gcc: Add arm_bf16.h.
3095 * config/aarch64/aarch64-builtins.c
3096 (aarch64_simd_builtin_std_type): Add BFmode.
3097 (aarch64_init_simd_builtin_types): Define element types for vector
3099 (aarch64_init_bf16_types): New function.
3100 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
3101 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
3103 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
3104 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
3106 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
3107 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
3108 * config/aarch64/aarch64.c
3109 (aarch64_classify_vector_mode): Add support for BF types.
3110 (aarch64_gimplify_va_arg_expr): Add support for BF types.
3111 (aarch64_vq_mode): Add support for BF types.
3112 (aarch64_simd_container_mode): Add support for BF types.
3113 (aarch64_mangle_type): Add support for BF scalar type.
3114 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
3115 * config/aarch64/arm_bf16.h: New file.
3116 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
3117 * config/aarch64/iterators.md: Add BF types to mode attributes.
3118 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
3120 2020-01-10 Jason Merrill <jason@redhat.com>
3122 PR c++/93173 - incorrect tree sharing.
3123 * gimplify.c (copy_if_shared): No longer static.
3124 * gimplify.h: Declare it.
3126 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3128 * doc/invoke.texi (-msve-vector-bits=): Document that
3129 -msve-vector-bits=128 now generates VL-specific code for
3130 little-endian targets.
3131 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
3132 build_vector_type_for_mode to construct the data vector types.
3133 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
3134 VL-specific code for -msve-vector-bits=128 on little-endian targets.
3135 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
3136 for 128-bit vectors.
3138 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3140 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
3143 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3145 * config/aarch64/aarch64-builtins.c
3146 (aarch64_builtin_vectorized_function): Check for specific vector modes,
3147 rather than checking the number of elements and the element mode.
3149 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3151 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
3152 get_related_vectype_for_scalar_type rather than build_vector_type
3153 to create the index type for a conditional reduction.
3155 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
3157 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
3158 for any type of gather or scatter, including strided accesses.
3160 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3162 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
3165 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
3167 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
3169 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
3170 parameter and its use to reset DR_OFFSET's.
3171 (vect_transform_loop): Remove orig_drs_init argument.
3172 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
3173 member of dr_vec_info rather than the offset of the associated
3174 data_reference's innermost_loop_behavior.
3175 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
3176 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
3177 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
3178 get_dr_vinfo_offset.
3179 (vectorizable_store): Likewise.
3180 (vectorizable_load): Likewise.
3182 2020-01-10 Richard Biener <rguenther@suse.de>
3184 * gimple-ssa-store-merging
3185 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3187 2020-01-10 Martin Liska <mliska@suse.cz>
3190 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3191 encapsulation that was there before r280040.
3193 2020-01-10 Richard Biener <rguenther@suse.de>
3196 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3197 sequences to avoid walking them again for secondary opportunities.
3198 (pass_lower_eh_dispatch::execute): Instead actually insert
3201 2020-01-10 Richard Biener <rguenther@suse.de>
3204 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3205 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3206 avoid quadraticness.
3208 2020-01-10 Martin Jambor <mjambor@suse.cz>
3210 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3211 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3212 to get param_ipa_sra_max_replacements.
3213 (param_splitting_across_edge): Pass the caller to
3214 pull_accesses_from_callee.
3216 2020-01-10 Martin Jambor <mjambor@suse.cz>
3218 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3219 * ipa-cp.c (max_new_size): Removed.
3220 (orig_overall_size): New variable.
3221 (get_max_overall_size): New function.
3222 (estimate_local_effects): Use it. Adjust dump.
3223 (decide_about_value): Likewise.
3224 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3225 orig_overall_size. Adjust dump.
3226 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3228 2020-01-10 Martin Jambor <mjambor@suse.cz>
3230 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3231 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3232 instead of param_ipa_max_agg_items.
3233 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3234 optimization info for the callee.
3236 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3238 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3239 markers if debug_inline_points is false.
3241 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3243 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3245 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3246 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3247 aarch64-sve-builtins-sve2.h.
3248 (aarch64-sve-builtins-sve2.o): New rule.
3249 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3250 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3251 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3252 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3253 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3254 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3256 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3257 instructions that are handled here.
3258 (@cond_asrd<mode>): Generalize to...
3259 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3260 (*cond_asrd<mode>_2): Generalize to...
3261 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3262 (*cond_asrd<mode>_z): Generalize to...
3263 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3264 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3265 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3266 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3267 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3269 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3270 (@aarch64_scatter_stnt<mode>): Likewise.
3271 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3272 (@aarch64_mul_lane_<mode>): Likewise.
3273 (@aarch64_sve_suqadd<mode>_const): Likewise.
3274 (*<sur>h<addsub><mode>): Generalize to...
3275 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3277 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3278 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3279 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3280 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3281 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3282 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3283 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3284 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3285 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3286 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3287 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3288 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3289 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3290 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3291 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3292 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3293 (@aarch64_sve2_xar<mode>): Likewise.
3294 (@aarch64_sve2_bcax<mode>): Likewise.
3295 (*aarch64_sve2_eor3<mode>): Rename to...
3296 (@aarch64_sve2_eor3<mode>): ...this.
3297 (@aarch64_sve2_bsl<mode>): New expander.
3298 (@aarch64_sve2_nbsl<mode>): Likewise.
3299 (@aarch64_sve2_bsl1n<mode>): Likewise.
3300 (@aarch64_sve2_bsl2n<mode>): Likewise.
3301 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3302 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3303 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3304 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3305 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3306 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3307 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3308 (<su>mull<bt><Vwide>): Generalize to...
3309 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3311 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3312 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3313 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3314 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3315 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3316 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3317 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3318 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3319 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3320 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3321 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3322 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3323 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3324 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3325 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3326 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3327 (<SHRNB:r>shrnb<mode>): Generalize to...
3328 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3330 (<SHRNT:r>shrnt<mode>): Generalize to...
3331 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3333 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3334 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3335 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3336 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3337 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3338 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3339 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3340 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3341 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3342 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3343 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3344 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3345 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3346 (@aarch64_sve2_cvtnt<mode>): Likewise.
3347 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3348 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3349 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3350 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3351 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3352 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3353 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3354 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3355 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3356 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3357 (@aarch64_sve2_pmul<mode>): Likewise.
3358 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3359 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3360 (@aarch64_sve2_tbl2<mode>): Likewise.
3361 (@aarch64_sve2_tbx<mode>): Likewise.
3362 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3363 (@aarch64_sve2_histcnt<mode>): Likewise.
3364 (@aarch64_sve2_histseg<mode>): Likewise.
3365 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3366 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3367 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3368 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3369 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3370 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3371 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3372 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3373 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3374 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3375 (SVE2_PMULL_PAIR_I): New mode iterators.
3376 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3377 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3378 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3379 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3380 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3381 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3382 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3383 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3384 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3385 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3386 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3387 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3388 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3389 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3390 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3391 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3392 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3393 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3394 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3395 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3396 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3397 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3398 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3399 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3400 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3401 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3402 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3403 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3404 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3405 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3406 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3408 (VNARROW, Ventype): New mode attributes.
3409 (Vewtype): Handle VNx2DI. Fix typo in comment.
3410 (VDOUBLE): New mode attribute.
3411 (sve_lane_con): Handle VNx8HI.
3412 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3413 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3414 (sve_int_op, sve_int_op_rev): Handle the above codes.
3415 (sve_pred_int_rhs2_operand): Likewise.
3416 (MULLBT, SHRNB, SHRNT): Delete.
3417 (SVE_INT_SHIFT_IMM): New int iterator.
3418 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3419 and UNSPEC_WHILEHS for TARGET_SVE2.
3420 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3421 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3422 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3423 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3424 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3425 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3426 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3427 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3428 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3429 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3430 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3431 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3432 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3433 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3434 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3435 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3436 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3437 (optab): Handle the new unspecs.
3438 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3440 (lr): Handle the new unspecs.
3442 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3443 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3444 (sve_int_qsub_op): New int attributes.
3445 (sve_fp_op, rot): Handle the new unspecs.
3446 * config/aarch64/aarch64-sve-builtins.h
3447 (function_resolver::require_matching_pointer_type): Declare.
3448 (function_resolver::resolve_unary): Add an optional boolean argument.
3449 (function_resolver::finish_opt_n_resolution): Add an optional
3450 type_suffix_index argument.
3451 (gimple_folder::redirect_call): Declare.
3452 (gimple_expander::prepare_gather_address_operands): Add an optional
3454 * config/aarch64/aarch64-sve-builtins.cc: Include
3455 aarch64-sve-builtins-sve2.h.
3456 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3457 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3458 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3459 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3460 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3461 (TYPES_s_unsigned): Likewise.
3462 (TYPES_s_integer): Use TYPES_s_unsigned.
3463 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3464 (TYPES_sd_integer): Use them.
3465 (TYPES_d_unsigned): New macro.
3466 (TYPES_d_integer): Use it.
3467 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3468 (TYPES_cvt_narrow): Likewise.
3469 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3470 (preds_mx): New variable.
3471 (function_builder::add_overloaded_function): Allow the new feature
3472 set to be more restrictive than the original one.
3473 (function_resolver::infer_pointer_type): Remove qualifiers from
3474 the pointer type before printing it.
3475 (function_resolver::require_matching_pointer_type): New function.
3476 (function_resolver::resolve_sv_displacement): Handle functions
3477 that don't support 32-bit vector indices or svint32_t vector offsets.
3478 (function_resolver::finish_opt_n_resolution): Take the inferred type
3479 as a separate argument.
3480 (function_resolver::resolve_unary): Optionally treat all forms in
3481 the same way as normal merging functions.
3482 (gimple_folder::redirect_call): New function.
3483 (function_expander::prepare_gather_address_operands): Add an argument
3484 that says whether scaled forms are available. If they aren't,
3485 handle scaling of vector indices and don't add the extension and
3487 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3488 fall back to using cond_* instead.
3489 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3490 Split out the member variables into...
3491 (rtx_code_function_base): ...this new base class.
3492 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3493 (unspec_based_function): Split out the member variables into...
3494 (unspec_based_function_base): ...this new base class.
3495 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3496 (unspec_based_function_exact_insn): New class.
3497 (unspec_based_add_function, unspec_based_add_lane_function)
3498 (unspec_based_lane_function, unspec_based_pred_function)
3499 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3500 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3501 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3503 (unspec_based_fused_function): New class.
3504 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3505 (unspec_based_fused_lane_function): New class.
3506 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3508 (CODE_FOR_MODE1): New macro.
3509 (fixed_insn_function): New class.
3510 (while_comparison): Likewise.
3511 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3512 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3513 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3514 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3515 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3516 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3517 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3518 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3519 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3520 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3521 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3522 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3523 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3524 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3525 Also add an initial argument for unary_convert_narrowt, regardless
3526 of the predication type.
3527 (build_32_64): Allow loads and stores to specify MODE_none.
3528 (build_sv_index64, build_sv_uint_offset): New functions.
3529 (long_type_suffix): New function.
3530 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3531 (binary_imm_long_base, load_gather_sv_base): Likewise.
3532 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3533 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3534 (unary_narrowb_base, unary_narrowt_base): Likewise.
3535 (binary_long_lane_def, binary_long_lane): New shape.
3536 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3537 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3538 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3539 (binary_to_uint_def, binary_to_uint): Likewise.
3540 (binary_wide_def, binary_wide): Likewise.
3541 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3542 (compare_def, compare): Likewise.
3543 (compare_ptr_def, compare_ptr): Likewise.
3544 (load_ext_gather_index_restricted_def,
3545 load_ext_gather_index_restricted): Likewise.
3546 (load_ext_gather_offset_restricted_def,
3547 load_ext_gather_offset_restricted): Likewise.
3548 (load_gather_sv_def): Inherit from load_gather_sv_base.
3549 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3550 (shift_left_imm_def, shift_left_imm): Likewise.
3551 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3552 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3553 (store_scatter_index_restricted_def,
3554 store_scatter_index_restricted): Likewise.
3555 (store_scatter_offset_restricted_def,
3556 store_scatter_offset_restricted): Likewise.
3557 (tbl_tuple_def, tbl_tuple): Likewise.
3558 (ternary_long_lane_def, ternary_long_lane): Likewise.
3559 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3560 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3561 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3562 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3563 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3564 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3565 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3566 (ternary_uint_def, ternary_uint): Likewise.
3567 (unary_convert): Fix typo in comment.
3568 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3569 (unary_long_def, unary_long): Likewise.
3570 (unary_narrowb_def, unary_narrowb): Likewise.
3571 (unary_narrowt_def, unary_narrowt): Likewise.
3572 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3573 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3574 (unary_to_int_def, unary_to_int): Likewise.
3575 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3576 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3577 (svasrd_impl): Delete.
3578 (svcadd_impl::expand): Handle integer operations too.
3579 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3580 new functions to derive the unspec numbers.
3581 (svmla_svmls_lane_impl): Replace with...
3582 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3583 integer operations too.
3584 (svwhile_impl): Rename to...
3585 (svwhilelx_impl): ...this and inherit from while_comparison.
3586 (svasrd): Use unspec_based_function.
3587 (svmla_lane): Use svmla_lane_impl.
3588 (svmls_lane): Use svmls_lane_impl.
3589 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3590 (svwhilele, svwhilelt): Use svwhilelx_impl.
3591 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3592 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3593 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3594 * config/aarch64/aarch64-sve-builtins.def: Include
3595 aarch64-sve-builtins-sve2.def.
3597 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3599 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3600 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3601 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3602 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3603 immediates as well as vector ones.
3604 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3605 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3606 (aarch64_sve_qsub_immediate): Update calls accordingly.
3608 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3610 * config/aarch64/aarch64-sve2.md: Add banner comments.
3611 (<su>mulh<r>s<mode>3): Move further up file.
3612 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3613 (*aarch64_sve2_sra<mode>): Move further down file.
3614 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3616 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3618 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3620 (while_optab_cmp): Handle them.
3621 * config/aarch64/aarch64-sve.md
3622 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3623 and add a "@" marker.
3624 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3625 instead of gen_aarch64_sve2_while_ptest.
3626 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3628 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3630 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3631 (UNSPEC_WHILELE): ...this.
3632 (UNSPEC_WHILE_LO): Rename to...
3633 (UNSPEC_WHILELO): ...this.
3634 (UNSPEC_WHILE_LS): Rename to...
3635 (UNSPEC_WHILELS): ...this.
3636 (UNSPEC_WHILE_LT): Rename to...
3637 (UNSPEC_WHILELT): ...this.
3638 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3639 (cmp_op, while_optab_cmp): Likewise.
3640 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3641 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3642 (svwhilelt): Likewise.
3644 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3646 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3647 (unary_to_uint): Define.
3648 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3649 (unary_count): Rename to...
3650 (unary_to_uint_def, unary_to_uint): ...this.
3651 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3653 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3655 * config/aarch64/aarch64-sve-builtins-functions.h
3656 (code_for_mode_function): New class.
3657 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3658 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3659 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3660 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3661 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3663 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3665 * config/aarch64/iterators.md (addsub): New code attribute.
3666 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3668 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3669 in the asm string and attributes. Fix indentation.
3670 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3672 (@aarch64_sve_<optab><mode>): ...this.
3673 * config/aarch64/aarch64-sve-builtins.h
3674 (function_expander::expand_signed_unpred_op): Delete.
3675 * config/aarch64/aarch64-sve-builtins.cc
3676 (function_expander::expand_signed_unpred_op): Likewise.
3677 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3678 try using code_for_aarch64_sve instead.
3679 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3680 (svqsub_impl): Likewise.
3681 (svqadd, svqsub): Use rtx_code_function instead.
3683 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3685 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3686 (HADDSUB, sur, addsub): Remove them.
3688 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3690 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3691 functions the same way, rather than singling out those that
3692 aren't mapped directly to optabs.
3694 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3696 * target.def (compatible_vector_types_p): New target hook.
3697 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3698 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3699 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3700 * doc/tm.texi: Regenerate.
3701 * gimple-expr.c: Include target.h.
3702 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3703 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3705 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3706 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3707 Use the original predicate if it already has a suitable type.
3709 2020-01-09 Martin Jambor <mjambor@suse.cz>
3711 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3712 resolve_speculation and redirect_call_stmt_to_callee static. Change
3713 return type of set_call_stmt to cgraph_edge *.
3714 * auto-profile.c (afdo_indirect_call): Adjust call to
3715 redirect_call_stmt_to_callee.
3716 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3717 make the this pointer explicit, adjust self-recursive calls and the
3718 call top make_direct. Return the resulting edge.
3719 (cgraph_edge::remove): Make this pointer explicit.
3720 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3721 (cgraph_edge::make_direct): Likewise, adjust call to
3722 resolve_speculation.
3723 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3724 call to set_call_stmt.
3725 (cgraph_update_edges_for_call_stmt_node): Update call to
3726 set_call_stmt and remove.
3727 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3728 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3729 (cgraph_node::create_edge_including_clones): Moved "first" definition
3730 of edge to the block where it was used. Adjusted calls to
3732 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3733 cgraph_edge::remove.
3734 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3735 make_direct and redirect_call_stmt_to_callee.
3736 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3737 resolve_speculation and make_direct.
3738 * ipa-inline-transform.c (inline_transform): Adjust call to
3739 redirect_call_stmt_to_callee.
3740 (check_speculations_1):: Adjust call to resolve_speculation.
3741 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3742 resolve-speculation.
3743 (inline_small_functions): Adjust call to resolve_speculation.
3744 (ipa_inline): Likewise.
3745 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3747 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3748 safe with regards to edge removal, adjust calls to
3749 redirect_call_stmt_to_callee.
3750 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3751 and redirect_call_stmt_to_callee.
3752 * multiple_target.c (create_dispatcher_calls): Adjust call to
3753 redirect_call_stmt_to_callee
3754 (redirect_to_specific_clone): Likewise.
3755 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3756 Adjust calls to cgraph_edge::remove.
3757 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3758 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3759 (expand_call_inline): Adjust call to cgraph_edge::remove.
3761 2020-01-09 Martin Liska <mliska@suse.cz>
3763 * params.opt: Set Optimization for
3764 param_max_speculative_devirt_maydefs.
3766 2020-01-09 Martin Sebor <msebor@redhat.com>
3770 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3772 2020-01-09 Martin Liska <mliska@suse.cz>
3774 * auto-profile.c (auto_profile): Use opt_for_fn
3776 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3777 (propagate_vals_across_arith_jfunc): Likewise.
3778 (hint_time_bonus): Likewise.
3779 (incorporate_penalties): Likewise.
3780 (good_cloning_opportunity_p): Likewise.
3781 (perform_estimation_of_a_value): Likewise.
3782 (estimate_local_effects): Likewise.
3783 (ipcp_propagate_stage): Likewise.
3784 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3785 (set_switch_stmt_execution_predicate): Likewise.
3786 (analyze_function_body): Likewise.
3787 * ipa-inline-analysis.c (offline_size): Likewise.
3788 * ipa-inline.c (early_inliner): Likewise.
3789 * ipa-prop.c (ipa_analyze_node): Likewise.
3790 (ipcp_transform_function): Likewise.
3791 * ipa-sra.c (process_scan_results): Likewise.
3792 (ipa_sra_summarize_function): Likewise.
3793 * params.opt: Rename ipcp-unit-growth to
3794 ipa-cp-unit-growth. Add Optimization for various
3795 IPA-related parameters.
3797 2020-01-09 Richard Biener <rguenther@suse.de>
3800 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3802 2020-01-09 Richard Biener <rguenther@suse.de>
3804 PR tree-optimization/93040
3805 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3807 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3809 * common/config/avr/avr-common.c (avr_option_optimization_table)
3810 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3812 2020-01-09 Martin Liska <mliska@suse.cz>
3814 * cgraphclones.c (symbol_table::materialize_all_clones):
3815 Use cgraph_node::dump_name.
3817 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3820 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3821 output_operand_lossage instead of gcc_unreachable.
3822 * doc/md.texi (riscv f constraint): Fix typo.
3825 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3826 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3827 CONST_SCALAR_INT_P instead of CONST_INT_P.
3828 (*subv<mode>4_1): Rename to ...
3829 (subv<mode>4_1): ... this.
3830 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3831 define_insn_and_split patterns.
3832 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3835 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3837 * vec.c (class selftest::count_dtor): New class.
3838 (selftest::test_auto_delete_vec): New test.
3839 (selftest::vec_c_tests): Call it.
3840 * vec.h (class auto_delete_vec): New class template.
3841 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3843 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3845 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3847 2020-01-08 Jim Wilson <jimw@sifive.com>
3849 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3850 use of TLS_MODEL_LOCAL_EXEC when not pic.
3852 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3854 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3857 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3860 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3861 *stack_protect_set_3 peephole2): Also check that the second
3862 insns source is general_operand.
3865 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3866 predicate for output operand instead of register_operand.
3867 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3868 memory destination and non-memory operands[2].
3870 2020-01-08 Martin Liska <mliska@suse.cz>
3872 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3873 ::dump_asm_name instead of (::name or ::asm_name).
3874 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3875 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3876 (analyze_functions): Likewise.
3877 (expand_all_functions): Likewise.
3878 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3879 (propagate_bits_across_jump_function): Likewise.
3880 (dump_profile_updates): Likewise.
3881 (ipcp_store_bits_results): Likewise.
3882 (ipcp_store_vr_results): Likewise.
3883 * ipa-devirt.c (dump_targets): Likewise.
3884 * ipa-fnsummary.c (analyze_function_body): Likewise.
3885 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3886 (process_hsa_functions): Likewise.
3887 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3888 (set_alias_uids): Likewise.
3889 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3890 * ipa-inline.c (recursive_inlining): Likewise.
3891 (inline_to_all_callers_1): Likewise.
3892 (ipa_inline): Likewise.
3893 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3894 (ipa_propagate_frequency): Likewise.
3895 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3896 (remove_described_reference): Likewise.
3897 * ipa-pure-const.c (worse_state): Likewise.
3898 (check_retval_uses): Likewise.
3899 (analyze_function): Likewise.
3900 (propagate_pure_const): Likewise.
3901 (propagate_nothrow): Likewise.
3902 (dump_malloc_lattice): Likewise.
3903 (propagate_malloc): Likewise.
3904 (pass_local_pure_const::execute): Likewise.
3905 * ipa-visibility.c (optimize_weakref): Likewise.
3906 (function_and_variable_visibility): Likewise.
3907 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3908 (ipa_discover_variable_flags): Likewise.
3909 * lto-streamer-out.c (output_function): Likewise.
3910 (output_constructor): Likewise.
3911 * tree-inline.c (copy_bb): Likewise.
3912 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3913 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3915 2020-01-08 Richard Biener <rguenther@suse.de>
3918 * tree-eh.c (sink_clobbers): Update virtual operands for
3919 the first and last stmt only. Add a dry-run capability.
3920 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3921 after CFG manipulations and in RPO order to catch all
3922 secondary opportunities reliably.
3924 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3927 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3929 2019-01-08 Richard Biener <rguenther@suse.de>
3932 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3933 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3934 virtual operand, also updating SSA use.
3935 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3936 Update stmt after resetting virtual operand.
3937 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3938 * gimple-iterator.c (gsi_remove): When not removing the stmt
3939 permanently do not delink immediate uses or mark the stmt modified.
3941 2020-01-08 Martin Liska <mliska@suse.cz>
3943 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3944 (ipa_call_context::estimate_size_and_time): Likewise.
3945 (inline_analyze_function): Likewise.
3947 2020-01-08 Martin Liska <mliska@suse.cz>
3949 * cgraph.c (cgraph_node::dump): Use systematically
3952 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3954 Add -nodevicespecs option for avr.
3957 * config/avr/avr.opt (-nodevicespecs): New driver option.
3958 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3959 "-specs=device-specs/..." if that option is not set.
3960 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3962 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3964 Implement 64-bit double functions for avr.
3967 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3968 --with-double-comparison.
3969 * doc/install.texi: Document them.
3970 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3971 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3972 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3973 * doc/invoke.texi (AVR Built-in Macros): Document them.
3974 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3975 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3976 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3978 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3981 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3982 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3983 when only building rm-profile multilibs.
3985 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3988 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3989 lattice for a value to check.
3990 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3991 finite propagation in self-recursive scc.
3993 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3995 * ipa-inline.c (caller_growth_limits): Restore the AND.
3997 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3999 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
4000 (VEC_ALLREG_ALT): New iterator.
4001 (VEC_ALLREG_INT_MODE): New iterator.
4002 (VCMP_MODE): New iterator.
4003 (VCMP_MODE_INT): New iterator.
4004 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
4005 (vec_cmp<u>v64qidi): New define_expand.
4006 (vec_cmp<mode>di_exec): Use VCMP_MODE.
4007 (vec_cmpu<mode>di_exec): New define_expand.
4008 (vec_cmp<u>v64qidi_exec): New define_expand.
4009 (vec_cmp<mode>di_dup): Use VCMP_MODE.
4010 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
4011 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
4012 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
4013 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
4014 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
4015 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
4016 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
4017 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
4018 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
4020 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
4021 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
4023 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
4025 * config/gcn/constraints.md (DA): Update description and match.
4027 (Db): New constraint.
4028 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
4030 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
4031 Implement 'Db' mixed immediate type.
4032 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
4033 (addcv64si3_dup<exec_vcc>): Delete.
4034 (subcv64si3<exec_vcc>): Rework constraints.
4035 (addv64di3): Rework constraints.
4036 (addv64di3_exec): Rework constraints.
4037 (subv64di3): Rework constraints.
4038 (addv64di3_dup): Delete.
4039 (addv64di3_dup_exec): Delete.
4040 (addv64di3_zext): Rework constraints.
4041 (addv64di3_zext_exec): Rework constraints.
4042 (addv64di3_zext_dup): Rework constraints.
4043 (addv64di3_zext_dup_exec): Rework constraints.
4044 (addv64di3_zext_dup2): Rework constraints.
4045 (addv64di3_zext_dup2_exec): Rework constraints.
4046 (addv64di3_sext_dup2): Rework constraints.
4047 (addv64di3_sext_dup2_exec): Rework constraints.
4049 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
4051 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
4052 existing target checks.
4054 2020-01-07 Richard Biener <rguenther@suse.de>
4056 * doc/install.texi: Bump minimal supported MPC version.
4058 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4060 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
4061 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
4062 * langhooks.c: Include stor-layout.h.
4063 (lhd_simulate_enum_decl): New function.
4064 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
4065 handle_arm_sve_h for the LTO frontend.
4066 (register_vector_type): Cope with null returns from pushdecl.
4068 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
4070 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
4071 (aarch64_sve::nvectors_if_data_type): Replace with...
4072 (aarch64_sve::builtin_type_p): ...this.
4073 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
4074 (find_vector_type): Delete.
4075 (add_sve_type_attribute): New function.
4076 (lookup_sve_type_attribute): Likewise.
4077 (register_builtin_types): Add an "SVE type" attribute to each type.
4078 (register_tuple_type): Likewise.
4079 (svbool_type_p, nvectors_if_data_type): Delete.
4080 (mangle_builtin_type): Use lookup_sve_type_attribute.
4081 (builtin_type_p): Likewise. Add an overload that returns the
4082 number of constituent vector and predicate registers.
4083 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
4084 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
4085 instead of aarch64_sve_argument_p.
4086 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
4087 (aarch64_pass_by_reference): Likewise.
4088 (aarch64_function_value_1): Likewise.
4089 (aarch64_return_in_memory): Likewise.
4090 (aarch64_layout_arg): Likewise.
4092 2020-01-07 Jakub Jelinek <jakub@redhat.com>
4094 PR tree-optimization/93156
4095 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
4096 least significant bit is always clear.
4098 PR tree-optimization/93118
4099 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
4100 simplifier with two intermediate conversions.
4102 2020-01-07 Martin Liska <mliska@suse.cz>
4104 * params.opt: Add Optimization for various parameters.
4106 2020-01-07 Martin Liska <mliska@suse.cz>
4109 * doc/extend.texi: Explain cloning for target_clone
4112 2020-01-07 Martin Liska <mliska@suse.cz>
4114 PR tree-optimization/92860
4115 * common.opt: Make in Optimization option
4116 as it is affected by -O0, which is an Optimization
4118 * tree-inline.c (tree_inlinable_function_p):
4119 Use opt_for_fn for warn_inline.
4120 (expand_call_inline): Likewise.
4122 2020-01-07 Martin Liska <mliska@suse.cz>
4124 PR tree-optimization/92860
4125 * common.opt: Make flag_ree as optimization
4128 2020-01-07 Martin Liska <mliska@suse.cz>
4130 PR optimization/92860
4131 * params.opt: Mark param_min_crossjump_insns with Optimization
4134 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
4136 * ipa-inline-analysis.c (estimate_growth): Fix typo.
4137 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
4139 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
4141 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
4142 helper function to return the valid addressing formats for a given
4143 hard register and mode.
4144 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
4146 * config/rs6000/constraints.md (Q constraint): Update
4148 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
4151 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
4152 Use 'Q' for doing vector extract from memory.
4153 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
4155 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
4156 doing vector extract from memory.
4157 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
4158 extract from memory.
4160 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
4161 for the offset being 34-bits when -mcpu=future is used.
4163 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
4165 * config/pa/pa.md: Revert change to use ordered_comparison_operator
4166 instead of cmpib_comparison_operator in cmpib patterns.
4167 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
4168 of cmpib_comparison_operator. Revise comment.
4170 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4172 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
4173 in an IFN_DIV_POW2 node to be equal.
4175 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4177 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
4178 (vect_check_scalar_mask): ...this.
4179 (vectorizable_store, vectorizable_load): Update call accordingly.
4180 (vectorizable_call): Use vect_check_scalar_mask to check the mask
4181 argument in calls to conditional internal functions.
4183 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4185 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4186 '0' matching inputs.
4187 (subv64di3_exec): Likewise.
4189 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4191 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4192 * doc/md.texi (movstr): Likewise.
4194 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4196 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4199 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4201 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4203 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4204 to a temporary file and use move-if-change to update the real
4205 file where necessary.
4207 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4209 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4210 rather than Upa for CPY /M.
4212 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4214 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4217 2020-01-06 Martin Liska <mliska@suse.cz>
4219 PR tree-optimization/92860
4220 * params.opt: Mark param_max_combine_insns with Optimization
4223 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4226 * config/i386/i386.md (SWIDWI): New mode iterator.
4227 (DWI, dwi): Add TImode variants.
4228 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4229 <general_hilo_operand> instead of <general_operand>. Use
4230 CONST_SCALAR_INT_P instead of CONST_INT_P.
4231 (*addv<mode>4_1): Rename to ...
4232 (addv<mode>4_1): ... this.
4233 (QWI): New mode attribute.
4234 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4235 define_insn_and_split patterns.
4236 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4238 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4239 <general_hilo_operand> instead of <general_operand>.
4240 (*addcarry<mode>_1): New define_insn.
4241 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4243 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4245 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4246 Use "call" instead of "set".
4248 2020-01-03 Martin Jambor <mjambor@suse.cz>
4251 * ipa-cp.c (print_all_lattices): Skip functions without info.
4253 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4256 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4257 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4258 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4259 for 'e' simd clones.
4262 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4264 (mprefer-vector-width=): Add Save.
4265 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4266 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4267 (ix86_debug_options, ix86_function_specific_print): Adjust
4268 ix86_target_string callers.
4269 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4270 (ix86_valid_target_attribute_tree): Likewise.
4271 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4272 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4273 ix86_target_string caller.
4276 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4277 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4278 instead of gen_int_shift_amount + convert_modes.
4280 PR rtl-optimization/93088
4281 * loop-iv.c (find_single_def_src): Punt after looking through
4282 128 reg copies for regs with single definitions. Move definitions
4285 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4287 * config/arm/arm-c.c (arm_cpu_builtins): Define
4288 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4289 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4290 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4291 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4292 * config/arm/arm-tables.opt: Regenerated.
4293 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4294 arm_arch_i8mm and arm_arch_bf16 when enabled.
4295 * config/arm/arm.h (TARGET_I8MM): New macro.
4296 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4297 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4298 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4299 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4300 (v8_6_a_simd_variants): New.
4301 (v8_*_a_simd_variants): Add i8mm and bf16.
4302 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4304 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4307 * predict.c (compute_function_frequency): Don't call
4308 warn_function_cold on functions that already have cold attribute.
4310 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4313 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4314 COMDAT group function labels in .data.rel.ro.local section.
4315 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4318 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4319 comparison_operator in B and S integer comparisons. Likewise, use
4320 ordered_comparison_operator instead of cmpib_comparison_operator in
4322 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4324 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4326 Update copyright years.
4328 * gcc.c (process_command): Update copyright notice dates.
4329 * gcov-dump.c (print_version): Ditto.
4330 * gcov.c (print_version): Ditto.
4331 * gcov-tool.c (print_version): Ditto.
4332 * gengtype.c (create_file): Ditto.
4333 * doc/cpp.texi: Bump @copying's copyright year.
4334 * doc/cppinternals.texi: Ditto.
4335 * doc/gcc.texi: Ditto.
4336 * doc/gccint.texi: Ditto.
4337 * doc/gcov.texi: Ditto.
4338 * doc/install.texi: Ditto.
4339 * doc/invoke.texi: Ditto.
4341 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4343 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4346 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4348 PR tree-optimization/93098
4349 * match.pd (popcount): For shift amounts, use integer_onep
4350 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4351 tests. Make sure that precision is power of two larger than or equal
4352 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4353 instead of ULL suffixed constants. Formatting fixes.
4355 Copyright (C) 2020 Free Software Foundation, Inc.
4357 Copying and distribution of this file, with or without modification,
4358 are permitted in any medium without royalty provided the copyright
4359 notice and this notice are preserved.