1 2020-02-18 Jakub Jelinek <jakub@redhat.com>
4 * cgraph.c (verify_speculative_call): Use speculative_id instead of
5 speculative_uid in messages. Remove trailing whitespace from error
6 message. Use num_speculative_call_targets instead of
7 num_speculative_targets in a message.
8 (cgraph_node::verify_node): Use call_stmt instead of cal_stmt in
9 edge messages and stmt instead of cal_stmt in reference message.
11 PR tree-optimization/93780
12 * tree-ssa.c (non_rewritable_lvalue_p): Check valid_vector_subparts_p
13 before calling build_vector_type.
14 (execute_update_addresses_taken): Likewise.
17 * params.opt (-param=ipa-max-switch-predicate-bounds=): Fix help
18 typo, functoin -> function.
19 * tree.c (free_lang_data_in_decl): Fix comment typo,
21 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
23 2020-02-17 David Malcolm <dmalcolm@redhat.com>
25 * diagnostic.c (print_any_cwe): Don't call get_cwe_url if URLs
27 (print_option_information): Don't call get_option_url if URLs
30 2020-02-17 Alexandre Oliva <oliva@adacore.com>
32 * tree-emutls.c (new_emutls_decl, emutls_common_1): Complete
33 handling of register_common-less targets.
35 2020-02-17 Martin Liska <mliska@suse.cz>
38 * ipa-devirt.c (odr_types_equivalent_p): Fix grammar.
40 2020-02-17 Martin Liska <mliska@suse.cz>
43 * config/rs6000/rs6000.c (rs6000_option_override_internal):
46 2020-02-17 Martin Liska <mliska@suse.cz>
49 * config/rx/elf.opt: Fix typo.
51 2020-02-17 Richard Biener <rguenther@suse.de>
54 * opts-global.c (print_ignored_options): Use inform and
57 2020-02-17 Jiufu Guo <guojiufu@linux.ibm.com>
60 * config/rs6000/rs6000.md (untyped_call): Add emit_clobber.
62 2020-02-16 Uroš Bizjak <ubizjak@gmail.com>
65 * config/i386/i386.md (atan2xf3): Swap operands 1 and 2.
66 (atan2<mode>3): Update operand order in the call to gen_atan2xf3.
68 2020-02-15 Jason Merrill <jason@redhat.com>
70 * doc/invoke.texi (C Dialect Options): Add -std=c++20.
72 2020-02-15 Jakub Jelinek <jakub@redhat.com>
74 PR tree-optimization/93744
75 * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0,
76 A - ((A - B) & -(C cmp D)) -> (C cmp D) ? B : A,
77 A + ((B - A) & -(C cmp D)) -> (C cmp D) ? B : A): For GENERIC, make
78 sure @2 in the first and @1 in the other patterns has no side-effects.
80 2020-02-15 David Malcolm <dmalcolm@redhat.com>
81 Bernd Edlinger <bernd.edlinger@hotmail.de>
85 * config.in (DIAGNOSTICS_URLS_DEFAULT): New define.
86 * configure.ac (--with-diagnostics-urls): New configuration
87 option, based on --with-diagnostics-color.
88 (DIAGNOSTICS_URLS_DEFAULT): New define.
89 * config.h: Regenerate.
90 * configure: Regenerate.
91 * diagnostic.c (diagnostic_urls_init): Handle -1 for
92 DIAGNOSTICS_URLS_DEFAULT from configure-time
93 --with-diagnostics-urls=auto-if-env by querying for a GCC_URLS
94 and TERM_URLS environment variable.
95 * diagnostic-url.h (diagnostic_url_format): New enum type.
96 (diagnostic_urls_enabled_p): rename to...
97 (determine_url_format): ... this, and change return type.
98 * diagnostic-color.c (parse_env_vars_for_urls): New helper function.
99 (auto_enable_urls): Disable URLs on xfce4-terminal, gnome-terminal,
100 the linux console, and mingw.
101 (diagnostic_urls_enabled_p): rename to...
102 (determine_url_format): ... this, and adjust.
103 * pretty-print.h (pretty_printer::show_urls): rename to...
104 (pretty_printer::url_format): ... this, and change to enum.
105 * pretty-print.c (pretty_printer::pretty_printer,
106 pp_begin_url, pp_end_url, test_urls): Adjust.
107 * doc/install.texi (--with-diagnostics-urls): Document the new
108 configuration option.
109 (--with-diagnostics-color): Document the existing interaction
110 with GCC_COLORS better.
111 * doc/invoke.texi (-fdiagnostics-urls): Add GCC_URLS and TERM_URLS
112 vindex reference. Update description of defaults based on the above.
113 (-fdiagnostics-color): Update description of how -fdiagnostics-color
114 interacts with GCC_COLORS.
116 2020-02-14 Eric Botcazou <ebotcazou@adacore.com>
119 * config/sparc/sparc.c (eligible_for_call_delay): Test HAVE_GNU_LD in
120 conjunction with TARGET_GNU_TLS in early return.
122 2020-02-14 Alexander Monakov <amonakov@ispras.ru>
124 * rtlanal.c (rtx_cost): Handle a SET up front. Avoid division if
125 the mode is not wider than UNITS_PER_WORD.
127 2020-02-14 Martin Jambor <mjambor@suse.cz>
129 PR tree-optimization/93516
130 * tree-sra.c (propagate_subaccesses_from_rhs): Do not create
131 access of the same type as the parent.
132 (propagate_subaccesses_from_lhs): Likewise.
134 2020-02-14 Hongtao Liu <hongtao.liu@intel.com>
137 * config/i386/avx512vbmi2intrin.h
138 (_mm512_shrdi_epi16, _mm512_mask_shrdi_epi16,
139 _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
140 _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
141 _m512_shrdi_epi64, _m512_mask_shrdi_epi64,
142 _m512_maskz_shrdi_epi64, _mm512_shldi_epi16,
143 _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16,
144 _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
145 _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
146 _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): Fix typo
147 of lacking a closing parenthesis.
148 * config/i386/avx512vbmi2vlintrin.h
149 (_mm256_shrdi_epi16, _mm256_mask_shrdi_epi16,
150 _mm256_maskz_shrdi_epi16, _mm256_shrdi_epi32,
151 _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32,
152 _m256_shrdi_epi64, _m256_mask_shrdi_epi64,
153 _m256_maskz_shrdi_epi64, _mm256_shldi_epi16,
154 _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
155 _mm256_shldi_epi32, _mm256_mask_shldi_epi32,
156 _mm256_maskz_shldi_epi32, _mm256_shldi_epi64,
157 _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64,
158 _mm_shrdi_epi16, _mm_mask_shrdi_epi16,
159 _mm_maskz_shrdi_epi16, _mm_shrdi_epi32,
160 _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32,
161 _mm_shrdi_epi64, _mm_mask_shrdi_epi64,
162 _m_maskz_shrdi_epi64, _mm_shldi_epi16,
163 _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16,
164 _mm_shldi_epi32, _mm_mask_shldi_epi32,
165 _mm_maskz_shldi_epi32, _mm_shldi_epi64,
166 _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64): Ditto.
168 2020-02-13 H.J. Lu <hongjiu.lu@intel.com>
171 * config/i386/i386.c (ix86_trampoline_init): Skip ENDBR32 at
172 the target function entry.
174 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
176 * common/config/arc/arc-common.c (arc_option_optimization_table):
177 Disable if-conversion step when optimized for size.
179 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
181 * config/arc/arc.c (arc_conditional_register_usage): R0-R3 and
182 R12-R15 are always in ARCOMPACT16_REGS register class.
183 * config/arc/arc.opt (mq-class): Deprecate.
184 * config/arc/constraint.md ("q"): Remove dependency on mq-class
186 * doc/invoke.texi (mq-class): Update text.
187 * common/config/arc/arc-common.c (arc_option_optimization_table):
190 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
192 * config/arc/arc.c (arc_insn_cost): New function.
193 (TARGET_INSN_COST): Define.
194 * config/arc/arc.md (cost): New attribute.
195 (add_n): Use arc_nonmemory_operand.
196 (ashlsi3_insn): Likewise, also update constraints.
197 (ashrsi3_insn): Likewise.
199 (add_shift): Likewise.
200 * config/arc/predicates.md (arc_nonmemory_operand): New predicate.
202 2020-02-13 Claudiu Zissulescu <claziss@synopsys.com>
204 * config/arc/arc.md (mulsidi_600): Correctly select mlo/mhi
206 (umulsidi_600): Likewise.
208 2020-02-13 Jakub Jelinek <jakub@redhat.com>
211 * config/i386/avx512bitalgintrin.h (_mm512_mask_popcnt_epi8,
212 _mm512_mask_popcnt_epi16, _mm256_mask_popcnt_epi8,
213 _mm256_mask_popcnt_epi16, _mm_mask_popcnt_epi8,
214 _mm_mask_popcnt_epi16): Rename __B argument to __A and __A to __W,
215 pass __A to the builtin followed by __W instead of __A followed by
217 * config/i386/avx512vpopcntdqintrin.h (_mm512_mask_popcnt_epi32,
218 _mm512_mask_popcnt_epi64): Likewise.
219 * config/i386/avx512vpopcntdqvlintrin.h (_mm_mask_popcnt_epi32,
220 _mm256_mask_popcnt_epi32, _mm_mask_popcnt_epi64,
221 _mm256_mask_popcnt_epi64): Likewise.
223 PR tree-optimization/93582
224 * fold-const.h (shift_bytes_in_array_left,
225 shift_bytes_in_array_right): Declare.
226 * fold-const.c (shift_bytes_in_array_left,
227 shift_bytes_in_array_right): New function, moved from
228 gimple-ssa-store-merging.c, no longer static.
229 * gimple-ssa-store-merging.c (shift_bytes_in_array): Move
230 to gimple-ssa-store-merging.c and rename to shift_bytes_in_array_left.
231 (shift_bytes_in_array_right): Move to gimple-ssa-store-merging.c.
232 (encode_tree_to_bitpos): Use shift_bytes_in_array_left instead of
233 shift_bytes_in_array.
234 (verify_shift_bytes_in_array): Rename to ...
235 (verify_shift_bytes_in_array_left): ... this. Use
236 shift_bytes_in_array_left instead of shift_bytes_in_array.
237 (store_merging_c_tests): Call verify_shift_bytes_in_array_left
238 instead of verify_shift_bytes_in_array.
239 * tree-ssa-sccvn.c (vn_reference_lookup_3): For native_encode_expr
240 / native_interpret_expr where the store covers all needed bits,
241 punt on PDP-endian, otherwise allow all involved offsets and sizes
242 not to be byte-aligned.
245 * config/i386/sse.md (k<code><mode>): Drop mode from last operand and
246 use const_0_to_255_operand predicate instead of immediate_operand.
247 (avx512dq_fpclass<mode><mask_scalar_merge_name>,
248 avx512dq_vmfpclass<mode><mask_scalar_merge_name>,
249 vgf2p8affineinvqb_<mode><mask_name>,
250 vgf2p8affineqb_<mode><mask_name>): Drop mode from
251 const_0_to_255_operand predicated operands.
253 2020-02-12 Jeff Law <law@redhat.com>
255 * config/h8300/h8300.md (comparison shortening peepholes): Use
256 a mode iterator to merge the HImode and SImode peepholes.
258 2020-02-12 Jakub Jelinek <jakub@redhat.com>
261 * real.c (is_even): Make static. Function comment fix.
262 (is_halfway_below): Make static, don't assert R is not inf/nan,
263 instead return false for those. Small formatting fixes.
265 2020-02-12 Martin Sebor <msebor@redhat.com>
268 * tree-ssa-strlen.c (handle_builtin_stxncpy): Rename...
269 (handle_builtin_stxncpy_strncat): ...to this. Change first argument.
270 Issue only -Wstringop-overflow strncat, never -Wstringop-truncation.
271 (strlen_check_and_optimize_call): Adjust callee name.
273 2020-02-12 Jeff Law <law@redhat.com>
275 * config/h8300/h8300.md (comparison shortening peepholes): Drop
276 (and (xor)) variant. Combine other two into single peephole.
278 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
280 PR rtl-optimization/93565
281 * config/aarch64/aarch64.c (aarch64_rtx_costs): Add CTZ costs.
283 2020-02-12 Wilco Dijkstra <wdijkstr@arm.com>
285 * config/aarch64/aarch64-simd.md
286 (aarch64_zero_extend<GPI:mode>_reduc_plus_<VDQV_E:mode>): New pattern.
287 * config/aarch64/aarch64.md (popcount<mode>2): Use it instead of
288 generating separate ADDV and zero_extend patterns.
289 * config/aarch64/iterators.md (VDQV_E): New iterator.
291 2020-02-12 Jeff Law <law@redhat.com>
293 * config/h8300/h8300.md (cpymemsi, movmd): Remove dead patterns,
294 expanders, splits, etc.
295 (movmd_internal_<mode>, movmd splitter, movstr, movsd): Likewise.
296 (stpcpy_internal_<mode>, stpcpy splitter): Likewise.
297 (peepholes to convert QI/HI mode pushes to SI mode pushes): Likewise.
298 * config/h8300/h8300.c (h8300_swap_into_er6): Remove unused function.
299 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise
300 * config/h8300/h8300-protos.h (h8300_swap_into_er6): Remove unused
302 (h8300_swap_out_of_er6, h8sx_emit_movmd): Likewise.
304 2020-02-12 Jakub Jelinek <jakub@redhat.com>
307 * config/i386/sse.md (VI48F_256_DQ): New mode iterator.
308 (avx512vl_vextractf128<mode>): Use it instead of VI48F_256. Remove
309 TARGET_AVX512DQ from condition.
310 (vec_extract_lo_<mode><mask_name>): Use <mask_avx512dq_condition>
311 instead of <mask_mode512bit_condition> in condition. If
312 TARGET_AVX512DQ is false, emit vextract*64x4 instead of
314 (vec_extract_lo_<mode><mask_name>): Drop <mask_avx512dq_condition>
317 2020-02-12 Kewen Lin <linkw@gcc.gnu.org>
320 * ira.c (combine_and_move_insns): Skip multiple_sets def_insn.
322 2020-02-12 Segher Boessenkool <segher@kernel.crashing.org>
324 * config/rs6000/rs6000.c (rs6000_debug_print_mode): Don't use sizeof
325 where strlen is more legible.
326 (rs6000_builtin_vectorized_libmass): Ditto.
327 (rs6000_print_options_internal): Ditto.
329 2020-02-11 Martin Sebor <msebor@redhat.com>
331 PR tree-optimization/93683
332 * tree-ssa-alias.c (stmt_kills_ref_p): Avoid using LHS when not set.
334 2020-02-11 Michael Meissner <meissner@linux.ibm.com>
336 * config/rs6000/predicates.md (cint34_operand): Rename the
337 -mprefixed-addr option to be -mprefixed.
338 * config/rs6000/rs6000-cpus.def (ISA_FUTURE_MASKS_SERVER): Rename
339 the -mprefixed-addr option to be -mprefixed.
340 (OTHER_FUTURE_MASKS): Likewise.
341 (POWERPC_MASKS): Likewise.
342 * config/rs6000/rs6000.c (rs6000_option_override_internal): Rename
343 the -mprefixed-addr option to be -mprefixed. Change error
344 messages to refer to -mprefixed.
345 (num_insns_constant_gpr): Rename the -mprefixed-addr option to be
347 (rs6000_legitimate_offset_address_p): Likewise.
348 (rs6000_mode_dependent_address): Likewise.
349 (rs6000_opt_masks): Change the spelling of "-mprefixed-addr" to be
350 "-mprefixed" for target attributes and pragmas.
351 (address_to_insn_form): Rename the -mprefixed-addr option to be
353 (rs6000_adjust_insn_length): Likewise.
354 * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Rename the
355 -mprefixed-addr option to be -mprefixed.
356 (ASM_OUTPUT_OPCODE): Likewise.
357 * config/rs6000/rs6000.md (prefixed insn attribute): Rename the
358 -mprefixed-addr option to be -mprefixed.
359 * config/rs6000/rs6000.opt (-mprefixed): Rename the
360 -mprefixed-addr option to be prefixed. Change the option from
361 being undocumented to being documented.
362 * doc/invoke.texi (RS/6000 and PowerPC Options): Document the
363 -mprefixed option. Update the -mpcrel documentation to mention
366 2020-02-11 Hans-Peter Nilsson <hp@axis.com>
368 * ira-conflicts.c (print_hard_reg_set): Correct output for sets
369 including FIRST_PSEUDO_REGISTER - 1.
370 * ira-color.c (print_hard_reg_set): Ditto.
372 2020-02-11 Stam Markianos-Wright <stam.markianos-wright@arm.com>
374 * config/arm/arm-builtins.c (enum arm_type_qualifiers):
375 (USTERNOP_QUALIFIERS): New define.
376 (USMAC_LANE_QUADTUP_QUALIFIERS): New define.
377 (SUMAC_LANE_QUADTUP_QUALIFIERS): New define.
378 (arm_expand_builtin_args): Add case ARG_BUILTIN_LANE_QUADTUP_INDEX.
379 (arm_expand_builtin_1): Add qualifier_lane_quadtup_index.
380 * config/arm/arm_neon.h (vusdot_s32): New.
381 (vusdot_lane_s32): New.
382 (vusdotq_lane_s32): New.
383 (vsudot_lane_s32): New.
384 (vsudotq_lane_s32): New.
385 * config/arm/arm_neon_builtins.def (usdot, usdot_lane,sudot_lane): New.
386 * config/arm/iterators.md (DOTPROD_I8MM): New.
387 (sup, opsuffix): Add <us/su>.
388 * config/arm/neon.md (neon_usdot, <us/su>dot_lane: New.
389 * config/arm/unspecs.md (UNSPEC_DOT_US, UNSPEC_DOT_SU): New.
391 2020-02-11 Richard Biener <rguenther@suse.de>
393 PR tree-optimization/93661
394 PR tree-optimization/93662
395 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly guard
397 * tree-sra.c (get_access_for_expr): Likewise.
399 2020-02-10 Jakub Jelinek <jakub@redhat.com>
402 * config/i386/sse.md (VI_256_AVX2): New mode iterator.
403 (vcond_mask_<mode><sseintvecmodelower>): Use it instead of VI_256.
404 Change condition from TARGET_AVX2 to TARGET_AVX.
406 2020-02-10 Iain Sandoe <iain@sandoe.co.uk>
409 * config/darwin-c.c (darwin_cfstring_ref_p): Fix up last
412 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
414 Try to generate zero-based comparisons.
415 * config/cris/cris.c (cris_reduce_compare): New function.
416 * config/cris/cris-protos.h (cris_reduce_compare): Add prototype.
417 * config/cris/cris.md ("cbranch<mode>4", "cbranchdi4", "cstoredi4")
418 (cstore<mode>4"): Apply cris_reduce_compare in expanders.
420 2020-02-10 Richard Earnshaw <rearnsha@arm.com>
423 * config/arm/arm.md (movsi_compare0): Allow SP as a source register
424 in Thumb state and also as a destination in Arm state. Add T16
427 2020-02-10 Hans-Peter Nilsson <hp@axis.com>
429 * md.texi (Define Subst): Match closing paren in example.
431 2020-02-10 Jakub Jelinek <jakub@redhat.com>
435 * config/i386/i386.c (x86_64_elf_section_type_flags): Fix up last
436 arguments of strncmp.
438 2020-02-10 Feng Xue <fxue@os.amperecomputing.com>
441 * ipa-cp.c (ipcp_lattice::add_value): Add source with same call edge
442 but different source value.
443 (adjust_callers_for_value_intersection): New function.
444 (gather_edges_for_value): Adjust order of callers to let a
445 non-self-recursive caller be the first element.
446 (self_recursive_pass_through_p): Add a new parameter "simple", and
447 check generalized self-recursive pass-through jump function.
448 (self_recursive_agg_pass_through_p): Likewise.
449 (find_more_scalar_values_for_callers_subset): Compute value from
450 pass-through jump function for self-recursive.
451 (intersect_with_plats): Cleanup previous implementation code for value
452 itersection with self-recursive call edge.
453 (intersect_with_agg_replacements): Likewise.
454 (intersect_aggregates_with_edge): Deduce value from pass-through jump
455 function for self-recursive call edge. Cleanup previous implementation
456 code for value intersection with self-recursive call edge.
457 (decide_whether_version_node): Remove dead callers and adjust order
458 to let a non-self-recursive caller be the first element.
460 2020-02-09 Uroš Bizjak <ubizjak@gmail.com>
462 * recog.c: Move pass_split_before_sched2 code in front of
463 pass_split_before_regstack.
464 (pass_data_split_before_sched2): Rename pass to split3 from split4.
465 (pass_data_split_before_regstack): Rename pass to split4 from split3.
466 (rest_of_handle_split_before_sched2): Remove.
467 (pass_split_before_sched2::execute): Unconditionally call
469 (enable_split_before_sched2): New function.
470 (pass_split_before_sched2::gate): Use enable_split_before_sched2.
471 (pass_split_before_regstack::gate): Ditto.
472 * config/nds32/nds32.c (nds32_split_double_word_load_store_p):
473 Update name check for renamed split4 pass.
474 * config/sh/sh.c (register_sh_passes): Update pass insertion
475 point for renamed split4 pass.
477 2020-02-09 Jakub Jelinek <jakub@redhat.com>
479 * gimplify.c (gimplify_adjust_omp_clauses_1): Promote
480 DECL_IN_CONSTANT_POOL variables into "omp declare target" to avoid
481 copying them around between host and target.
483 2020-02-08 Andrew Pinski <apinski@marvell.com>
486 * config/aarch64/aarch64-simd.md (movmisalign<mode>): Check
487 STRICT_ALIGNMENT also.
489 2020-02-08 Jim Wilson <jimw@sifive.com>
492 * config/riscv/riscv.h (HARD_REGNO_CALLER_SAVE_MODE): Define.
494 2020-02-08 Uroš Bizjak <ubizjak@gmail.com>
495 Jakub Jelinek <jakub@redhat.com>
498 * config/i386/i386.h (CALL_USED_REGISTERS): Make
499 xmm16-xmm31 call-used even in 64-bit ms-abi.
501 2020-02-07 Dennis Zhang <dennis.zhang@arm.com>
503 * config/aarch64/aarch64-simd-builtins.def (simd_smmla): New entry.
504 (simd_ummla, simd_usmmla): Likewise.
505 * config/aarch64/aarch64-simd.md (aarch64_simd_<sur>mmlav16qi): New.
506 * config/aarch64/arm_neon.h (vmmlaq_s32, vmmlaq_u32): New.
509 2020-02-07 Richard Biener <rguenther@suse.de>
512 * tree-inline.c (fold_marked_statements): Do a PRE walk,
513 skipping unreachable regions.
514 (optimize_inline_calls): Skip folding stmts when we didn't
517 2020-02-07 H.J. Lu <hongjiu.lu@intel.com>
520 * config/i386/i386.c (function_arg_ms_64): Add a type argument.
521 Don't return aggregates with only SFmode and DFmode in SSE
523 (ix86_function_arg): Pass arg.type to function_arg_ms_64.
525 2020-02-07 Jakub Jelinek <jakub@redhat.com>
528 * config/rs6000/rs6000-logue.c
529 (rs6000_emit_probe_stack_range_stack_clash): Always use gen_add3_insn,
530 if it fails, move rs into end_addr and retry. Add
531 REG_FRAME_RELATED_EXPR note whenever it returns more than one insn or
532 the insn pattern doesn't describe well what exactly happens to
536 * config/i386/predicates.md (avx_identity_operand): Remove.
537 * config/i386/sse.md (*avx_vec_concat<mode>_1): Remove.
538 (avx_<castmode><avxsizesuffix>_<castmode>,
539 avx512f_<castmode><avxsizesuffix>_256<castmode>): Change patterns to
540 a VEC_CONCAT of the operand and UNSPEC_CAST.
541 (avx512f_<castmode><avxsizesuffix>_<castmode>): Change pattern to
542 a VEC_CONCAT of VEC_CONCAT of the operand and UNSPEC_CAST with
546 * config/i386/i386.c (ix86_lea_outperforms): Make sure to clear
547 recog_data.insn if distance_non_agu_define changed it.
549 2020-02-06 Michael Meissner <meissner@linux.ibm.com>
552 * config/rs6000/rs6000.c (reg_to_non_prefixed): Before ISA 3.0
553 we only had X-FORM (reg+reg) addressing for vectors. Also before
554 ISA 3.0, we only had X-FORM addressing for scalars in the
555 traditional Altivec registers.
557 2020-02-06 <zhongyunde@huawei.com>
558 Vladimir Makarov <vmakarov@redhat.com>
560 PR rtl-optimization/93561
561 * lra-assigns.c (spill_for): Check that tested hard regno is not out of
564 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
566 * config/aarch64/aarch64.md (aarch64_movk<mode>): Add a type
569 2020-02-06 Segher Boessenkool <segher@kernel.crashing.org>
571 * config/rs6000/rs6000.c (rs6000_emit_set_long_const): Handle the case
572 where the low and the high 32 bits are equal to each other specially,
573 with an rldimi instruction.
575 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
577 * config/arm/arm-cpus.in: Set profile M for armv8.1-m.main.
579 2020-02-06 Mihail Ionescu <mihail.ionescu@arm.com>
581 * config/arm/arm-tables.opt: Regenerate.
583 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
586 * config/aarch64/aarch64-protos.h (aarch64_movk_shift): Declare.
587 * config/aarch64/aarch64.c (aarch64_movk_shift): New function.
588 * config/aarch64/aarch64.md (aarch64_movk<mode>): New pattern.
590 2020-02-06 Richard Sandiford <richard.sandiford@arm.com>
592 PR rtl-optimization/87763
593 * config/aarch64/aarch64.md (*ashiftsi_extvdi_bfiz): New pattern.
595 2020-02-06 Delia Burduv <delia.burduv@arm.com>
597 * config/aarch64/aarch64-simd-builtins.def
598 (bfmlaq): New built-in function.
599 (bfmlalb): New built-in function.
600 (bfmlalt): New built-in function.
601 (bfmlalb_lane): New built-in function.
602 (bfmlalt_lane): New built-in function.
603 * config/aarch64/aarch64-simd.md
604 (aarch64_bfmmlaqv4sf): New pattern.
605 (aarch64_bfmlal<bt>v4sf): New pattern.
606 (aarch64_bfmlal<bt>_lane<q>v4sf): New pattern.
607 * config/aarch64/arm_neon.h (vbfmmlaq_f32): New intrinsic.
608 (vbfmlalbq_f32): New intrinsic.
609 (vbfmlaltq_f32): New intrinsic.
610 (vbfmlalbq_lane_f32): New intrinsic.
611 (vbfmlaltq_lane_f32): New intrinsic.
612 (vbfmlalbq_laneq_f32): New intrinsic.
613 (vbfmlaltq_laneq_f32): New intrinsic.
614 * config/aarch64/iterators.md (BF_MLA): New int iterator.
615 (bt): New int attribute.
617 2020-02-06 Uroš Bizjak <ubizjak@gmail.com>
619 * config/i386/i386.md (*pushtf): Emit "#" instead of
620 calling gcc_unreachable in insn output.
623 (*pushsf_rex64): Ditto for alternatives other than 1.
624 (*pushsf): Ditto for alternatives other than 1.
626 2020-02-06 Martin Liska <mliska@suse.cz>
628 PR gcov-profile/91971
629 PR gcov-profile/93466
630 * coverage.c (coverage_init): Revert mangling of
631 path into filename. It can lead to huge filename length.
632 Creation of subfolders seem more natural.
634 2020-02-06 Stam Markianos-Wright <stam.markianos-wright@arm.com>
637 * config/arm/arm.c (arm_block_arith_comp_libfuncs_for_mode): New.
638 (arm_init_libfuncs): Add BFmode support to block spurious BF libfuncs.
639 Use arm_block_arith_comp_libfuncs_for_mode for HFmode.
641 2020-02-06 Jakub Jelinek <jakub@redhat.com>
644 * config/i386/predicates.md (avx_identity_operand): New predicate.
645 * config/i386/sse.md (*avx_vec_concat<mode>_1): New
646 define_insn_and_split.
649 * omp-low.c (use_pointer_for_field): For nested constructs, also
650 look for map clauses on target construct.
651 (scan_omp_1_stmt) <case GIMPLE_OMP_TARGET>: Bump temporarily
652 taskreg_nesting_level.
655 * gimplify.c (gimplify_scan_omp_clauses) <do_notice>: If adding
656 shared clause, call omp_notice_variable on outer context if any.
658 2020-02-05 Jason Merrill <jason@redhat.com>
661 * symtab.c (symtab_node::nonzero_address): A DECL_COMDAT decl has
662 non-zero address even if weak and not yet defined.
664 2020-02-05 Martin Sebor <msebor@redhat.com>
666 PR tree-optimization/92765
667 * gimple-fold.c (get_range_strlen_tree): Handle MEM_REF and PARM_DECL.
668 * tree-ssa-strlen.c (compute_string_length): Remove.
669 (determine_min_objsize): Remove.
670 (get_len_or_size): Add an argument. Call get_range_strlen_dynamic.
671 Avoid using type size as the upper bound on string length.
672 (handle_builtin_string_cmp): Add an argument. Adjust.
673 (strlen_check_and_optimize_call): Pass additional argument to
674 handle_builtin_string_cmp.
676 2020-02-05 Uroš Bizjak <ubizjak@gmail.com>
678 * config/i386/i386.md (*pushdi2_rex64 peephole2): Remove.
679 (*pushdi2_rex64 peephole2): Unconditionally split after
681 (*ashl<mode>3_doubleword): Ditto.
682 (*<shift_insn><mode>3_doubleword): Ditto.
684 2020-02-05 Michael Meissner <meissner@linux.ibm.com>
687 * config/rs6000/rs6000.c (get_vector_offset): Fix
689 2020-02-05 Andrew Stubbs <ams@codesourcery.com>
691 * config/gcn/t-gcn-hsa (MULTILIB_OPTIONS): Use / not space.
693 2020-02-05 David Malcolm <dmalcolm@redhat.com>
696 (Special Functions for Debugging the Analyzer): Update description
697 of __analyzer_dump_exploded_nodes.
699 2020-02-05 Jakub Jelinek <jakub@redhat.com>
702 * config/i386/i386-features.c (ix86_add_reg_usage_to_vzeroupper): Only
703 include sets and not clobbers in the vzeroupper pattern.
704 * config/i386/sse.md (*avx_vzeroupper): Require in insn condition that
705 the parallel has 17 (64-bit) or 9 (32-bit) elts.
706 (*avx_vzeroupper_1): New define_insn_and_split.
709 * recog.c (pass_split_after_reload::gate): For STACK_REGS targets,
710 don't run when !optimize.
711 (pass_split_before_regstack::gate): For STACK_REGS targets, run even
714 2020-02-05 Richard Biener <rguenther@suse.de>
717 * genmatch.c (dt_node::gen_kids_1): Emit number of argument
718 checks before matching calls.
720 2020-02-05 Jakub Jelinek <jakub@redhat.com>
722 * tree-ssa-alias.c (aliasing_matching_component_refs_p): Fix up
723 function comment typo.
726 * omp-simd-clone.c (expand_simd_clones): If simd_clone_mangle or
727 simd_clone_create failed when i == 0, adjust clone->nargs by
730 2020-02-05 Martin Liska <mliska@suse.cz>
733 * doc/invoke.texi: Document that one should
734 not combine ASLR and -fpch.
736 2020-02-04 Richard Biener <rguenther@suse.de>
738 PR tree-optimization/93538
739 * match.pd (addr EQ/NE ptr): Amend to handle &ptr->x EQ/NE ptr.
741 2020-02-04 Richard Biener <rguenther@suse.de>
743 PR tree-optimization/91123
744 * tree-ssa-sccvn.c (vn_walk_cb_data::finish): New method.
745 (vn_walk_cb_data::last_vuse): New member.
746 (vn_walk_cb_data::saved_operands): Likewsie.
747 (vn_walk_cb_data::~vn_walk_cb_data): Release saved_operands.
748 (vn_walk_cb_data::push_partial_def): Use finish.
749 (vn_reference_lookup_2): Update last_vuse and use finish if
750 we've saved operands.
751 (vn_reference_lookup_3): Use finish and update calls to
752 push_partial_defs everywhere. When translating through
753 memcpy or aggregate copies save off operands and alias-set.
754 (eliminate_dom_walker::eliminate_stmt): Restore VN_WALKREWRITE
755 operation for redundant store removal.
757 2020-02-04 Richard Biener <rguenther@suse.de>
759 PR tree-optimization/92819
760 * tree-ssa-forwprop.c (simplify_vector_constructor): Avoid
761 generating more stmts than before.
763 2020-02-04 Martin Liska <mliska@suse.cz>
765 * config/arm/arm.c (arm_gen_far_branch): Move the function
766 outside of selftests.
768 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
770 * config/rs6000/rs6000.c (adjust_vec_address_pcrel): New helper
771 function to adjust PC-relative vector addresses.
772 (rs6000_adjust_vec_address): Call adjust_vec_address_pcrel to
773 handle vectors with PC-relative addresses.
775 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
777 * config/rs6000/rs6000.c (reg_to_non_prefixed): Add forward
779 (hard_reg_and_mode_to_addr_mask): Delete.
780 (rs6000_adjust_vec_address): If the original vector address
781 was REG+REG or REG+OFFSET and the element is not zero, do the add
782 of the elements in the original address before adding the offset
783 for the vector element. Use address_to_insn_form to validate the
784 address using the register being loaded, rather than guessing
785 whether the address is a DS-FORM or DQ-FORM address.
787 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
789 * config/rs6000/rs6000.c (get_vector_offset): New helper function
790 to calculate the offset in memory from the start of a vector of a
791 particular element. Add code to keep the element number in
792 bounds if the element number is variable.
793 (rs6000_adjust_vec_address): Move calculation of offset of the
794 vector element to get_vector_offset.
795 (rs6000_split_vec_extract_var): Do not do the initial AND of
796 element here, move the code to get_vector_offset.
798 2020-02-03 Michael Meissner <meissner@linux.ibm.com>
800 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add some
803 2020-02-03 Segher Boessenkool <segher@kernel.crashing.org>
805 * config/rs6000/constraints.md: Improve documentation.
807 2020-02-03 Richard Earnshaw <rearnsha@arm.com>
810 * config/arm/t-arm: ($(srcdir)/config/arm/arm-tune.md)
811 ($(srcdir)/config/arm/arm-tables.opt): Use move-if-change.
813 2020-02-03 Andrew Stubbs <ams@codesourcery.com>
815 * config.gcc: Remove "carrizo" support.
816 * config/gcn/gcn-opts.h (processor_type): Likewise.
817 * config/gcn/gcn.c (gcn_omp_device_kind_arch_isa): Likewise.
818 * config/gcn/gcn.opt (gpu_type): Likewise.
819 * config/gcn/t-omp-device: Likewise.
821 2020-02-03 Stam Markianos-Wright <stam.markianos-wright@arm.com>
824 * config/arm/arm-protos.h: New function arm_gen_far_branch prototype.
825 * config/arm/arm.c (arm_gen_far_branch): New function
827 * config/arm/arm.md: Update b<cond> for Thumb2 range checks.
829 2020-02-03 Julian Brown <julian@codesourcery.com>
830 Tobias Burnus <tobias@codesourcery.com>
832 * doc/invoke.texi: Update mention of OpenACC version to 2.6.
834 2020-02-03 Jakub Jelinek <jakub@redhat.com>
837 * config/s390/s390.md (popcounthi2_z196): Fix up expander to emit
838 valid RTL to sum up the lowest and second lowest bytes of the popcnt
841 2020-02-02 Vladimir Makarov <vmakarov@redhat.com>
843 PR rtl-optimization/91333
844 * ira-color.c (struct allocno_color_data): Add member
846 (init_allocno_threads): Set the member up.
847 (bucket_allocno_compare_func): Add compare hard reg
850 2020-01-31 Sandra Loosemore <sandra@codesourcery.com>
852 nios2: Support for GOT-relative DW_EH_PE_datarel encoding.
854 * configure.ac [nios2-*-*]: Check HAVE_AS_NIOS2_GOTOFF_RELOCATION.
855 * config.in: Regenerated.
856 * configure: Regenerated.
857 * config/nios2/nios2.h (ASM_PREFERRED_EH_DATA_FORMAT): Fix handling
858 for PIC when HAVE_AS_NIOS2_GOTOFF_RELOCATION.
859 (ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX): New.
861 2020-02-01 Andrew Burgess <andrew.burgess@embecosm.com>
863 * configure: Regenerate.
865 2020-01-31 Vladimir Makarov <vmakarov@redhat.com>
867 PR rtl-optimization/91333
868 * ira-color.c (bucket_allocno_compare_func): Move conflict hard
869 reg preferences comparison up.
871 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
873 * config/aarch64/aarch64.h (TARGET_SVE_BF16): New macro.
874 * config/aarch64/aarch64-sve-builtins-sve2.h (svcvtnt): Move to
875 aarch64-sve-builtins-base.h.
876 * config/aarch64/aarch64-sve-builtins-sve2.cc (svcvtnt): Move to
877 aarch64-sve-builtins-base.cc.
878 * config/aarch64/aarch64-sve-builtins-base.h (svbfdot, svbfdot_lane)
879 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
881 * config/aarch64/aarch64-sve-builtins-base.cc (svbfdot, svbfdot_lane)
882 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
883 (svcvtnt): New functions.
884 * config/aarch64/aarch64-sve-builtins-base.def (svbfdot, svbfdot_lane)
885 (svbfmlalb, svbfmlalb_lane, svbfmlalt, svbfmlalt_lane, svbfmmla)
886 (svcvtnt): New functions.
887 (svcvt): Add a form that converts f32 to bf16.
888 * config/aarch64/aarch64-sve-builtins-shapes.h (ternary_bfloat)
889 (ternary_bfloat_lane, ternary_bfloat_lanex2, ternary_bfloat_opt_n):
891 * config/aarch64/aarch64-sve-builtins-shapes.cc (parse_element_type):
892 Treat B as bfloat16_t.
893 (ternary_bfloat_lane_base): New class.
894 (ternary_bfloat_def): Likewise.
895 (ternary_bfloat): New shape.
896 (ternary_bfloat_lane_def): New class.
897 (ternary_bfloat_lane): New shape.
898 (ternary_bfloat_lanex2_def): New class.
899 (ternary_bfloat_lanex2): New shape.
900 (ternary_bfloat_opt_n_def): New class.
901 (ternary_bfloat_opt_n): New shape.
902 * config/aarch64/aarch64-sve-builtins.cc (TYPES_cvt_bfloat): New macro.
903 * config/aarch64/aarch64-sve.md (@aarch64_sve_<sve_fp_op>vnx4sf)
904 (@aarch64_sve_<sve_fp_op>_lanevnx4sf): New patterns.
905 (@aarch64_sve_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>)
906 (@cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
907 (*cond_<optab>_trunc<VNx4SF_ONLY:mode><VNx8BF_ONLY:mode>): Likewise.
908 (@aarch64_sve_cvtnt<VNx8BF_ONLY:mode>): Likewise.
909 * config/aarch64/aarch64-sve2.md (@aarch64_sve2_cvtnt<mode>): Key
910 the pattern off the narrow mode instead of the wider one.
911 * config/aarch64/iterators.md (VNx8BF_ONLY): New mode iterator.
912 (UNSPEC_BFMLALB, UNSPEC_BFMLALT, UNSPEC_BFMMLA): New unspecs.
913 (sve_fp_op): Handle them.
914 (SVE_BFLOAT_TERNARY_LONG): New int itertor.
915 (SVE_BFLOAT_TERNARY_LONG_LANE): Likewise.
917 2020-01-31 Richard Sandiford <richard.sandiford@arm.com>
919 * config/aarch64/arm_sve.h: Include arm_bf16.h.
920 * config/aarch64/aarch64-modes.def (BF): Move definition before
921 VECTOR_MODES. Remove separate VECTOR_MODES for V4BF and V8BF.
922 (SVE_MODES): Handle BF modes.
923 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
925 (aarch64_full_sve_mode): Likewise.
926 * config/aarch64/iterators.md (SVE_STRUCT): Add VNx16BF, VNx24BF
928 (SVE_FULL, SVE_FULL_HSD, SVE_ALL): Add VNx8BF.
929 (Vetype, Vesize, Vctype, VEL, Vel, VEL_INT, V128, v128, vwcore)
930 (V_INT_EQUIV, v_int_equiv, V_FP_EQUIV, v_fp_equiv, vector_count)
931 (insn_length, VSINGLE, vsingle, VPRED, vpred, VDOUBLE): Handle the
933 * config/aarch64/aarch64-sve-builtins.h (TYPE_bfloat): New
935 * config/aarch64/aarch64-sve-builtins.cc (TYPES_all_arith): New macro.
936 (TYPES_all_data): Add bf16.
937 (TYPES_reinterpret1, TYPES_reinterpret): Likewise.
938 (register_tuple_type): Increase buffer size.
939 * config/aarch64/aarch64-sve-builtins.def (svbfloat16_t): New type.
940 (bf16): New type suffix.
941 * config/aarch64/aarch64-sve-builtins-base.def (svabd, svadd, svaddv)
942 (svcmpeq, svcmpge, svcmpgt, svcmple, svcmplt, svcmpne, svmad, svmax)
943 (svmaxv, svmin, svminv, svmla, svmls, svmsb, svmul, svsub, svsubr):
944 Change type from all_data to all_arith.
945 * config/aarch64/aarch64-sve-builtins-sve2.def (svaddp, svmaxp)
948 2020-01-31 Dennis Zhang <dennis.zhang@arm.com>
949 Matthew Malcomson <matthew.malcomson@arm.com>
950 Richard Sandiford <richard.sandiford@arm.com>
952 * doc/invoke.texi (f32mm): Document new AArch64 -march= extension.
953 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Define
954 __ARM_FEATURE_SVE_MATMUL_INT8, __ARM_FEATURE_SVE_MATMUL_FP32 and
955 __ARM_FEATURE_SVE_MATMUL_FP64 as appropriate. Don't define
956 __ARM_FEATURE_MATMUL_FP64.
957 * config/aarch64/aarch64-option-extensions.def (fp, simd, fp16)
958 (sve): Add AARCH64_FL_F32MM to the list of extensions that should
959 be disabled at the same time.
960 (f32mm): New extension.
961 * config/aarch64/aarch64.h (AARCH64_FL_F32MM): New macro.
962 (AARCH64_FL_F64MM): Bump to the next bit up.
963 (AARCH64_ISA_F32MM, TARGET_SVE_I8MM, TARGET_F32MM, TARGET_SVE_F32MM)
964 (TARGET_SVE_F64MM): New macros.
965 * config/aarch64/iterators.md (SVE_MATMULF): New mode iterator.
966 (UNSPEC_FMMLA, UNSPEC_SMATMUL, UNSPEC_UMATMUL, UNSPEC_USMATMUL)
967 (UNSPEC_TRN1Q, UNSPEC_TRN2Q, UNSPEC_UZP1Q, UNSPEC_UZP2Q, UNSPEC_ZIP1Q)
968 (UNSPEC_ZIP2Q): New unspeccs.
969 (DOTPROD_US_ONLY, PERMUTEQ, MATMUL, FMMLA): New int iterators.
970 (optab, sur, perm_insn): Handle the new unspecs.
971 (sve_fp_op): Handle UNSPEC_FMMLA. Resort.
972 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use
973 TARGET_SVE_F64MM instead of separate tests.
974 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod<vsi2qi>): New pattern.
975 (@aarch64_<DOTPROD_US_ONLY:sur>dot_prod_lane<vsi2qi>): Likewise.
976 (@aarch64_sve_add_<MATMUL:optab><vsi2qi>): Likewise.
977 (@aarch64_sve_<FMMLA:sve_fp_op><mode>): Likewise.
978 (@aarch64_sve_<PERMUTEQ:optab><mode>): Likewise.
979 * config/aarch64/aarch64-sve-builtins.cc (TYPES_s_float): New macro.
980 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): Use it.
981 (TYPES_s_signed): New macro.
982 (TYPES_s_integer): Use it.
983 (TYPES_d_float): New macro.
984 (TYPES_d_data): Use it.
985 * config/aarch64/aarch64-sve-builtins-shapes.h (mmla): Declare.
986 (ternary_intq_uintq_lane, ternary_intq_uintq_opt_n, ternary_uintq_intq)
987 (ternary_uintq_intq_lane, ternary_uintq_intq_opt_n): Likewise.
988 * config/aarch64/aarch64-sve-builtins-shapes.cc (mmla_def): New class.
990 (ternary_resize2_opt_n_base): Add TYPE_CLASS2 and TYPE_CLASS3
992 (ternary_resize2_lane_base): Likewise.
993 (ternary_resize2_base): New class.
994 (ternary_qq_lane_base): Likewise.
995 (ternary_intq_uintq_lane_def): Likewise.
996 (ternary_intq_uintq_lane): New shape.
997 (ternary_intq_uintq_opt_n_def): New class
998 (ternary_intq_uintq_opt_n): New shape.
999 (ternary_qq_lane_def): Inherit from ternary_qq_lane_base.
1000 (ternary_uintq_intq_def): New class.
1001 (ternary_uintq_intq): New shape.
1002 (ternary_uintq_intq_lane_def): New class.
1003 (ternary_uintq_intq_lane): New shape.
1004 (ternary_uintq_intq_opt_n_def): New class.
1005 (ternary_uintq_intq_opt_n): New shape.
1006 * config/aarch64/aarch64-sve-builtins-base.h (svmmla, svsudot)
1007 (svsudot_lane, svtrn1q, svtrn2q, svusdot, svusdot_lane, svusmmla)
1008 (svuzp1q, svuzp2q, svzip1q, svzip2q): Declare.
1009 * config/aarch64/aarch64-sve-builtins-base.cc (svdot_lane_impl):
1011 (svdotprod_lane_impl): ...this new class.
1012 (svmmla_impl, svusdot_impl): New classes.
1013 (svdot_lane): Update to use svdotprod_lane_impl.
1014 (svmmla, svsudot, svsudot_lane, svtrn1q, svtrn2q, svusdot)
1015 (svusdot_lane, svusmmla, svuzp1q, svuzp2q, svzip1q, svzip2q): New
1017 * config/aarch64/aarch64-sve-builtins-base.def (svmmla): New base
1018 function, with no types defined.
1019 (svmmla, svusmmla, svsudot, svsudot_lane, svusdot, svusdot_lane): New
1020 AARCH64_FL_I8MM functions.
1021 (svmmla): New AARCH64_FL_F32MM function.
1022 (svld1ro): Depend only on AARCH64_FL_F64MM, not on AARCH64_FL_V8_6.
1023 (svmmla, svtrn1q, svtrn2q, svuz1q, svuz2q, svzip1q, svzip2q): New
1024 AARCH64_FL_F64MM function.
1025 (REQUIRED_EXTENSIONS):
1027 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1029 * config/gcn/gcn-valu.md (addv64di3_exec): Allow one '0' in each
1032 2020-01-31 Uroš Bizjak <ubizjak@gmail.com>
1034 * config/i386/i386.md (*movoi_internal_avx): Do not check for
1035 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL. Remove MODE_V8SF handling.
1036 (*movti_internal): Do not check for
1037 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1038 (*movtf_internal): Move check for TARGET_SSE2 and size optimization
1039 just after check for TARGET_AVX.
1040 (*movdf_internal): Ditto.
1041 * config/i386/mmx.md (*mov<mode>_internal): Do not check for
1042 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL.
1043 * config/i386/sse.md (mov<mode>_internal): Only check
1044 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL with V2DFmode. Move check
1045 for TARGET_SSE2 and size optimization just after check for TARGET_AVX.
1046 (<sse>_andnot<mode>3<mask_name>): Move check for
1047 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL after check for TARGET_AVX.
1048 (<code><mode>3<mask_name>): Ditto.
1049 (*andnot<mode>3): Ditto.
1050 (*andnottf3): Ditto.
1051 (*<code><mode>3): Ditto.
1052 (*<code>tf3): Ditto.
1053 (*andnot<VI:mode>3): Remove
1054 TARGET_SSE_PACKED_SINGLE_INSN_OPTIMAL handling.
1055 (<mask_codefor><code><VI48_AVX_AVX512F:mode>3<mask_name>): Ditto.
1056 (*<code><VI12_AVX_AVX512F:mode>3): Ditto.
1057 (sse4_1_blendv<ssemodesuffix>): Ditto.
1058 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL):
1059 Explain that tune applies to 128bit instructions only.
1061 2020-01-31 Kwok Cheung Yeung <kcy@codesourcery.com>
1063 * config/gcn/mkoffload.c (process_asm): Add sgpr_count and vgpr_count
1064 to definition of hsa_kernel_description. Parse assembly to find SGPR
1065 and VGPR count of kernel and store in hsa_kernel_description.
1067 2020-01-31 Tamar Christina <tamar.christina@arm.com>
1069 PR rtl-optimization/91838
1070 * simplify-rtx.c (simplify_binary_operation_1): Update LSHIFTRT case
1071 to truncate if allowed or reject combination.
1073 2020-01-31 Andrew Stubbs <ams@codesourcery.com>
1075 * tree-ssa-loop-ivopts.c (get_iv): Use sizetype for zero-step.
1076 (find_inv_vars_cb): Likewise.
1078 2020-01-31 David Malcolm <dmalcolm@redhat.com>
1080 * calls.c (special_function_p): Split out the check for DECL_NAME
1081 being non-NULL and fndecl being extern at file scope into a
1082 new maybe_special_function_p and call it. Drop check for fndecl
1083 being non-NULL that was after a usage of DECL_NAME (fndecl).
1084 * tree.h (maybe_special_function_p): New inline function.
1086 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1088 * config/gcn/gcn-valu.md (gather<mode>_exec): Move contents ...
1089 (mask_gather_load<mode>): ... here, and zero-initialize the
1091 (maskload<mode>di): Zero-initialize the destination.
1094 2020-01-30 David Malcolm <dmalcolm@redhat.com>
1097 * doc/analyzer.texi (Limitations): Note that constraints on
1098 floating-point values are currently ignored.
1100 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1103 * symtab.c (symtab_node::noninterposable_alias): If localalias
1104 already exists, but is not usable, append numbers after it until
1105 a unique name is found. Formatting fix.
1108 * combine.c (simplify_comparison) <case ROTATE>: Punt on out of range
1111 2020-01-30 Andrew Stubbs <ams@codesourcery.com>
1113 * config/gcn/gcn.c (print_operand): Handle LTGT.
1114 * config/gcn/predicates.md (gcn_fp_compare_operator): Allow ltgt.
1116 2020-01-30 Richard Biener <rguenther@suse.de>
1118 * tree-pretty-print.c (dump_generic_node): Wrap VECTOR_CST
1119 and CONSTRUCTOR in _Literal (type) with TDF_GIMPLE.
1121 2020-01-30 John David Anglin <danglin@gcc.gnu.org>
1123 * config/pa/pa.c (pa_elf_select_rtx_section): Place function pointers
1124 without a DECL in .data.rel.ro.local.
1126 2020-01-30 Jakub Jelinek <jakub@redhat.com>
1129 * config/arm/arm.md (uaddvdi4): Actually emit what gen_uaddvsi4
1133 * config/i386/sse.md
1134 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext): Renamed to ...
1135 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext): ... this. Use
1136 any_extend code iterator instead of always zero_extend.
1137 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_lt): Renamed to ...
1138 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_lt): ... this.
1139 Use any_extend code iterator instead of always zero_extend.
1140 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_zext_shift): Renamed to ...
1141 (*<sse>_movmsk<ssemodesuffix><avxsizesuffix>_<u>ext_shift): ... this.
1142 Use any_extend code iterator instead of always zero_extend.
1143 (*sse2_pmovmskb_ext): New define_insn.
1144 (*sse2_pmovmskb_ext_lt): New define_insn_and_split.
1147 * config/i386/i386.md (*popcountsi2_zext): New define_insn_and_split.
1148 (*popcountsi2_zext_falsedep): New define_insn.
1150 2020-01-30 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1152 * config.in: Regenerated.
1153 * configure: Regenerated.
1155 2020-01-29 Tobias Burnus <tobias@codesourcery.com>
1158 * config/gcn/gcn-hsa.h (ASM_SPEC): Add -mattr=-code-object-v3 as
1159 LLVM's assembler changed the default in version 9.
1161 2020-01-24 Jeff Law <law@redhat.com>
1163 PR tree-optimization/89689
1164 * builtins.def (BUILT_IN_OBJECT_SIZE): Make it const rather than pure.
1166 2020-01-29 Richard Sandiford <richard.sandiford@arm.com>
1170 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1172 PR rtl-optimization/87763
1173 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1174 simplification to handle subregs as well as bare regs.
1175 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1177 2020-01-29 Joel Hutton <Joel.Hutton@arm.com>
1180 * ira.c (ira): Revert use of simplified LRA algorithm.
1182 2020-01-29 Martin Jambor <mjambor@suse.cz>
1184 PR tree-optimization/92706
1185 * tree-sra.c (struct access): Fields first_link, last_link,
1186 next_queued and grp_queued renamed to first_rhs_link, last_rhs_link,
1187 next_rhs_queued and grp_rhs_queued respectively, new fields
1188 first_lhs_link, last_lhs_link, next_lhs_queued and grp_lhs_queued.
1189 (struct assign_link): Field next renamed to next_rhs, new field
1190 next_lhs. Updated comment.
1191 (work_queue_head): Renamed to rhs_work_queue_head.
1192 (lhs_work_queue_head): New variable.
1193 (add_link_to_lhs): New function.
1194 (relink_to_new_repr): Also relink LHS lists.
1195 (add_access_to_work_queue): Renamed to add_access_to_rhs_work_queue.
1196 (add_access_to_lhs_work_queue): New function.
1197 (pop_access_from_work_queue): Renamed to
1198 pop_access_from_rhs_work_queue.
1199 (pop_access_from_lhs_work_queue): New function.
1200 (build_accesses_from_assign): Also add links to LHS lists and to LHS
1202 (child_would_conflict_in_lacc): Renamed to
1203 child_would_conflict_in_acc. Adjusted parameter names.
1204 (create_artificial_child_access): New parameter set_grp_read, use it.
1205 (subtree_mark_written_and_enqueue): Renamed to
1206 subtree_mark_written_and_rhs_enqueue.
1207 (propagate_subaccesses_across_link): Renamed to
1208 propagate_subaccesses_from_rhs.
1209 (propagate_subaccesses_from_lhs): New function.
1210 (propagate_all_subaccesses): Also propagate subaccesses from LHSs to
1213 2020-01-29 Martin Jambor <mjambor@suse.cz>
1215 PR tree-optimization/92706
1216 * tree-sra.c (struct access): Adjust comment of
1217 grp_total_scalarization.
1218 (find_access_in_subtree): Look for single children spanning an entire
1220 (scalarizable_type_p): Allow register accesses, adjust callers.
1221 (completely_scalarize): Remove function.
1222 (scalarize_elem): Likewise.
1223 (create_total_scalarization_access): Likewise.
1224 (sort_and_splice_var_accesses): Do not track total scalarization
1226 (analyze_access_subtree): New parameter totally, adjust to new meaning
1227 of grp_total_scalarization.
1228 (analyze_access_trees): Pass new parameter to analyze_access_subtree.
1229 (can_totally_scalarize_forest_p): New function.
1230 (create_total_scalarization_access): Likewise.
1231 (create_total_access_and_reshape): Likewise.
1232 (total_should_skip_creating_access): Likewise.
1233 (totally_scalarize_subtree): Likewise.
1234 (analyze_all_variable_accesses): Perform total scalarization after
1235 subaccess propagation using the new functions above.
1236 (initialize_constant_pool_replacements): Output initializers by
1237 traversing the access tree.
1239 2020-01-29 Martin Jambor <mjambor@suse.cz>
1241 * tree-sra.c (verify_sra_access_forest): New function.
1242 (verify_all_sra_access_forests): Likewise.
1243 (create_artificial_child_access): Set parent.
1244 (analyze_all_variable_accesses): Call the verifier.
1246 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1248 * cgraph.c (cgraph_edge::resolve_speculation): Only lookup direct edge
1249 if called on indirect edge.
1250 (cgraph_edge::redirect_call_stmt_to_callee): Lookup indirect edge of
1251 speculative call if needed.
1253 2020-01-29 Richard Biener <rguenther@suse.de>
1255 PR tree-optimization/93428
1256 * tree-vect-slp.c (vect_build_slp_tree_2): Compute the load
1257 permutation when the load node is created.
1258 (vect_analyze_slp_instance): Re-use it here.
1260 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1262 * ipa-prop.c (update_indirect_edges_after_inlining): Fix warning.
1264 2020-01-28 Vladimir Makarov <vmakarov@redhat.com>
1266 PR rtl-optimization/93272
1267 * ira-lives.c (process_out_of_region_eh_regs): New function.
1268 (process_bb_node_lives): Call it.
1270 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1272 * coverage.c (read_counts_file): Make error message lowercase.
1274 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1276 * profile-count.c (profile_quality_display_names): Fix ordering.
1278 2020-01-28 Jan Hubicka <hubicka@ucw.cz>
1281 * cgraph.c (cgraph_add_edge_to_call_site_hash): Update call site
1282 hash only when edge is first within the sequence.
1283 (cgraph_edge::set_call_stmt): Update handling of speculative calls.
1284 (symbol_table::create_edge): Do not set target_prob.
1285 (cgraph_edge::remove_caller): Watch for speculative calls when updating
1287 (cgraph_edge::make_speculative): Drop target_prob parameter.
1288 (cgraph_edge::speculative_call_info): Remove.
1289 (cgraph_edge::first_speculative_call_target): New member function.
1290 (update_call_stmt_hash_for_removing_direct_edge): New function.
1291 (cgraph_edge::resolve_speculation): Rewrite to new API.
1292 (cgraph_edge::speculative_call_for_target): New member function.
1293 (cgraph_edge::make_direct): Rewrite to new API; fix handling of
1294 multiple speculation targets.
1295 (cgraph_edge::redirect_call_stmt_to_callee): Likewise; fix updating
1297 (verify_speculative_call): Verify that targets form an interval.
1298 * cgraph.h (cgraph_edge::speculative_call_info): Remove.
1299 (cgraph_edge::first_speculative_call_target): New member function.
1300 (cgraph_edge::next_speculative_call_target): New member function.
1301 (cgraph_edge::speculative_call_target_ref): New member function.
1302 (cgraph_edge;:speculative_call_indirect_edge): New member funtion.
1303 (cgraph_edge): Remove target_prob.
1304 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
1305 Fix handling of speculative calls.
1306 * ipa-devirt.c (ipa_devirt): Fix handling of speculative cals.
1307 * ipa-fnsummary.c (analyze_function_body): Likewise.
1308 * ipa-inline.c (speculation_useful_p): Use new speculative call API.
1309 * ipa-profile.c (dump_histogram): Fix formating.
1310 (ipa_profile_generate_summary): Watch for overflows.
1311 (ipa_profile): Do not require probablity to be 1/2; update to new API.
1312 * ipa-prop.c (ipa_make_edge_direct_to_target): Update to new API.
1313 (update_indirect_edges_after_inlining): Update to new API.
1314 * ipa-utils.c (ipa_merge_profiles): Rewrite merging of speculative call
1316 * profile-count.h: (profile_probability::adjusted): New.
1317 * tree-inline.c (copy_bb): Update to new speculative call API; fix
1318 updating of profile.
1319 * value-prof.c (gimple_ic_transform): Rename to ...
1320 (dump_ic_profile): ... this one; update dumping.
1321 (stream_in_histogram_value): Fix formating.
1322 (gimple_value_profile_transformations): Update.
1324 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1327 * config/i386/i386.md (*movoi_internal_avx): Remove
1328 TARGET_SSE_TYPELESS_STORES check.
1329 (*movti_internal): Prefer TARGET_AVX over
1330 TARGET_SSE_TYPELESS_STORES.
1331 (*movtf_internal): Likewise.
1332 * config/i386/sse.md (mov<mode>_internal): Prefer TARGET_AVX over
1333 TARGET_SSE_TYPELESS_STORES. Remove "<MODE_SIZE> == 16" check
1334 from TARGET_SSE_TYPELESS_STORES.
1336 2020-01-28 David Malcolm <dmalcolm@redhat.com>
1338 * diagnostic-core.h (warning_at): Rename overload to...
1339 (warning_meta): ...this.
1340 (emit_diagnostic_valist): Delete decl of overload taking
1341 diagnostic_metadata.
1342 * diagnostic.c (emit_diagnostic_valist): Likewise for defn.
1343 (warning_at): Rename overload taking diagnostic_metadata to...
1344 (warning_meta): ...this.
1346 2020-01-28 Richard Biener <rguenther@suse.de>
1348 PR tree-optimization/93439
1349 * tree-parloops.c (create_loop_fn): Move clique bookkeeping...
1350 * tree-cfg.c (move_sese_region_to_fn): ... here.
1351 (verify_types_in_gimple_reference): Verify used cliques are
1354 2020-01-28 H.J. Lu <hongjiu.lu@intel.com>
1357 * config/i386/i386-options.c (set_ix86_tune_features): Add an
1358 argument of a pointer to struct gcc_options and pass it to
1359 parse_mtune_ctrl_str.
1360 (ix86_function_specific_restore): Pass opts to
1361 set_ix86_tune_features.
1362 (ix86_option_override_internal): Likewise.
1363 (parse_mtune_ctrl_str): Add an argument of a pointer to struct
1364 gcc_options and use it for x_ix86_tune_ctrl_string.
1366 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1368 PR rtl-optimization/87763
1369 * simplify-rtx.c (simplify_truncation): Extend sign/zero_extract
1370 simplification to handle subregs as well as bare regs.
1371 * config/i386/i386.md (*testqi_ext_3): Match QI extracts too.
1373 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1375 * tree-vect-loop.c (vectorizable_reduction): Fail gracefully
1376 for reduction chains that (now) include a call.
1378 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1380 PR tree-optimization/92822
1381 * tree-ssa-forwprop.c (simplify_vector_constructor): When filling
1382 out the don't-care elements of a vector whose significant elements
1383 are duplicates, make the don't-care elements duplicates too.
1385 2020-01-28 Richard Sandiford <richard.sandiford@arm.com>
1387 PR tree-optimization/93434
1388 * tree-predcom.c (split_data_refs_to_components): Record which
1389 components have had aliasing loads removed. Prevent store-store
1390 commoning for all such components.
1392 2020-01-28 Jakub Jelinek <jakub@redhat.com>
1395 * config/i386/i386.c (ix86_fold_builtin) <do_shift>: If mask is not
1396 -1 or is_vshift is true, use new_vector with number of elts npatterns
1397 rather than new_unary_operation.
1399 PR tree-optimization/93454
1400 * gimple-fold.c (fold_array_ctor_reference): Perform
1401 elt_size.to_uhwi () just once, instead of calling it in every
1402 iteration. Punt if that value is above size of the temporary
1403 buffer. Decrease third native_encode_expr argument when
1404 bufoff + elt_sz is above size of buf.
1406 2020-01-27 Joseph Myers <joseph@codesourcery.com>
1408 * config/mips/mips.c (mips_declare_object_name)
1409 [USE_GNU_UNIQUE_OBJECT]: Support use of gnu_unique_object.
1411 2020-01-27 Martin Liska <mliska@suse.cz>
1413 PR gcov-profile/93403
1414 * tree-profile.c (gimple_init_gcov_profiler): Generate
1415 both __gcov_indirect_call_profiler_v4 and
1416 __gcov_indirect_call_profiler_v4_atomic.
1418 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1421 * config/aarch64/aarch64-simd.md (aarch64_get_half<mode>): New
1423 (@aarch64_split_simd_mov<mode>): Use it.
1424 (aarch64_simd_mov_from_<mode>low): Add a GPR alternative.
1425 Leave the vec_extract patterns to handle 2-element vectors.
1426 (aarch64_simd_mov_from_<mode>high): Likewise.
1427 (vec_extract<VQMOV_NO2E:mode><Vhalf>): New expander.
1428 (vec_extractv2dfv1df): Likewise.
1430 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1432 * config/aarch64/aarch64.c (aarch64_if_then_else_costs): Match
1433 jump conditions for *compare_condjump<GPI:mode>.
1435 2020-01-27 David Malcolm <dmalcolm@redhat.com>
1438 * digraph.cc (test_edge::test_edge): Specify template for base
1441 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1443 * config/arc/arc.c (arc_rtx_costs): Update mul64 cost.
1445 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1447 * config/arc/arc-protos.h (gen_mlo): Remove.
1448 (gen_mhi): Likewise.
1449 * config/arc/arc.c (AUX_MULHI): Define.
1450 (arc_must_save_reister): Special handling for r58/59.
1451 (arc_compute_frame_size): Consider mlo/mhi registers.
1452 (arc_save_callee_saves): Emit fp/sp move only when emit_move
1454 (arc_conditional_register_usage): Remove TARGET_BIG_ENDIAN from
1455 mlo/mhi name selection.
1456 (arc_restore_callee_saves): Don't early restore blink when ISR.
1457 (arc_expand_prologue): Add mlo/mhi saving.
1458 (arc_expand_epilogue): Add mlo/mhi restoring.
1461 * config/arc/arc.h (DBX_REGISTER_NUMBER): Correct register
1462 numbering when MUL64 option is used.
1463 (DWARF2_FRAME_REG_OUT): Define.
1464 * config/arc/arc.md (arc600_stall): New pattern.
1465 (VUNSPEC_ARC_ARC600_STALL): Define.
1466 (mulsi64): Use correct mlo/mhi registers.
1467 (mulsi_600): Clean it up.
1468 * config/arc/predicates.md (mlo_operand): Remove any dependency on
1470 (mhi_operand): Likewise.
1472 2020-01-27 Claudiu Zissulescu <claziss@synopsys.com>
1473 Petro Karashchenko <petro.karashchenko@ring.com>
1475 * config/arc/arc.c (arc_is_uncached_mem_p): Check struct
1476 attributes if needed.
1477 (prepare_move_operands): Generate special unspec instruction for
1479 (arc_isuncached_mem_p): Propagate uncached attribute to each
1481 * config/arc/arc.md (VUNSPEC_ARC_LDDI): Define.
1482 (VUNSPEC_ARC_STDI): Likewise.
1483 (ALLI): New mode iterator.
1484 (mALLI): New mode attribute.
1485 (lddi): New instruction pattern.
1487 (stdidi_split): Split instruction for architectures which are not
1488 supporting ll64 option.
1489 (lddidi_split): Likewise.
1491 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1493 PR rtl-optimization/92989
1494 * lra-lives.c (process_bb_lives): Update the live-in set before
1495 processing additional clobbers.
1497 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1499 PR rtl-optimization/93170
1500 * cselib.c (cselib_invalidate_regno_val): New function, split out
1502 (cselib_invalidate_regno): ...here.
1503 (cselib_invalidated_by_call_p): New function.
1504 (cselib_process_insn): Iterate over all the hard-register entries in
1505 REG_VALUES and invalidate any that cross call-clobbered registers.
1507 2020-01-27 Richard Sandiford <richard.sandiford@arm.com>
1509 * dojump.c (split_comparison): Use HONOR_NANS rather than
1510 HONOR_SNANS when splitting LTGT.
1512 2020-01-27 Martin Liska <mliska@suse.cz>
1515 * opts.c (print_filtered_help): Exclude language-specific
1516 options from --help=common unless enabled in all FEs.
1518 2020-01-27 Martin Liska <mliska@suse.cz>
1520 * opts.c (print_help): Exclude params from
1521 all except --help=param.
1523 2020-01-27 Martin Liska <mliska@suse.cz>
1526 * config/i386/i386-features.c (make_resolver_func):
1527 Align the code with ppc64 target implementation.
1528 Do not generate a unique name for resolver function.
1530 2020-01-27 Richard Biener <rguenther@suse.de>
1532 PR tree-optimization/93397
1533 * tree-vect-slp.c (vect_analyze_slp_instance): Delay
1534 converted reduction chain SLP graph adjustment.
1536 2020-01-26 Marek Polacek <polacek@redhat.com>
1539 * sanopt.c (sanitize_rewrite_addressable_params): Avoid crash on
1542 2020-01-26 Jason Merrill <jason@redhat.com>
1545 * tree.c (verify_type_variant): Only verify TYPE_NEEDS_CONSTRUCTING
1548 2020-01-26 Darius Galis <darius.galis@cyberthorstudios.com>
1550 * config/rx/rx.md (setmemsi): Added rx_allow_string_insns constraint
1551 (rx_setmem): Likewise.
1553 2020-01-26 Jakub Jelinek <jakub@redhat.com>
1556 * config/i386/i386.md (*addv<dwi>4_doubleword, *subv<dwi>4_doubleword):
1557 Use nonimmediate_operand instead of x86_64_hilo_general_operand and
1558 drop <di> from constraint of last operand.
1561 * config/i386/sse.md (*avx_vperm_broadcast_<mode>): Disallow for
1562 TARGET_AVX2 and V4DFmode not in the split condition, but in the
1563 pattern condition, though allow { 0, 0, 0, 0 } broadcast always.
1565 2020-01-25 Feng Xue <fxue@os.amperecomputing.com>
1568 * ipa-cp.c (get_info_about_necessary_edges): Remove value
1571 2020-01-24 Jeff Law <law@redhat.com>
1573 PR tree-optimization/92788
1574 * tree-ssa-threadedge.c (thread_across_edge): Check EDGE_COMPLEX
1577 2020-01-24 Jakub Jelinek <jakub@redhat.com>
1580 * config/i386/sse.md (*avx_vperm_broadcast_v4sf,
1581 *avx_vperm_broadcast_<mode>,
1582 <sse2_avx_avx512f>_vpermil<mode><mask_name>,
1583 *<sse2_avx_avx512f>_vpermilp<mode><mask_name>):
1584 Move before avx2_perm<mode>/avx512f_perm<mode>.
1587 * simplify-rtx.c (simplify_const_unary_operation,
1588 simplify_const_binary_operation): Punt for mode precision above
1589 MAX_BITSIZE_MODE_ANY_INT.
1591 2020-01-24 Andrew Pinski <apinski@marvell.com>
1593 * config/arm/aarch-cost-tables.h (cortexa57_extra_costs): Change
1596 2020-01-24 Jeff Law <law@redhat.com>
1599 * config/h8300/h8300.c (h8300_print_operand): Only call byte_reg
1600 for REGs. Call output_operand_lossage to get more reasonable
1603 2020-01-24 Andrew Stubbs <ams@codesourcery.com>
1605 * config/gcn/gcn-valu.md (vec_cmp<mode>di): Use
1606 gcn_fp_compare_operator.
1607 (vec_cmpu<mode>di): Use gcn_compare_operator.
1608 (vec_cmp<u>v64qidi): Use gcn_compare_operator.
1609 (vec_cmp<mode>di_exec): Use gcn_fp_compare_operator.
1610 (vec_cmpu<mode>di_exec): Use gcn_compare_operator.
1611 (vec_cmp<u>v64qidi_exec): Use gcn_compare_operator.
1612 (vec_cmp<mode>di_dup): Use gcn_fp_compare_operator.
1613 (vec_cmp<mode>di_dup_exec): Use gcn_fp_compare_operator.
1614 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): Use
1615 gcn_fp_compare_operator.
1616 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): Use
1617 gcn_fp_compare_operator.
1618 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): Use
1619 gcn_fp_compare_operator.
1620 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): Use
1621 gcn_fp_compare_operator.
1623 2020-01-24 Maciej W. Rozycki <macro@wdc.com>
1625 * doc/install.texi (Cross-Compiler-Specific Options): Document
1626 `--with-toolexeclibdir' option.
1628 2020-01-24 Hans-Peter Nilsson <hp@axis.com>
1630 * target.def (flags_regnum): Also mention effect on delay slot filling.
1631 * doc/tm.texi: Regenerate.
1633 2020-01-23 Jeff Law <law@redhat.com>
1635 PR translation/90162
1636 * config/h8300/h8300.c (h8300_option_override): Fix diagnostic text.
1638 2020-01-23 Mikael Tillenius <mti-1@tillenius.com>
1641 * config/h8300/h8300.h (FUNCTION_PROFILER): Fix emission of
1644 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1646 PR rtl-optimization/93402
1647 * postreload.c (reload_combine_recognize_pattern): Don't try to adjust
1650 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1652 * config.in: Regenerated.
1653 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to 1
1654 for TARGET_LIBC_GNUSTACK.
1655 * configure: Regenerated.
1656 * configure.ac: Define TARGET_LIBC_GNUSTACK if glibc version is
1657 found to be 2.31 or greater.
1659 2020-01-23 Dragan Mladjenovic <dmladjenovic@wavecomp.com>
1661 * config/mips/linux.h (NEED_INDICATE_EXEC_STACK): Define to
1663 * config/mips/mips.c (TARGET_ASM_FILE_END): Define to ...
1664 (mips_asm_file_end): New function. Delegate to
1665 file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK is true.
1666 * config/mips/mips.h (NEED_INDICATE_EXEC_STACK): Define to 0.
1668 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1671 * config/i386/i386-modes.def (POImode): New mode.
1672 (MAX_BITSIZE_MODE_ANY_INT): Change from 128 to 160.
1673 * config/i386/i386.md (DPWI): New mode attribute.
1674 (addv<mode>4, subv<mode>4): Use <DPWI> instead of <DWI>.
1676 (QPWI): ... this. Use POI instead of OI for TImode.
1677 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1,
1678 *subv<dwi>4_doubleword, *subv<dwi>4_doubleword_1): Use <QPWI>
1681 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1684 * config/aarch64/aarch64.md (UNSPEC_SPECULATION_TRACKER_REV): New
1686 (speculation_tracker_rev): New pattern.
1687 * config/aarch64/aarch64-speculation.cc (aarch64_do_track_speculation):
1688 Use speculation_tracker_rev to track the inverse condition.
1690 2020-01-23 Richard Biener <rguenther@suse.de>
1692 PR tree-optimization/93381
1693 * tree-ssa-sccvn.c (vn_walk_cb_data::push_partial_def): Take
1694 alias-set of the def as argument and record the first one.
1695 (vn_walk_cb_data::first_set): New member.
1696 (vn_reference_lookup_3): Pass the alias-set of the current def
1697 to push_partial_def. Fix alias-set used in the aggregate copy
1699 (vn_reference_lookup): Consistently set *last_vuse_ptr.
1700 * real.c (clear_significand_below): Fix out-of-bound access.
1702 2020-01-23 Jakub Jelinek <jakub@redhat.com>
1705 * config/i386/i386.md (*bmi2_bzhi_<mode>3_2, *bmi2_bzhi_<mode>3_3):
1706 New define_insn patterns.
1708 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1710 * doc/sourcebuild.texi (check-function-bodies): Add an
1711 optional target/xfail selector.
1713 2020-01-23 Richard Sandiford <richard.sandiford@arm.com>
1715 PR rtl-optimization/93124
1716 * auto-inc-dec.c (merge_in_block): Don't add auto inc/decs to
1717 bare USE and CLOBBER insns.
1719 2020-01-22 Andrew Pinski <apinski@marvell.com>
1721 * config/arc/arc.c (output_short_suffix): Check insn for nullness.
1723 2020-01-22 David Malcolm <dmalcolm@redhat.com>
1726 * gdbinit.in (break-on-saved-diagnostic): Update for move of
1727 diagnostic_manager into "ana" namespace.
1728 * selftest-run-tests.c (selftest::run_tests): Update for move of
1729 selftest::run_analyzer_selftests to
1730 ana::selftest::run_analyzer_selftests.
1732 2020-01-22 Richard Sandiford <richard.sandiford@arm.com>
1734 * cfgexpand.c (union_stack_vars): Update the size.
1736 2020-01-22 Richard Biener <rguenther@suse.de>
1738 PR tree-optimization/93381
1739 * tree-ssa-structalias.c (find_func_aliases): Assume offsetting
1740 throughout, handle all conversions the same.
1742 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1745 * config/aarch64/aarch64.c (aarch64_expand_subvti): Only use
1746 gen_subdi3_compare1_imm if low_in2 satisfies aarch64_plus_immediate
1747 predicate, not whenever it is CONST_INT. Otherwise, force_reg it.
1748 Call force_reg on high_in2 unconditionally.
1750 2020-01-22 Martin Liska <mliska@suse.cz>
1752 PR tree-optimization/92924
1753 * profile.c (compute_value_histograms): Divide
1756 2020-01-22 Jakub Jelinek <jakub@redhat.com>
1759 * output.h (assemble_name_resolve): Declare.
1760 * varasm.c (assemble_name_resolve): New function.
1761 (assemble_name): Use it.
1762 * config/i386/i386.h (ASM_OUTPUT_SYMBOL_REF): Define.
1764 2020-01-22 Joseph Myers <joseph@codesourcery.com>
1766 * doc/sourcebuild.texi (Texinfo Manuals, Front End): Refer to
1767 update_web_docs_git instead of update_web_docs_svn.
1769 2020-01-21 Andrew Pinski <apinski@marvell.com>
1772 * config/aarch64/aarch64.md (tlsgd_small_<mode>): Have operand 0
1773 as PTR mode. Have operand 1 as being modeless, it can be P mode.
1774 (*tlsgd_small_<mode>): Likewise.
1775 * config/aarch64/aarch64.c (aarch64_load_symref_appropriately)
1776 <case SYMBOL_SMALL_TLSGD>: Call gen_tlsgd_small_* with a ptr_mode
1777 register. Convert that register back to dest using convert_mode.
1779 2020-01-21 Jim Wilson <jimw@sifive.com>
1781 * config/riscv/riscv-sr.c (riscv_sr_match_prologue): Use INTVAL
1784 2020-01-21 H.J. Lu <hongjiu.lu@intel.com>
1785 Uros Bizjak <ubizjak@gmail.com>
1788 * config/i386/i386.c (ix86_tls_module_base): Replace Pmode
1790 (legitimize_tls_address): Do GNU2 TLS address computation in
1791 ptr_mode and zero-extend result to Pmode.
1792 * config/i386/i386.md (@tls_dynamic_gnu2_64_<mode>): Replace
1793 :P with :PTR and Pmode with ptr_mode.
1794 (*tls_dynamic_gnu2_lea_64_<mode>): Likewise.
1795 (*tls_dynamic_gnu2_call_64_<mode>): Likewise.
1796 (*tls_dynamic_gnu2_combine_64_<mode>): Likewise.
1798 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1801 * config/riscv/riscv.c (riscv_rtx_costs) <case ZERO_EXTRACT>: Verify
1802 the last two operands are CONST_INT_P before using them as such.
1804 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1806 * config/aarch64/aarch64-sve-builtins.def: Use get_typenode_from_name
1807 to get the integer element types.
1809 2020-01-21 Richard Sandiford <richard.sandiford@arm.com>
1811 * config/aarch64/aarch64-sve-builtins.h
1812 (function_expander::convert_to_pmode): Declare.
1813 * config/aarch64/aarch64-sve-builtins.cc
1814 (function_expander::convert_to_pmode): New function.
1815 (function_expander::get_contiguous_base): Use it.
1816 (function_expander::prepare_gather_address_operands): Likewise.
1817 * config/aarch64/aarch64-sve-builtins-sve2.cc
1818 (svwhilerw_svwhilewr_impl::expand): Likewise.
1820 2020-01-21 Szabolcs Nagy <szabolcs.nagy@arm.com>
1823 * config/aarch64/aarch64.c (aarch64_declare_function_name): Set
1824 cfun->machine->label_is_assembled.
1825 (aarch64_print_patchable_function_entry): New.
1826 (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY): Define.
1827 * config/aarch64/aarch64.h (struct machine_function): New field,
1830 2020-01-21 David Malcolm <dmalcolm@redhat.com>
1833 * ipa-profile.c (ipa_profile): Delete call_sums and set it to
1836 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1839 * cgraph.c (cgraph_edge::resolve_speculation,
1840 cgraph_edge::redirect_call_stmt_to_callee): Fix update of
1841 call_stmt_site_hash.
1843 2020-01-21 Martin Liska <mliska@suse.cz>
1845 * config/rs6000/rs6000.c (common_mode_defined): Remove
1848 2020-01-21 Richard Biener <rguenther@suse.de>
1850 PR tree-optimization/92328
1851 * tree-ssa-sccvn.c (vn_reference_lookup_3): Preserve
1852 type when value-numbering same-sized store by inserting a
1854 (eliminate_dom_walker::eliminate_stmt): When eliminating
1855 a redundant store handle bit-reinterpretation of the same value.
1857 2020-01-21 Andrew Pinski <apinski@marvel.com>
1860 * tree-into-ssa.c (prepare_block_for_update_1): Split out
1862 (prepare_block_for_update): This. Use a worklist instead of
1865 2020-01-21 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
1867 * gcc/config/arm/arm.c (clear_operation_p):
1868 Initialise last_regno, skip first iteration
1869 based on the first_set value and use ints instead
1870 of the unnecessary HOST_WIDE_INTs.
1872 2020-01-21 Jakub Jelinek <jakub@redhat.com>
1875 * config/rs6000/rs6000.c (rs6000_emit_cmove): If using fsel, punt for
1876 compare_mode other than SFmode or DFmode.
1878 2020-01-21 Kito Cheng <kito.cheng@sifive.com>
1881 * config/riscv/riscv-protos.h (riscv_hard_regno_rename_ok): New.
1882 * config/riscv/riscv.c (riscv_hard_regno_rename_ok): New.
1883 * config/riscv/riscv.h (HARD_REGNO_RENAME_OK): Defined.
1885 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1887 * config/aarch64/aarch64.c (neoversen1_tunings): Set jump_align to 4.
1889 2020-01-20 Andrew Pinski <apinski@marvell.com>
1892 * targhooks.c (default_print_patchable_function_entry): Use
1893 output_asm_insn to emit the nop instruction.
1895 2020-01-20 Fangrui Song <maskray@google.com>
1898 * targhooks.c (default_print_patchable_function_entry): Align to
1901 2020-01-20 H.J. Lu <hongjiu.lu@intel.com>
1904 * config/i386/i386.c (legitimize_tls_address): Pass Pmode to
1905 gen_tls_dynamic_gnu2_64. Compute GNU2 TLS address in ptr_mode.
1906 * config/i386/i386.md (tls_dynamic_gnu2_64): Renamed to ...
1907 (@tls_dynamic_gnu2_64_<mode>): This. Replace DI with P.
1908 (*tls_dynamic_gnu2_lea_64): Renamed to ...
1909 (*tls_dynamic_gnu2_lea_64_<mode>): This. Replace DI with P.
1910 Remove the {q} suffix from lea.
1911 (*tls_dynamic_gnu2_call_64): Renamed to ...
1912 (*tls_dynamic_gnu2_call_64_<mode>): This. Replace DI with P.
1913 (*tls_dynamic_gnu2_combine_64): Renamed to ...
1914 (*tls_dynamic_gnu2_combine_64_<mode>): This. Replace DI with P.
1915 Pass Pmode to gen_tls_dynamic_gnu2_64.
1917 2020-01-20 Wilco Dijkstra <wdijkstr@arm.com>
1919 * config/aarch64/aarch64.h (SLOW_BYTE_ACCESS): Set to 1.
1921 2020-01-20 Richard Sandiford <richard.sandiford@arm.com>
1923 * config/aarch64/aarch64-sve-builtins-base.cc
1924 (svld1ro_impl::memory_vector_mode): Remove parameter name.
1926 2020-01-20 Richard Biener <rguenther@suse.de>
1929 * dwarf2out.c (prune_unused_types): Unconditionally mark
1930 called function DIEs.
1932 2020-01-20 Martin Liska <mliska@suse.cz>
1934 PR tree-optimization/93199
1935 * tree-eh.c (struct leh_state): Add
1936 new field outer_non_cleanup.
1937 (cleanup_is_dead_in): Pass leh_state instead
1938 of eh_region. Add a checking that state->outer_non_cleanup
1939 points to outer non-clean up region.
1940 (lower_try_finally): Record outer_non_cleanup
1942 (lower_catch): Likewise.
1943 (lower_eh_filter): Likewise.
1944 (lower_eh_must_not_throw): Likewise.
1945 (lower_cleanup): Likewise.
1947 2020-01-20 Richard Biener <rguenther@suse.de>
1949 PR tree-optimization/93094
1950 * tree-vectorizer.h (vect_loop_versioning): Adjust.
1951 (vect_transform_loop): Likewise.
1952 * tree-vectorizer.c (try_vectorize_loop_1): Pass down
1953 loop_vectorized_call to vect_transform_loop.
1954 * tree-vect-loop.c (vect_transform_loop): Pass down
1955 loop_vectorized_call to vect_loop_versioning.
1956 * tree-vect-loop-manip.c (vect_loop_versioning): Use
1957 the earlier discovered loop_vectorized_call.
1959 2020-01-19 Eric S. Raymond <esr@thyrsus.com>
1961 * doc/contribute.texi: Update for SVN -> Git transition.
1962 * doc/install.texi: Likewise.
1964 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1967 * cgraph.c (cgraph_edge::make_speculative): Increase number of
1968 speculative targets.
1969 (verify_speculative_call): New function
1970 (cgraph_node::verify_node): Use it.
1971 * ipa-profile.c (ipa_profile): Fix formating; do not set number of
1974 2020-01-18 Jan Hubicka <hubicka@ucw.cz>
1977 * cgraph.c (cgraph_edge::resolve_speculation): Fix foramting.
1978 (cgraph_edge::make_direct): Remove all indirect targets.
1979 (cgraph_edge::redirect_call_stmt_to_callee): Use make_direct..
1980 (cgraph_node::verify_node): Verify that only one call_stmt or
1981 lto_stmt_uid is set.
1982 * cgraphclones.c (cgraph_edge::clone): Set only one call_stmt or
1984 * lto-cgraph.c (lto_output_edge): Simplify streaming of stmt.
1985 (lto_output_ref): Simplify streaming of stmt.
1986 * lto-streamer-in.c (fixup_call_stmt_edges_1): Clear lto_stmt_uid.
1988 2020-01-18 Tamar Christina <tamar.christina@arm.com>
1990 * config/aarch64/aarch64-sve-builtins-base.cc (memory_vector_mode):
1991 Mark parameter unused.
1993 2020-01-18 Hans-Peter Nilsson <hp@axis.com>
1995 * config.gcc <obsolete targets>: Add crisv32-*-* and cris-*-linux*
1997 2019-01-18 Gerald Pfeifer <gerald@pfeifer.com>
1999 * varpool.c (ctor_useable_for_folding_p): Fix grammar.
2001 2020-01-18 Iain Sandoe <iain@sandoe.co.uk>
2003 * Makefile.in: Add coroutine-passes.o.
2004 * builtin-types.def (BT_CONST_SIZE): New.
2005 (BT_FN_BOOL_PTR): New.
2006 (BT_FN_PTR_PTR_CONST_SIZE_BOOL): New.
2007 * builtins.def (DEF_COROUTINE_BUILTIN): New.
2008 * coroutine-builtins.def: New file.
2009 * coroutine-passes.cc: New file.
2010 * function.h (struct GTY function): Add a bit to indicate that the
2011 function is a coroutine component.
2012 * internal-fn.c (expand_CO_FRAME): New.
2013 (expand_CO_YIELD): New.
2014 (expand_CO_SUSPN): New.
2015 (expand_CO_ACTOR): New.
2016 * internal-fn.def (CO_ACTOR): New.
2020 * passes.def: Add pass_coroutine_lower_builtins,
2021 pass_coroutine_early_expand_ifns.
2022 * tree-pass.h (make_pass_coroutine_lower_builtins): New.
2023 (make_pass_coroutine_early_expand_ifns): New.
2024 * doc/invoke.texi: Document the fcoroutines command line
2027 2020-01-18 Jakub Jelinek <jakub@redhat.com>
2029 * config/arm/vfp.md (*clear_vfp_multiple): Remove unused variable.
2032 * config/arm/arm.c (clear_operation_p): Don't use REGNO until
2033 after checking the argument is a REG. Don't use REGNO (reg)
2034 again to set last_regno, reuse regno variable instead.
2036 2020-01-17 David Malcolm <dmalcolm@redhat.com>
2038 * doc/analyzer.texi (Limitations): Add note about NaN.
2040 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2041 Sudakshina Das <sudi.das@arm.com>
2043 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for both reg
2044 and valid immediate.
2045 (ashrdi3): Generate thumb2_asrl for both reg and valid immediate.
2046 (lshrdi3): Generate thumb2_lsrl for valid immediates.
2047 * config/arm/constraints.md (Pg): New.
2048 * config/arm/predicates.md (long_shift_imm): New.
2049 (arm_reg_or_long_shift_imm): Likewise.
2050 * config/arm/thumb2.md (thumb2_asrl): New immediate alternative.
2051 (thumb2_lsll): Likewise.
2054 2020-01-17 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2055 Sudakshina Das <sudi.das@arm.com>
2057 * config/arm/arm.md (ashldi3): Generate thumb2_lsll for TARGET_HAVE_MVE.
2058 (ashrdi3): Generate thumb2_asrl for TARGET_HAVE_MVE.
2059 * config/arm/arm.c (arm_hard_regno_mode_ok): Allocate even odd
2060 register pairs for doubleword quantities for ARMv8.1M-Mainline.
2061 * config/arm/thumb2.md (thumb2_asrl): New.
2062 (thumb2_lsll): Likewise.
2064 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2066 * config/arm/arm.c (cmse_nonsecure_call_inline_register_clear): Remove
2069 2020-01-17 Alexander Monakov <amonakov@ispras.ru>
2071 * gdbinit.in (help-gcc-hooks): New command.
2072 (pp, pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, ptc, pdn, ptn, pdd, prc,
2073 pi, pbm, pel, trt): Take $arg0 instead of $ if supplied. Update
2076 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2078 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): Use the
2079 correct target macro.
2081 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2083 * config/aarch64/aarch64-protos.h
2084 (aarch64_sve_ld1ro_operand_p): New.
2085 * config/aarch64/aarch64-sve-builtins-base.cc
2086 (class load_replicate): New.
2087 (class svld1ro_impl): New.
2088 (class svld1rq_impl): Change to inherit from load_replicate.
2089 (svld1ro): New sve intrinsic function base.
2090 * config/aarch64/aarch64-sve-builtins-base.def (svld1ro):
2091 New DEF_SVE_FUNCTION.
2092 * config/aarch64/aarch64-sve-builtins-base.h
2093 (svld1ro): New decl.
2094 * config/aarch64/aarch64-sve-builtins.cc
2095 (function_expander::add_mem_operand): Modify assert to allow
2097 * config/aarch64/aarch64-sve.md (@aarch64_sve_ld1ro<mode>): New
2099 * config/aarch64/aarch64.c
2100 (aarch64_sve_ld1rq_operand_p): Implement in terms of ...
2101 (aarch64_sve_ld1rq_ld1ro_operand_p): This.
2102 (aarch64_sve_ld1ro_operand_p): New.
2103 * config/aarch64/aarch64.md (UNSPEC_LD1RO): New unspec.
2104 * config/aarch64/constraints.md (UOb,UOh,UOw,UOd): New.
2105 * config/aarch64/predicates.md
2106 (aarch64_sve_ld1ro_operand_{b,h,w,d}): New.
2108 2020-01-17 Matthew Malcomson <matthew.malcomson@arm.com>
2110 * config/aarch64/aarch64-c.c (_ARM_FEATURE_MATMUL_FLOAT64):
2111 Introduce this ACLE specified predefined macro.
2112 * config/aarch64/aarch64-option-extensions.def (f64mm): New.
2113 (fp): Disabling this disables f64mm.
2114 (simd): Disabling this disables f64mm.
2115 (fp16): Disabling this disables f64mm.
2116 (sve): Disabling this disables f64mm.
2117 * config/aarch64/aarch64.h (AARCH64_FL_F64MM): New.
2118 (AARCH64_ISA_F64MM): New.
2119 (TARGET_F64MM): New.
2120 * doc/invoke.texi (f64mm): Document new option.
2122 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2124 * config/aarch64/aarch64.c (generic_tunings): Add branch fusion.
2125 (neoversen1_tunings): Likewise.
2127 2020-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2130 * config/aarch64/aarch64.c (aarch64_split_compare_and_swap)
2131 Add assert to ensure prolog has been emitted.
2132 (aarch64_split_atomic_op): Likewise.
2133 * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>)
2134 Use epilogue_completed rather than reload_completed.
2135 (aarch64_atomic_exchange<mode>): Likewise.
2136 (aarch64_atomic_<atomic_optab><mode>): Likewise.
2137 (atomic_nand<mode>): Likewise.
2138 (aarch64_atomic_fetch_<atomic_optab><mode>): Likewise.
2139 (atomic_fetch_nand<mode>): Likewise.
2140 (aarch64_atomic_<atomic_optab>_fetch<mode>): Likewise.
2141 (atomic_nand_fetch<mode>): Likewise.
2143 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2146 * config/aarch64/aarch64.h (REVERSIBLE_CC_MODE): Return false
2148 (REVERSE_CONDITION): Delete.
2149 * config/aarch64/iterators.md (CC_ONLY): New mode iterator.
2150 (CCFP_CCFPE): Likewise.
2151 (e): New mode attribute.
2152 * config/aarch64/aarch64.md (ccmp<GPI:mode>): Rename to...
2153 (@ccmp<CC_ONLY:mode><GPI:mode>): ...this, using CC_ONLY instead of CC.
2154 (fccmp<GPF:mode>, fccmpe<GPF:mode>): Merge into...
2155 (@ccmp<CCFP_CCFPE:mode><GPF:mode>): ...this combined pattern.
2156 (@ccmp<CC_ONLY:mode><GPI:mode>_rev): New pattern.
2157 (@ccmp<CCFP_CCFPE:mode><GPF:mode>_rev): Likewise.
2158 * config/aarch64/aarch64.c (aarch64_gen_compare_reg): Update
2159 name of generator from gen_ccmpdi to gen_ccmpccdi.
2160 (aarch64_gen_ccmp_next): Use code_for_ccmp. If we want to reverse
2161 the previous comparison but aren't able to, use the new ccmp_rev
2164 2020-01-17 Richard Sandiford <richard.sandiford@arm.com>
2166 * gimplify.c (gimplify_return_expr): Use poly_int_tree_p rather
2167 than testing directly for INTEGER_CST.
2168 (gimplify_target_expr, gimplify_omp_depend): Likewise.
2170 2020-01-17 Jakub Jelinek <jakub@redhat.com>
2172 PR tree-optimization/93292
2173 * tree-vect-stmts.c (vectorizable_comparison): Punt also if
2174 get_vectype_for_scalar_type returns NULL.
2176 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2178 * params.opt (-param=max-predicted-iterations): Increase range from 0.
2179 * predict.c (estimate_loops): Add 1 to param_max_predicted_iterations.
2181 2020-01-16 Jan Hubicka <hubicka@ucw.cz>
2183 * ipa-fnsummary.c (estimate_calls_size_and_time): Fix formating of
2185 * params.opt: (max-predicted-iterations): Set bounds.
2186 * predict.c (real_almost_one, real_br_prob_base,
2187 real_inv_br_prob_base, real_one_half, real_bb_freq_max): Remove.
2188 (propagate_freq): Add max_cyclic_prob parameter; cap cyclic
2189 probabilities; do not truncate to reg_br_prob_bases.
2190 (estimate_loops_at_level): Pass max_cyclic_prob.
2191 (estimate_loops): Compute max_cyclic_prob.
2192 (estimate_bb_frequencies): Do not initialize real_*; update calculation
2194 * profile-count.c (profile_probability::to_sreal): New.
2195 * profile-count.h (class sreal): Move up in file.
2196 (profile_probability::to_sreal): Declare.
2198 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2201 (arm_invalid_conversion): New function for target hook.
2202 (arm_invalid_unary_op): New function for target hook.
2203 (arm_invalid_binary_op): New function for target hook.
2205 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2207 * config.gcc: Add arm_bf16.h.
2208 * config/arm/arm-builtins.c (arm_mangle_builtin_type): Fix comment.
2209 (arm_simd_builtin_std_type): Add BFmode.
2210 (arm_init_simd_builtin_types): Define element types for vector types.
2211 (arm_init_bf16_types): New function.
2212 (arm_init_builtins): Add arm_init_bf16_types function call.
2213 * config/arm/arm-modes.def: Add BFmode and V4BF, V8BF vector modes.
2214 * config/arm/arm-simd-builtin-types.def: Add V4BF, V8BF.
2215 * config/arm/arm.c (aapcs_vfp_sub_candidate): Add BFmode.
2216 (arm_hard_regno_mode_ok): Add BFmode and tidy up statements.
2217 (arm_vector_mode_supported_p): Add V4BF, V8BF.
2218 (arm_mangle_type): Add __bf16.
2219 * config/arm/arm.h: Add V4BF, V8BF to VALID_NEON_DREG_MODE,
2220 VALID_NEON_QREG_MODE respectively. Add export arm_bf16_type_node,
2221 arm_bf16_ptr_type_node.
2222 * config/arm/arm.md: Add BFmode to movhf expand, mov pattern and
2223 define_split between ARM registers.
2224 * config/arm/arm_bf16.h: New file.
2225 * config/arm/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2226 * config/arm/iterators.md: (ANY64_BF, VDXMOV, VHFBF, HFBF, fporbf): New.
2228 * config/arm/neon.md: Add BF vector types to movhf NEON move patterns.
2229 * config/arm/vfp.md: Add BFmode to movhf patterns.
2231 2020-01-16 Mihail Ionescu <mihail.ionescu@arm.com>
2232 Andre Vieira <andre.simoesdiasvieira@arm.com>
2234 * config/arm/arm-cpus.in (mve, mve_float): New features.
2235 (dsp, mve, mve.fp): New options.
2236 * config/arm/arm.h (TARGET_HAVE_MVE, TARGET_HAVE_MVE_FLOAT): Define.
2237 * config/arm/t-rmprofile: Map v8.1-M multilibs to v8-M.
2238 * doc/invoke.texi: Document the armv8.1-m mve and dps options.
2240 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2241 Thomas Preud'homme <thomas.preudhomme@arm.com>
2243 * config/arm/arm-cpus.in (ARMv8_1m_main): Redefine as an extension to
2245 * config/arm/arm.c (arm_options_perform_arch_sanity_checks): Remove
2246 error for using -mcmse when targeting Armv8.1-M Mainline.
2248 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2249 Thomas Preud'homme <thomas.preudhomme@arm.com>
2251 * config/arm/arm.md (nonsecure_call_internal): Do not force memory
2252 address in r4 when targeting Armv8.1-M Mainline.
2253 (nonsecure_call_value_internal): Likewise.
2254 * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Make memory address
2255 a register match_operand again. Emit BLXNS when targeting
2257 (nonsecure_call_value_reg_thumb2): Likewise.
2259 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2260 Thomas Preud'homme <thomas.preudhomme@arm.com>
2262 * config/arm/arm.c (arm_add_cfa_adjust_cfa_note): Declare early.
2263 (cmse_nonsecure_call_inline_register_clear): Define new lazy_fpclear
2264 variable as true when floating-point ABI is not hard. Replace
2265 check against TARGET_HARD_FLOAT_ABI by checks against lazy_fpclear.
2266 Generate VLSTM and VLLDM instruction respectively before and
2267 after a function call to cmse_nonsecure_call function.
2268 * config/arm/unspecs.md (VUNSPEC_VLSTM): Define unspec.
2269 (VUNSPEC_VLLDM): Likewise.
2270 * config/arm/vfp.md (lazy_store_multiple_insn): New define_insn.
2271 (lazy_load_multiple_insn): Likewise.
2273 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2274 Thomas Preud'homme <thomas.preudhomme@arm.com>
2276 * config/arm/arm.c (vfp_emit_fstmd): Declare early.
2277 (arm_emit_vfp_multi_reg_pop): Likewise.
2278 (cmse_nonsecure_call_inline_register_clear): Abstract number of VFP
2279 registers to clear in max_fp_regno. Emit VPUSH and VPOP to save and
2280 restore callee-saved VFP registers.
2282 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2283 Thomas Preud'homme <thomas.preudhomme@arm.com>
2285 * config/arm/arm.c (arm_emit_multi_reg_pop): Declare early.
2286 (cmse_nonsecure_call_clear_caller_saved): Rename into ...
2287 (cmse_nonsecure_call_inline_register_clear): This. Save and clear
2288 callee-saved GPRs as well as clear ip register before doing a nonsecure
2289 call then restore callee-saved GPRs after it when targeting
2291 (arm_reorg): Adapt to function rename.
2293 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2294 Thomas Preud'homme <thomas.preudhomme@arm.com>
2296 * config/arm/arm-protos.h (clear_operation_p): Adapt prototype.
2297 * config/arm/arm.c (clear_operation_p): Extend to be able to check a
2298 clear_vfp_multiple pattern based on a new vfp parameter.
2299 (cmse_clear_registers): Generate VSCCLRM to clear VFP registers when
2300 targeting Armv8.1-M Mainline.
2301 (cmse_nonsecure_entry_clear_before_return): Clear VFP registers
2302 unconditionally when targeting Armv8.1-M Mainline architecture. Check
2303 whether VFP registers are available before looking call_used_regs for a
2305 * config/arm/predicates.md (clear_multiple_operation): Adapt to change
2306 of prototype of clear_operation_p.
2307 (clear_vfp_multiple_operation): New predicate.
2308 * config/arm/unspecs.md (VUNSPEC_VSCCLRM_VPR): New volatile unspec.
2309 * config/arm/vfp.md (clear_vfp_multiple): New define_insn.
2311 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2312 Thomas Preud'homme <thomas.preudhomme@arm.com>
2314 * config/arm/arm-protos.h (clear_operation_p): Declare.
2315 * config/arm/arm.c (clear_operation_p): New function.
2316 (cmse_clear_registers): Generate clear_multiple instruction pattern if
2317 targeting Armv8.1-M Mainline or successor.
2318 (output_return_instruction): Only output APSR register clearing if
2319 Armv8.1-M Mainline instructions not available.
2320 (thumb_exit): Likewise.
2321 * config/arm/predicates.md (clear_multiple_operation): New predicate.
2322 * config/arm/thumb2.md (clear_apsr): New define_insn.
2323 (clear_multiple): Likewise.
2324 * config/arm/unspecs.md (VUNSPEC_CLRM_APSR): New volatile unspec.
2326 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2327 Thomas Preud'homme <thomas.preudhomme@arm.com>
2329 * config/arm/arm.c (fp_sysreg_names): Declare and define.
2330 (use_return_insn): Also return false for Armv8.1-M Mainline.
2331 (output_return_instruction): Skip FPSCR clearing if Armv8.1-M
2332 Mainline instructions are available.
2333 (arm_compute_frame_layout): Allocate space in frame for FPCXTNS
2334 when targeting Armv8.1-M Mainline Security Extensions.
2335 (arm_expand_prologue): Save FPCXTNS if this is an Armv8.1-M
2336 Mainline entry function.
2337 (cmse_nonsecure_entry_clear_before_return): Clear IP and r4 if
2338 targeting Armv8.1-M Mainline or successor.
2339 (arm_expand_epilogue): Fix indentation of caller-saved register
2340 clearing. Restore FPCXTNS if this is an Armv8.1-M Mainline
2342 * config/arm/arm.h (TARGET_HAVE_FP_CMSE): New macro.
2343 (FP_SYSREGS): Likewise.
2344 (enum vfp_sysregs_encoding): Define enum.
2345 (fp_sysreg_names): Declare.
2346 * config/arm/unspecs.md (VUNSPEC_VSTR_VLDR): New volatile unspec.
2347 * config/arm/vfp.md (push_fpsysreg_insn): New define_insn.
2348 (pop_fpsysreg_insn): Likewise.
2350 2020-01-16 Mihail-Calin Ionescu <mihail.ionescu@arm.com>
2351 Thomas Preud'homme <thomas.preudhomme@arm.com>
2353 * config/arm/arm-cpus.in (armv8_1m_main): New feature.
2354 (ARMv4, ARMv4t, ARMv5t, ARMv5te, ARMv5tej, ARMv6, ARMv6j, ARMv6k,
2355 ARMv6z, ARMv6kz, ARMv6zk, ARMv6t2, ARMv6m, ARMv7, ARMv7a, ARMv7ve,
2356 ARMv7r, ARMv7m, ARMv7em, ARMv8a, ARMv8_1a, ARMv8_2a, ARMv8_3a,
2357 ARMv8_4a, ARMv8_5a, ARMv8m_base, ARMv8m_main, ARMv8r): Reindent.
2358 (ARMv8_1m_main): New feature group.
2359 (armv8.1-m.main): New architecture.
2360 * config/arm/arm-tables.opt: Regenerate.
2361 * config/arm/arm.c (arm_arch8_1m_main): Define and default initialize.
2362 (arm_option_reconfigure_globals): Initialize arm_arch8_1m_main.
2363 (arm_options_perform_arch_sanity_checks): Error out when targeting
2364 Armv8.1-M Mainline Security Extensions.
2365 * config/arm/arm.h (arm_arch8_1m_main): Declare.
2367 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2369 * config/aarch64/aarch64-simd-builtins.def (aarch64_bfdot,
2370 aarch64_bfdot_lane, aarch64_bfdot_laneq): New.
2371 * config/aarch64/aarch64-simd.md (aarch64_bfdot, aarch64_bfdot_lane,
2372 aarch64_bfdot_laneq): New.
2373 * config/aarch64/arm_bf16.h (vbfdot_f32, vbfdotq_f32,
2374 vbfdot_lane_f32, vbfdotq_lane_f32, vbfdot_laneq_f32,
2375 vbfdotq_laneq_f32): New.
2376 * config/aarch64/iterators.md (UNSPEC_BFDOT, Vbfdottype,
2377 VBFMLA_W, VBF): New.
2378 (isquadop): Add V4BF, V8BF.
2380 2020-01-16 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2382 * config/aarch64/aarch64-builtins.c: (enum aarch64_type_qualifiers):
2383 New qualifier_lane_quadtup_index, TYPES_TERNOP_SSUS,
2384 TYPES_QUADOPSSUS_LANE_QUADTUP, TYPES_QUADOPSSSU_LANE_QUADTUP.
2385 (aarch64_simd_expand_args): Add case SIMD_ARG_LANE_QUADTUP_INDEX.
2386 (aarch64_simd_expand_builtin): Add qualifier_lane_quadtup_index.
2387 * config/aarch64/aarch64-simd-builtins.def (usdot, usdot_lane,
2388 usdot_laneq, sudot_lane,sudot_laneq): New.
2389 * config/aarch64/aarch64-simd.md (aarch64_usdot): New.
2390 (aarch64_<sur>dot_lane): New.
2391 * config/aarch64/arm_neon.h (vusdot_s32): New.
2393 (vusdot_lane_s32): New.
2394 (vsudot_lane_s32): New.
2395 * config/aarch64/iterators.md (DOTPROD_I8MM): New iterator.
2396 (UNSPEC_USDOT, UNSPEC_SUDOT): New unspecs.
2398 2020-01-16 Martin Liska <mliska@suse.cz>
2400 * value-prof.c (dump_histogram_value): Fix
2401 obvious spacing issue.
2403 2020-01-16 Andrew Pinski <apinski@marvell.com>
2405 * tree-ssa-sccvn.c(vn_reference_lookup_3): Check lhs for
2406 !storage_order_barrier_p.
2408 2020-01-16 Andrew Pinski <apinski@marvell.com>
2410 * sched-int.h (_dep): Add unused bit-field field for the padding.
2411 * sched-deps.c (init_dep_1): Init unused field.
2413 2020-01-16 Andrew Pinski <apinski@marvell.com>
2415 * optabs.h (create_expand_operand): Initialize target field also.
2417 2020-01-16 Andre Vieira <andre.simoesdiasvieira@arm.com>
2419 PR tree-optimization/92429
2420 * tree-ssa-loop-niter.h (simplify_replace_tree): Add parameter.
2421 * tree-ssa-loop-niter.c (simplify_replace_tree): Add parameter to
2423 * tree-vect-loop.c (update_epilogue_vinfo): Do not fold when replacing
2426 2020-01-16 Richard Sandiford <richard.sandiford@arm.com>
2428 * config/aarch64/aarch64.c (aarch64_split_sve_subreg_move): Apply
2429 aarch64_sve_int_mode to each mode.
2431 2020-01-15 David Malcolm <dmalcolm@redhat.com>
2433 * doc/analyzer.texi (Overview): Add note about
2434 -fdump-ipa-analyzer.
2436 2020-01-15 Wilco Dijkstra <wdijkstr@arm.com>
2438 PR tree-optimization/93231
2439 * tree-ssa-forwprop.c (optimize_count_trailing_zeroes): Check
2440 input_type is unsigned. Use tree_to_shwi for shift constant.
2441 Check CST_STRING element size is CHAR_TYPE_SIZE bits.
2442 (simplify_count_trailing_zeroes): Add test to handle known non-zero
2443 inputs more efficiently.
2445 2020-01-15 Uroš Bizjak <ubizjak@gmail.com>
2447 * config/i386/i386.md (*movsf_internal): Do not require
2448 SSE2 ISA for alternatives 14 and 15.
2450 2020-01-15 Richard Biener <rguenther@suse.de>
2453 * tree-eh.c (sink_clobbers): If we already visited the destination
2454 block do not defer insertion.
2455 (pass_lower_eh_dispatch::execute): Maintain BB_VISITED for
2456 the purpose of defered insertion.
2458 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2460 * BASE-VER: Bump to 10.0.1.
2462 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2464 PR tree-optimization/93247
2465 * tree-vect-loop.c (update_epilogue_loop_vinfo): Check the access
2466 type of the stmt that we're going to vectorize.
2468 2020-01-15 Richard Sandiford <richard.sandiford@arm.com>
2470 * tree-vect-slp.c (vectorize_slp_instance_root_stmt): Use a
2471 VIEW_CONVERT_EXPR if the vectorized constructor has a diffeent
2474 2020-01-15 Martin Liska <mliska@suse.cz>
2476 * ipa-profile.c (ipa_profile_read_edge_summary): Do not allow
2477 2 calls of streamer_read_hwi in a function call.
2479 2020-01-15 Richard Biener <rguenther@suse.de>
2481 * alias.c (record_alias_subset): Avoid redundant work when
2482 subset is already recorded.
2484 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2486 * doc/invoke.texi (-fdiagnostics-show-cwe): Add note that some of
2487 the analyzer options provide CWE identifiers.
2489 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2491 * tree-diagnostic-path.cc (path_summary::event_range::print):
2492 When testing for UNKNOWN_LOCATION, look through ad-hoc wrappers
2493 using get_pure_location.
2495 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2497 PR tree-optimization/93262
2498 * tree-ssa-dse.c (maybe_trim_memstar_call): For *_chk builtins,
2499 perform head trimming only if the last argument is constant,
2500 either all ones, or larger or equal to head trim, in the latter
2501 case decrease the last argument by head_trim.
2503 PR tree-optimization/93249
2504 * tree-ssa-dse.c: Include builtins.h and gimple-fold.h.
2505 (maybe_trim_memstar_call): Move head_trim and tail_trim vars to
2506 function body scope, reindent. For BUILTIN_IN_STRNCPY*, don't
2507 perform head trim unless we can prove there are no '\0' chars
2508 from the source among the first head_trim chars.
2510 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2512 * Makefile.in (ANALYZER_OBJS): Add analyzer/function-set.o.
2514 2020-01-15 Jakub Jelinek <jakub@redhat.com>
2517 * config/i386/sse.md
2518 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1,
2519 *<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1,
2520 *<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1,
2521 *<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1): Use
2522 just a single alternative instead of two, make operands 1 and 2
2525 2020-01-14 Jan Hubicka <hubicka@ucw.cz>
2528 * ipa-devirt.c (odr_types_equivalent_p): Compare TREE_ADDRESSABLE and
2531 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2533 * Makefile.in (lang_opt_files): Add analyzer.opt.
2534 (ANALYZER_OBJS): New.
2535 (OBJS): Add digraph.o, graphviz.o, ordered-hash-map-tests.o,
2536 tristate.o and ANALYZER_OBJS.
2537 (TEXI_GCCINT_FILES): Add analyzer.texi.
2538 * common.opt (-fanalyzer): New driver option.
2539 * config.in: Regenerate.
2540 * configure: Regenerate.
2541 * configure.ac (--disable-analyzer, ENABLE_ANALYZER): New option.
2542 (gccdepdir): Also create depdir for "analyzer" subdir.
2543 * digraph.cc: New file.
2544 * digraph.h: New file.
2545 * doc/analyzer.texi: New file.
2546 * doc/gccint.texi ("Static Analyzer") New menu item.
2547 (analyzer.texi): Include it.
2548 * doc/invoke.texi ("Static Analyzer Options"): New list and new section.
2549 ("Warning Options"): Add static analysis warnings to the list.
2550 (-Wno-analyzer-double-fclose): New option.
2551 (-Wno-analyzer-double-free): New option.
2552 (-Wno-analyzer-exposure-through-output-file): New option.
2553 (-Wno-analyzer-file-leak): New option.
2554 (-Wno-analyzer-free-of-non-heap): New option.
2555 (-Wno-analyzer-malloc-leak): New option.
2556 (-Wno-analyzer-possible-null-argument): New option.
2557 (-Wno-analyzer-possible-null-dereference): New option.
2558 (-Wno-analyzer-null-argument): New option.
2559 (-Wno-analyzer-null-dereference): New option.
2560 (-Wno-analyzer-stale-setjmp-buffer): New option.
2561 (-Wno-analyzer-tainted-array-index): New option.
2562 (-Wno-analyzer-use-after-free): New option.
2563 (-Wno-analyzer-use-of-pointer-in-stale-stack-frame): New option.
2564 (-Wno-analyzer-use-of-uninitialized-value): New option.
2565 (-Wanalyzer-too-complex): New option.
2566 (-fanalyzer-call-summaries): New warning.
2567 (-fanalyzer-checker=): New warning.
2568 (-fanalyzer-fine-grained): New warning.
2569 (-fno-analyzer-state-merge): New warning.
2570 (-fno-analyzer-state-purge): New warning.
2571 (-fanalyzer-transitivity): New warning.
2572 (-fanalyzer-verbose-edges): New warning.
2573 (-fanalyzer-verbose-state-changes): New warning.
2574 (-fanalyzer-verbosity=): New warning.
2575 (-fdump-analyzer): New warning.
2576 (-fdump-analyzer-callgraph): New warning.
2577 (-fdump-analyzer-exploded-graph): New warning.
2578 (-fdump-analyzer-exploded-nodes): New warning.
2579 (-fdump-analyzer-exploded-nodes-2): New warning.
2580 (-fdump-analyzer-exploded-nodes-3): New warning.
2581 (-fdump-analyzer-supergraph): New warning.
2582 * doc/sourcebuild.texi (dg-require-dot): New.
2583 (dg-check-dot): New.
2584 * gdbinit.in (break-on-saved-diagnostic): New command.
2585 * graphviz.cc: New file.
2586 * graphviz.h: New file.
2587 * ordered-hash-map-tests.cc: New file.
2588 * ordered-hash-map.h: New file.
2589 * passes.def (pass_analyzer): Add before
2590 pass_ipa_whole_program_visibility.
2591 * selftest-run-tests.c (selftest::run_tests): Call
2592 selftest::ordered_hash_map_tests_cc_tests.
2593 * selftest.h (selftest::ordered_hash_map_tests_cc_tests): New
2595 * shortest-paths.h: New file.
2596 * timevar.def (TV_ANALYZER): New timevar.
2597 (TV_ANALYZER_SUPERGRAPH): Likewise.
2598 (TV_ANALYZER_STATE_PURGE): Likewise.
2599 (TV_ANALYZER_PLAN): Likewise.
2600 (TV_ANALYZER_SCC): Likewise.
2601 (TV_ANALYZER_WORKLIST): Likewise.
2602 (TV_ANALYZER_DUMP): Likewise.
2603 (TV_ANALYZER_DIAGNOSTICS): Likewise.
2604 (TV_ANALYZER_SHORTEST_PATHS): Likewise.
2605 * tree-pass.h (make_pass_analyzer): New decl.
2606 * tristate.cc: New file.
2607 * tristate.h: New file.
2609 2020-01-14 Uroš Bizjak <ubizjak@gmail.com>
2612 * config/i386/i386.md (*movsf_internal): Require SSE2 ISA for
2613 alternatives 9 and 10.
2615 2020-01-14 David Malcolm <dmalcolm@redhat.com>
2617 * attribs.c (excl_hash_traits::empty_zero_p): New static constant.
2618 * gcov.c (function_start_pair_hash::empty_zero_p): Likewise.
2619 * graphite.c (struct sese_scev_hash::empty_zero_p): Likewise.
2620 * hash-map-tests.c (selftest::test_nonzero_empty_key): New selftest.
2621 (selftest::hash_map_tests_c_tests): Call it.
2622 * hash-map-traits.h (simple_hashmap_traits::empty_zero_p):
2623 New static constant, using the value of = H::empty_zero_p.
2624 (unbounded_hashmap_traits::empty_zero_p): Likewise, using the value
2625 from default_hash_traits <Value>.
2626 * hash-map.h (hash_map::empty_zero_p): Likewise, using the value
2628 * hash-set-tests.c (value_hash_traits::empty_zero_p): Likewise.
2629 * hash-table.h (hash_table::alloc_entries): Guard the loop of
2630 calls to mark_empty with !Descriptor::empty_zero_p.
2631 (hash_table::empty_slow): Conditionalize the memset call with a
2632 check that Descriptor::empty_zero_p; otherwise, loop through the
2633 entries calling mark_empty on them.
2634 * hash-traits.h (int_hash::empty_zero_p): New static constant.
2635 (pointer_hash::empty_zero_p): Likewise.
2636 (pair_hash::empty_zero_p): Likewise.
2637 * ipa-devirt.c (default_hash_traits <type_pair>::empty_zero_p):
2639 * ipa-prop.c (ipa_bit_ggc_hash_traits::empty_zero_p): Likewise.
2640 (ipa_vr_ggc_hash_traits::empty_zero_p): Likewise.
2641 * profile.c (location_triplet_hash::empty_zero_p): Likewise.
2642 * sanopt.c (sanopt_tree_triplet_hash::empty_zero_p): Likewise.
2643 (sanopt_tree_couple_hash::empty_zero_p): Likewise.
2644 * tree-hasher.h (int_tree_hasher::empty_zero_p): Likewise.
2645 * tree-ssa-sccvn.c (vn_ssa_aux_hasher::empty_zero_p): Likewise.
2646 * tree-vect-slp.c (bst_traits::empty_zero_p): Likewise.
2648 (default_hash_traits<scalar_cond_masked_key>::empty_zero_p):
2651 2020-01-14 Kewen Lin <linkw@gcc.gnu.org>
2653 * cfgloopanal.c (average_num_loop_insns): Free bbs when early return,
2654 fix typo on return value.
2656 2020-01-14 Xiong Hu Luo <luoxhu@linux.ibm.com>
2659 * cgraph.c (symbol_table::create_edge): Init speculative_id and
2661 (cgraph_edge::make_speculative): Add param for setting speculative_id
2663 (cgraph_edge::speculative_call_info): Update comments and find reference
2664 by speculative_id for multiple indirect targets.
2665 (cgraph_edge::resolve_speculation): Decrease the speculations
2666 for indirect edge, drop it's speculative if not direct target
2667 left. Update comments.
2668 (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
2669 (cgraph_node::dump): Print num_speculative_call_targets.
2670 (cgraph_node::verify_node): Don't report error if speculative
2671 edge not include statement.
2672 (cgraph_edge::num_speculative_call_targets_p): New function.
2673 * cgraph.h (int common_target_id): Remove.
2674 (int common_target_probability): Remove.
2675 (num_speculative_call_targets): New variable.
2676 (make_speculative): Add param for setting speculative_id.
2677 (cgraph_edge::num_speculative_call_targets_p): New declare.
2678 (target_prob): New variable.
2679 (speculative_id): New variable.
2680 * ipa-fnsummary.c (analyze_function_body): Create and duplicate
2681 call summaries for multiple speculative call targets.
2682 * cgraphclones.c (cgraph_node::create_clone): Clone speculative_id.
2683 * ipa-profile.c (struct speculative_call_target): New struct.
2684 (class speculative_call_summary): New class.
2685 (class speculative_call_summaries): New class.
2686 (call_sums): New variable.
2687 (ipa_profile_generate_summary): Generate indirect multiple targets summaries.
2688 (ipa_profile_write_edge_summary): New function.
2689 (ipa_profile_write_summary): Stream out indirect multiple targets summaries.
2690 (ipa_profile_dump_all_summaries): New function.
2691 (ipa_profile_read_edge_summary): New function.
2692 (ipa_profile_read_summary_section): New function.
2693 (ipa_profile_read_summary): Stream in indirect multiple targets summaries.
2694 (ipa_profile): Generate num_speculative_call_targets from
2696 * ipa-ref.h (speculative_id): New variable.
2697 * ipa-utils.c (ipa_merge_profiles): Update with target_prob.
2698 * lto-cgraph.c (lto_output_edge): Remove indirect common_target_id and
2699 common_target_probability. Stream out speculative_id and
2700 num_speculative_call_targets.
2701 (input_edge): Likewise.
2702 * predict.c (dump_prediction): Remove edges count assert to be
2704 * symtab.c (symtab_node::create_reference): Init speculative_id.
2705 (symtab_node::clone_references): Clone speculative_id.
2706 (symtab_node::clone_referring): Clone speculative_id.
2707 (symtab_node::clone_reference): Clone speculative_id.
2708 (symtab_node::clear_stmts_in_references): Clear speculative_id.
2709 * tree-inline.c (copy_bb): Duplicate all the speculative edges
2710 if indirect call contains multiple speculative targets.
2711 * value-prof.h (check_ic_target): Remove.
2712 * value-prof.c (gimple_value_profile_transformations):
2713 Use void function gimple_ic_transform.
2714 * value-prof.c (gimple_ic_transform): Handle topn case.
2715 Fix comment typos. Change it to a void function.
2717 2020-01-13 Andrew Pinski <apinski@marvell.com>
2719 * config/aarch64/aarch64-cores.def (octeontx2): New define.
2720 (octeontx2t98): New define.
2721 (octeontx2t96): New define.
2722 (octeontx2t93): New define.
2723 (octeontx2f95): New define.
2724 (octeontx2f95n): New define.
2725 (octeontx2f95mm): New define.
2726 * config/aarch64/aarch64-tune.md: Regenerate.
2727 * doc/invoke.texi (-mcpu=): Document the new cpu types.
2729 2020-01-13 Jason Merrill <jason@redhat.com>
2731 PR c++/33799 - destroy return value if local cleanup throws.
2732 * gimplify.c (gimplify_return_expr): Handle COMPOUND_EXPR.
2734 2020-01-13 Martin Liska <mliska@suse.cz>
2736 * ipa-cp.c (get_max_overall_size): Use newly
2737 renamed param param_ipa_cp_unit_growth.
2738 * params.opt: Remove legacy param name.
2740 2020-01-13 Martin Sebor <msebor@redhat.com>
2742 PR tree-optimization/93213
2743 * tree-ssa-strlen.c (handle_store): Only allow single-byte nul-over-nul
2744 stores to be eliminated.
2746 2020-01-13 Martin Liska <mliska@suse.cz>
2748 * opts.c (print_help): Do not print CL_PARAM
2749 and CL_WARNING for CL_OPTIMIZATION.
2751 2020-01-13 Jonathan Wakely <jwakely@redhat.com>
2754 * doc/invoke.texi (Warning Options): Add caveat about some warnings
2755 depending on optimization settings.
2757 2020-01-13 Jakub Jelinek <jakub@redhat.com>
2759 PR tree-optimization/90838
2760 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2761 SCALAR_INT_TYPE_MODE directly in CTZ_DEFINED_VALUE_AT_ZERO macro
2762 argument rather than to initialize temporary for targets that
2763 don't use the mode argument at all. Initialize ctzval to avoid
2766 2020-01-10 Thomas Schwinge <thomas@codesourcery.com>
2768 * tree.h (OMP_CLAUSE_USE_DEVICE_PTR_IF_PRESENT): New definition.
2769 * tree-core.h: Document it.
2770 * gimplify.c (gimplify_omp_workshare): Set it.
2771 * omp-low.c (lower_omp_target): Use it.
2772 * tree-pretty-print.c (dump_omp_clause): Print it.
2774 * omp-low.c (lower_omp_target) <OMP_CLAUSE_USE_DEVICE_PTR etc.>:
2775 Assert that for OpenACC we always have 'GOMP_MAP_USE_DEVICE_PTR'.
2777 2020-01-10 David Malcolm <dmalcolm@redhat.com>
2779 * Makefile.in (OBJS): Add tree-diagnostic-path.o.
2780 * common.opt (fdiagnostics-path-format=): New option.
2781 (diagnostic_path_format): New enum.
2782 (fdiagnostics-show-path-depths): New option.
2783 * coretypes.h (diagnostic_event_id_t): New forward decl.
2784 * diagnostic-color.c (color_dict): Add "path".
2785 * diagnostic-event-id.h: New file.
2786 * diagnostic-format-json.cc (json_from_expanded_location): Make
2788 (json_end_diagnostic): Call context->make_json_for_path if it
2789 exists and the diagnostic has a path.
2790 (diagnostic_output_format_init): Clear context->print_path.
2791 * diagnostic-path.h: New file.
2792 * diagnostic-show-locus.c (colorizer::set_range): Special-case
2793 when printing a run of events in a diagnostic_path so that they
2794 all get the same color.
2795 (layout::m_diagnostic_path_p): New field.
2796 (layout::layout): Initialize it.
2797 (layout::print_any_labels): Don't colorize the label text for an
2798 event in a diagnostic_path.
2799 (gcc_rich_location::add_location_if_nearby): Add
2800 "restrict_to_current_line_spans" and "label" params. Pass the
2801 former to layout.maybe_add_location_range; pass the latter
2802 when calling add_range.
2803 * diagnostic.c: Include "diagnostic-path.h".
2804 (diagnostic_initialize): Initialize context->path_format and
2805 context->show_path_depths.
2806 (diagnostic_show_any_path): New function.
2807 (diagnostic_path::interprocedural_p): New function.
2808 (diagnostic_report_diagnostic): Call diagnostic_show_any_path.
2809 (simple_diagnostic_path::num_events): New function.
2810 (simple_diagnostic_path::get_event): New function.
2811 (simple_diagnostic_path::add_event): New function.
2812 (simple_diagnostic_event::simple_diagnostic_event): New ctor.
2813 (simple_diagnostic_event::~simple_diagnostic_event): New dtor.
2814 (debug): New overload taking a diagnostic_path *.
2815 * diagnostic.def (DK_DIAGNOSTIC_PATH): New.
2816 * diagnostic.h (enum diagnostic_path_format): New enum.
2817 (json::value): New forward decl.
2818 (diagnostic_context::path_format): New field.
2819 (diagnostic_context::show_path_depths): New field.
2820 (diagnostic_context::print_path): New callback field.
2821 (diagnostic_context::make_json_for_path): New callback field.
2822 (diagnostic_show_any_path): New decl.
2823 (json_from_expanded_location): New decl.
2824 * doc/invoke.texi (-fdiagnostics-path-format=): New option.
2825 (-fdiagnostics-show-path-depths): New option.
2826 (-fdiagnostics-color): Add "path" to description of default
2827 GCC_COLORS; describe it.
2828 (-fdiagnostics-format=json): Document how diagnostic paths are
2829 represented in the JSON output format.
2830 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
2831 Add optional params "restrict_to_current_line_spans" and "label".
2832 * opts.c (common_handle_option): Handle
2833 OPT_fdiagnostics_path_format_ and
2834 OPT_fdiagnostics_show_path_depths.
2835 * pretty-print.c: Include "diagnostic-event-id.h".
2836 (pp_format): Implement "%@" format code for printing
2837 diagnostic_event_id_t *.
2838 (selftest::test_pp_format): Add tests for "%@".
2839 * selftest-run-tests.c (selftest::run_tests): Call
2840 selftest::tree_diagnostic_path_cc_tests.
2841 * selftest.h (selftest::tree_diagnostic_path_cc_tests): New decl.
2842 * toplev.c (general_init): Initialize global_dc->path_format and
2843 global_dc->show_path_depths.
2844 * tree-diagnostic-path.cc: New file.
2845 * tree-diagnostic.c (maybe_unwind_expanded_macro_loc): Make
2846 non-static. Drop "diagnostic" param in favor of storing the
2847 original value of "where" and re-using it.
2848 (virt_loc_aware_diagnostic_finalizer): Update for dropped param of
2849 maybe_unwind_expanded_macro_loc.
2850 (tree_diagnostics_defaults): Initialize context->print_path and
2851 context->make_json_for_path.
2852 * tree-diagnostic.h (default_tree_diagnostic_path_printer): New
2854 (default_tree_make_json_for_path): New decl.
2855 (maybe_unwind_expanded_macro_loc): New decl.
2857 2020-01-10 Jakub Jelinek <jakub@redhat.com>
2859 PR tree-optimization/93210
2860 * fold-const.h (native_encode_initializer,
2861 can_native_interpret_type_p): Declare.
2862 * fold-const.c (native_encode_string): Fix up handling with off != -1,
2864 (native_encode_initializer): New function, moved from dwarf2out.c.
2865 Adjust to native_encode_expr compatible arguments, including dry-run
2866 and partial extraction modes. Don't handle STRING_CST.
2867 (can_native_interpret_type_p): No longer static.
2868 * gimple-fold.c (fold_ctor_reference): For native_encode_expr, verify
2869 offset / BITS_PER_UNIT fits into int and don't call it if
2870 can_native_interpret_type_p fails. If suboff is NULL and for
2871 CONSTRUCTOR fold_{,non}array_ctor_reference returns NULL, retry with
2872 native_encode_initializer.
2873 (fold_const_aggregate_ref_1): Formatting fix.
2874 * dwarf2out.c (native_encode_initializer): Moved to fold-const.c.
2875 (tree_add_const_value_attribute): Adjust caller.
2877 PR tree-optimization/90838
2878 * tree-ssa-forwprop.c (simplify_count_trailing_zeroes): Use
2879 SCALAR_INT_TYPE_MODE instead of TYPE_MODE as operand of
2880 CTZ_DEFINED_VALUE_AT_ZERO.
2882 2020-01-10 Vladimir Makarov <vmakarov@redhat.com>
2885 * lra-constraints.c (match_reload): Permit input operands have the
2886 same mode as output while other input operands have a different
2889 2020-01-10 Wilco Dijkstra <wdijkstr@arm.com>
2891 PR tree-optimization/90838
2892 * tree-ssa-forwprop.c (check_ctz_array): Add new function.
2893 (check_ctz_string): Likewise.
2894 (optimize_count_trailing_zeroes): Likewise.
2895 (simplify_count_trailing_zeroes): Likewise.
2896 (pass_forwprop::execute): Try ctz simplification.
2897 * match.pd: Add matching for ctz idioms.
2899 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2901 * config/aarch64/aarch64.c (aarch64_invalid_conversion): New function
2903 (aarch64_invalid_unary_op): New function for target hook.
2904 (aarch64_invalid_binary_op): New function for target hook.
2906 2020-01-10 Stam Markianos-Wright <stam.markianos-wright@arm.com>
2908 * config.gcc: Add arm_bf16.h.
2909 * config/aarch64/aarch64-builtins.c
2910 (aarch64_simd_builtin_std_type): Add BFmode.
2911 (aarch64_init_simd_builtin_types): Define element types for vector
2913 (aarch64_init_bf16_types): New function.
2914 (aarch64_general_init_builtins): Add arm_init_bf16_types function call.
2915 * config/aarch64/aarch64-modes.def: Add BFmode and V4BF, V8BF vector
2917 * config/aarch64/aarch64-simd-builtin-types.def: Add BF SIMD types.
2918 * config/aarch64/aarch64-simd.md: Add BF vector types to NEON move
2920 * config/aarch64/aarch64.h (AARCH64_VALID_SIMD_DREG_MODE): Add V4BF.
2921 (AARCH64_VALID_SIMD_QREG_MODE): Add V8BF.
2922 * config/aarch64/aarch64.c
2923 (aarch64_classify_vector_mode): Add support for BF types.
2924 (aarch64_gimplify_va_arg_expr): Add support for BF types.
2925 (aarch64_vq_mode): Add support for BF types.
2926 (aarch64_simd_container_mode): Add support for BF types.
2927 (aarch64_mangle_type): Add support for BF scalar type.
2928 * config/aarch64/aarch64.md: Add BFmode to movhf pattern.
2929 * config/aarch64/arm_bf16.h: New file.
2930 * config/aarch64/arm_neon.h: Add arm_bf16.h and Bfloat vector types.
2931 * config/aarch64/iterators.md: Add BF types to mode attributes.
2932 (HFBF, GPF_TF_F16_MOV, VDMOV, VQMOV, VQMOV_NO2Em VALL_F16MOV): New.
2934 2020-01-10 Jason Merrill <jason@redhat.com>
2936 PR c++/93173 - incorrect tree sharing.
2937 * gimplify.c (copy_if_shared): No longer static.
2938 * gimplify.h: Declare it.
2940 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2942 * doc/invoke.texi (-msve-vector-bits=): Document that
2943 -msve-vector-bits=128 now generates VL-specific code for
2944 little-endian targets.
2945 * config/aarch64/aarch64-sve-builtins.cc (register_builtin_types): Use
2946 build_vector_type_for_mode to construct the data vector types.
2947 * config/aarch64/aarch64.c (aarch64_convert_sve_vector_bits): Generate
2948 VL-specific code for -msve-vector-bits=128 on little-endian targets.
2949 (aarch64_simd_container_mode): Always prefer Advanced SIMD modes
2950 for 128-bit vectors.
2952 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2954 * config/aarch64/aarch64.c (aarch64_evpc_sel): Fix gen_vcond_mask
2957 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2959 * config/aarch64/aarch64-builtins.c
2960 (aarch64_builtin_vectorized_function): Check for specific vector modes,
2961 rather than checking the number of elements and the element mode.
2963 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2965 * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
2966 get_related_vectype_for_scalar_type rather than build_vector_type
2967 to create the index type for a conditional reduction.
2969 2020-01-10 Richard Sandiford <richard.sandiford@arm.com>
2971 * tree-vect-loop.c (update_epilogue_loop_vinfo): Update DR_REF
2972 for any type of gather or scatter, including strided accesses.
2974 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2976 * tree-vectorizer.h (get_dr_vinfo_offset): Add missing function
2979 2020-01-10 Andre Vieira <andre.simoesdiasvieira@arm.com>
2981 * tree-vect-data-refs.c (vect_create_addr_base_for_vector_ref): Use
2983 * tree-vect-loop.c (update_epilogue_loop_vinfo): Remove orig_drs_init
2984 parameter and its use to reset DR_OFFSET's.
2985 (vect_transform_loop): Remove orig_drs_init argument.
2986 * tree-vect-loop-manip.c (vect_update_init_of_dr): Update the offset
2987 member of dr_vec_info rather than the offset of the associated
2988 data_reference's innermost_loop_behavior.
2989 (vect_update_init_of_dr): Pass dr_vec_info instead of data_reference.
2990 (vect_do_peeling): Remove orig_drs_init parameter and its construction.
2991 * tree-vect-stmts.c (check_scan_store): Replace use of DR_OFFSET with
2992 get_dr_vinfo_offset.
2993 (vectorizable_store): Likewise.
2994 (vectorizable_load): Likewise.
2996 2020-01-10 Richard Biener <rguenther@suse.de>
2998 * gimple-ssa-store-merging
2999 (pass_store_merging::terminate_all_aliasing_chains): Cache alias info.
3001 2020-01-10 Martin Liska <mliska@suse.cz>
3004 * ipa-inline-analysis.c (offline_size): Make proper parenthesis
3005 encapsulation that was there before r280040.
3007 2020-01-10 Richard Biener <rguenther@suse.de>
3010 * tree-eh.c (sink_clobbers): Move clobbers to out-of-IL
3011 sequences to avoid walking them again for secondary opportunities.
3012 (pass_lower_eh_dispatch::execute): Instead actually insert
3015 2020-01-10 Richard Biener <rguenther@suse.de>
3018 * tree-eh.c (redirect_eh_edge_1): Avoid some work if possible.
3019 (cleanup_all_empty_eh): Walk landing pads in reverse order to
3020 avoid quadraticness.
3022 2020-01-10 Martin Jambor <mjambor@suse.cz>
3024 * params.opt (param_ipa_sra_max_replacements): Mark as Optimization.
3025 * ipa-sra.c (pull_accesses_from_callee): New parameter caller, use it
3026 to get param_ipa_sra_max_replacements.
3027 (param_splitting_across_edge): Pass the caller to
3028 pull_accesses_from_callee.
3030 2020-01-10 Martin Jambor <mjambor@suse.cz>
3032 * params.opt (param_ipcp_unit_growth): Mark as Optimization.
3033 * ipa-cp.c (max_new_size): Removed.
3034 (orig_overall_size): New variable.
3035 (get_max_overall_size): New function.
3036 (estimate_local_effects): Use it. Adjust dump.
3037 (decide_about_value): Likewise.
3038 (ipcp_propagate_stage): Do not calculate max_new_size, just store
3039 orig_overall_size. Adjust dump.
3040 (ipa_cp_c_finalize): Clear orig_overall_size instead of max_new_size.
3042 2020-01-10 Martin Jambor <mjambor@suse.cz>
3044 * params.opt (param_ipa_max_agg_items): Mark as Optimization
3045 * ipa-cp.c (merge_agg_lats_step): New parameter max_agg_items, use
3046 instead of param_ipa_max_agg_items.
3047 (merge_aggregate_lattices): Extract param_ipa_max_agg_items from
3048 optimization info for the callee.
3050 2020-01-09 Kwok Cheung Yeung <kcy@codesourcery.com>
3052 * lto-streamer-in.c (input_function): Remove streamed-in inline debug
3053 markers if debug_inline_points is false.
3055 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3057 * config.gcc (aarch64*-*-*): Add aarch64-sve-builtins-sve2.o to
3059 * config/aarch64/t-aarch64 (aarch64-sve-builtins.o): Depend on
3060 aarch64-sve-builtins-base.def, aarch64-sve-builtins-sve2.def and
3061 aarch64-sve-builtins-sve2.h.
3062 (aarch64-sve-builtins-sve2.o): New rule.
3063 * config/aarch64/aarch64.h (AARCH64_ISA_SVE2_AES): New macro.
3064 (AARCH64_ISA_SVE2_BITPERM, AARCH64_ISA_SVE2_SHA3): Likewise.
3065 (AARCH64_ISA_SVE2_SM4, TARGET_SVE2_AES, TARGET_SVE2_BITPERM): Likewise.
3066 (TARGET_SVE2_SHA, TARGET_SVE2_SM4): Likewise.
3067 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
3068 TARGET_SVE2_AES, TARGET_SVE2_BITPERM, TARGET_SVE2_SHA3 and
3070 * config/aarch64/aarch64-sve.md: Update comments with SVE2
3071 instructions that are handled here.
3072 (@cond_asrd<mode>): Generalize to...
3073 (@cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>): ...this.
3074 (*cond_asrd<mode>_2): Generalize to...
3075 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_2): ...this.
3076 (*cond_asrd<mode>_z): Generalize to...
3077 (*cond_<SVE_INT_SHIFT_IMM:sve_int_op><mode>_z): ...this.
3078 * config/aarch64/aarch64.md (UNSPEC_LDNT1_GATHER): New unspec.
3079 (UNSPEC_STNT1_SCATTER, UNSPEC_WHILEGE, UNSPEC_WHILEGT): Likewise.
3080 (UNSPEC_WHILEHI, UNSPEC_WHILEHS): Likewise.
3081 * config/aarch64/aarch64-sve2.md (@aarch64_gather_ldnt<mode>): New
3083 (@aarch64_gather_ldnt_<ANY_EXTEND:optab><SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3084 (@aarch64_scatter_stnt<mode>): Likewise.
3085 (@aarch64_scatter_stnt_<SVE_FULL_SDI:mode><SVE_PARTIAL_I:mode>)
3086 (@aarch64_mul_lane_<mode>): Likewise.
3087 (@aarch64_sve_suqadd<mode>_const): Likewise.
3088 (*<sur>h<addsub><mode>): Generalize to...
3089 (@aarch64_pred_<SVE2_COND_INT_BINARY_REV:sve_int_op><mode>): ...this
3091 (@cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>): New expander.
3092 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_2): New pattern.
3093 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_3): Likewise.
3094 (*cond_<SVE2_COND_INT_BINARY:sve_int_op><mode>_any): Likewise.
3095 (*cond_<SVE2_COND_INT_BINARY_NOREV:sve_int_op><mode>_z): Likewise.
3096 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op><mode>):: Likewise.
3097 (@aarch64_sve_<SVE2_INT_BINARY:sve_int_op>_lane_<mode>): Likewise.
3098 (@aarch64_pred_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): Likewise.
3099 (@cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>): New expander.
3100 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_2): New pattern.
3101 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_3): Likewise.
3102 (*cond_<SVE2_COND_INT_SHIFT:sve_int_op><mode>_any): Likewise.
3103 (@aarch64_sve_<SVE2_INT_TERNARY:sve_int_op><mode>): Likewise.
3104 (@aarch64_sve_<SVE2_INT_TERNARY_LANE:sve_int_op>_lane_<mode>)
3105 (@aarch64_sve_add_mul_lane_<mode>): Likewise.
3106 (@aarch64_sve_sub_mul_lane_<mode>): Likewise.
3107 (@aarch64_sve2_xar<mode>): Likewise.
3108 (@aarch64_sve2_bcax<mode>): Likewise.
3109 (*aarch64_sve2_eor3<mode>): Rename to...
3110 (@aarch64_sve2_eor3<mode>): ...this.
3111 (@aarch64_sve2_bsl<mode>): New expander.
3112 (@aarch64_sve2_nbsl<mode>): Likewise.
3113 (@aarch64_sve2_bsl1n<mode>): Likewise.
3114 (@aarch64_sve2_bsl2n<mode>): Likewise.
3115 (@aarch64_sve_add_<SHIFTRT:sve_int_op><mode>): Likewise.
3116 (*aarch64_sve2_sra<mode>): Add MOVPRFX support.
3117 (@aarch64_sve_add_<VRSHR_N:sve_int_op><mode>): New pattern.
3118 (@aarch64_sve_<SVE2_INT_SHIFT_INSERT:sve_int_op><mode>): Likewise.
3119 (@aarch64_sve2_<USMAX:su>aba<mode>): New expander.
3120 (*aarch64_sve2_<USMAX:su>aba<mode>): New pattern.
3121 (@aarch64_sve_<SVE2_INT_BINARY_WIDE:sve_int_op><mode>): Likewise.
3122 (<su>mull<bt><Vwide>): Generalize to...
3123 (@aarch64_sve_<SVE2_INT_BINARY_LONG:sve_int_op><mode>): ...this new
3125 (@aarch64_sve_<SVE2_INT_BINARY_LONG_lANE:sve_int_op>_lane_<mode>)
3126 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_LONG:sve_int_op><mode>)
3127 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG:sve_int_op><mode>)
3128 (@aarch64_sve_add_<SVE2_INT_ADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3129 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG:sve_int_op><mode>)
3130 (@aarch64_sve_qadd_<SVE2_INT_QADD_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3131 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG:sve_int_op><mode>)
3132 (@aarch64_sve_sub_<SVE2_INT_SUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3133 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG:sve_int_op><mode>)
3134 (@aarch64_sve_qsub_<SVE2_INT_QSUB_BINARY_LONG_LANE:sve_int_op>_lane_<mode>)
3135 (@aarch64_sve_<SVE2_FP_TERNARY_LONG:sve_fp_op><mode>): New patterns.
3136 (@aarch64_<SVE2_FP_TERNARY_LONG_LANE:sve_fp_op>_lane_<mode>)
3137 (@aarch64_sve_<SVE2_INT_UNARY_NARROWB:sve_int_op><mode>): Likewise.
3138 (@aarch64_sve_<SVE2_INT_UNARY_NARROWT:sve_int_op><mode>): Likewise.
3139 (@aarch64_sve_<SVE2_INT_BINARY_NARROWB:sve_int_op><mode>): Likewise.
3140 (@aarch64_sve_<SVE2_INT_BINARY_NARROWT:sve_int_op><mode>): Likewise.
3141 (<SHRNB:r>shrnb<mode>): Generalize to...
3142 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWB:sve_int_op><mode>): ...this
3144 (<SHRNT:r>shrnt<mode>): Generalize to...
3145 (@aarch64_sve_<SVE2_INT_SHIFT_IMM_NARROWT:sve_int_op><mode>): ...this
3147 (@aarch64_pred_<SVE2_INT_BINARY_PAIR:sve_int_op><mode>): New pattern.
3148 (@aarch64_pred_<SVE2_FP_BINARY_PAIR:sve_fp_op><mode>): Likewise.
3149 (@cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>): New expander.
3150 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_2): New pattern.
3151 (*cond_<SVE2_INT_BINARY_PAIR_LONG:sve_int_op><mode>_z): Likewise.
3152 (@aarch64_sve_<SVE2_INT_CADD:optab><mode>): Likewise.
3153 (@aarch64_sve_<SVE2_INT_CMLA:optab><mode>): Likewise.
3154 (@aarch64_<SVE2_INT_CMLA:optab>_lane_<mode>): Likewise.
3155 (@aarch64_sve_<SVE2_INT_CDOT:optab><mode>): Likewise.
3156 (@aarch64_<SVE2_INT_CDOT:optab>_lane_<mode>): Likewise.
3157 (@aarch64_pred_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): Likewise.
3158 (@cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New expander.
3159 (*cond_<SVE2_COND_FP_UNARY_LONG:sve_fp_op><mode>): New pattern.
3160 (@aarch64_sve2_cvtnt<mode>): Likewise.
3161 (@aarch64_pred_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): Likewise.
3162 (@cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>): New expander.
3163 (*cond_<SVE2_COND_FP_UNARY_NARROWB:sve_fp_op><mode>_any): New pattern.
3164 (@aarch64_sve2_cvtxnt<mode>): Likewise.
3165 (@aarch64_pred_<SVE2_U32_UNARY:sve_int_op><mode>): Likewise.
3166 (@cond_<SVE2_U32_UNARY:sve_int_op><mode>): New expander.
3167 (*cond_<SVE2_U32_UNARY:sve_int_op><mode>): New pattern.
3168 (@aarch64_pred_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): Likewise.
3169 (@cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New expander.
3170 (*cond_<SVE2_COND_INT_UNARY_FP:sve_fp_op><mode>): New pattern.
3171 (@aarch64_sve2_pmul<mode>): Likewise.
3172 (@aarch64_sve_<SVE2_PMULL:optab><mode>): Likewise.
3173 (@aarch64_sve_<SVE2_PMULL_PAIR:optab><mode>): Likewise.
3174 (@aarch64_sve2_tbl2<mode>): Likewise.
3175 (@aarch64_sve2_tbx<mode>): Likewise.
3176 (@aarch64_sve_<SVE2_INT_BITPERM:sve_int_op><mode>): Likewise.
3177 (@aarch64_sve2_histcnt<mode>): Likewise.
3178 (@aarch64_sve2_histseg<mode>): Likewise.
3179 (@aarch64_pred_<SVE2_MATCH:sve_int_op><mode>): Likewise.
3180 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_cc): Likewise.
3181 (*aarch64_pred_<SVE2_MATCH:sve_int_op><mode>_ptest): Likewise.
3182 (aarch64_sve2_aes<CRYPTO_AES:aes_op>): Likewise.
3183 (aarch64_sve2_aes<CRYPTO_AESMC:aesmc_op>): Likewise.
3184 (*aarch64_sve2_aese_fused, *aarch64_sve2_aesd_fused): Likewise.
3185 (aarch64_sve2_rax1, aarch64_sve2_sm4e, aarch64_sve2_sm4ekey): Likewise.
3186 (<su>mulh<r>s<mode>3): Update after above pattern name changes.
3187 * config/aarch64/iterators.md (VNx16QI_ONLY, VNx4SF_ONLY)
3188 (SVE_STRUCT2, SVE_FULL_BHI, SVE_FULL_HSI, SVE_FULL_HDI)
3189 (SVE2_PMULL_PAIR_I): New mode iterators.
3190 (UNSPEC_ADCLB, UNSPEC_ADCLT, UNSPEC_ADDHNB, UNSPEC_ADDHNT, UNSPEC_BDEP)
3191 (UNSPEC_BEXT, UNSPEC_BGRP, UNSPEC_CADD90, UNSPEC_CADD270, UNSPEC_CDOT)
3192 (UNSPEC_CDOT90, UNSPEC_CDOT180, UNSPEC_CDOT270, UNSPEC_CMLA)
3193 (UNSPEC_CMLA90, UNSPEC_CMLA180, UNSPEC_CMLA270, UNSPEC_COND_FCVTLT)
3194 (UNSPEC_COND_FCVTNT, UNSPEC_COND_FCVTX, UNSPEC_COND_FCVTXNT)
3195 (UNSPEC_COND_FLOGB, UNSPEC_EORBT, UNSPEC_EORTB, UNSPEC_FADDP)
3196 (UNSPEC_FMAXP, UNSPEC_FMAXNMP, UNSPEC_FMLALB, UNSPEC_FMLALT)
3197 (UNSPEC_FMLSLB, UNSPEC_FMLSLT, UNSPEC_FMINP, UNSPEC_FMINNMP)
3198 (UNSPEC_HISTCNT, UNSPEC_HISTSEG, UNSPEC_MATCH, UNSPEC_NMATCH)
3199 (UNSPEC_PMULLB, UNSPEC_PMULLB_PAIR, UNSPEC_PMULLT, UNSPEC_PMULLT_PAIR)
3200 (UNSPEC_RADDHNB, UNSPEC_RADDHNT, UNSPEC_RSUBHNB, UNSPEC_RSUBHNT)
3201 (UNSPEC_SLI, UNSPEC_SRI, UNSPEC_SABDLB, UNSPEC_SABDLT, UNSPEC_SADDLB)
3202 (UNSPEC_SADDLBT, UNSPEC_SADDLT, UNSPEC_SADDWB, UNSPEC_SADDWT)
3203 (UNSPEC_SBCLB, UNSPEC_SBCLT, UNSPEC_SMAXP, UNSPEC_SMINP)
3204 (UNSPEC_SQCADD90, UNSPEC_SQCADD270, UNSPEC_SQDMULLB, UNSPEC_SQDMULLBT)
3205 (UNSPEC_SQDMULLT, UNSPEC_SQRDCMLAH, UNSPEC_SQRDCMLAH90)
3206 (UNSPEC_SQRDCMLAH180, UNSPEC_SQRDCMLAH270, UNSPEC_SQRSHRNB)
3207 (UNSPEC_SQRSHRNT, UNSPEC_SQRSHRUNB, UNSPEC_SQRSHRUNT, UNSPEC_SQSHRNB)
3208 (UNSPEC_SQSHRNT, UNSPEC_SQSHRUNB, UNSPEC_SQSHRUNT, UNSPEC_SQXTNB)
3209 (UNSPEC_SQXTNT, UNSPEC_SQXTUNB, UNSPEC_SQXTUNT, UNSPEC_SSHLLB)
3210 (UNSPEC_SSHLLT, UNSPEC_SSUBLB, UNSPEC_SSUBLBT, UNSPEC_SSUBLT)
3211 (UNSPEC_SSUBLTB, UNSPEC_SSUBWB, UNSPEC_SSUBWT, UNSPEC_SUBHNB)
3212 (UNSPEC_SUBHNT, UNSPEC_TBL2, UNSPEC_UABDLB, UNSPEC_UABDLT)
3213 (UNSPEC_UADDLB, UNSPEC_UADDLT, UNSPEC_UADDWB, UNSPEC_UADDWT)
3214 (UNSPEC_UMAXP, UNSPEC_UMINP, UNSPEC_UQRSHRNB, UNSPEC_UQRSHRNT)
3215 (UNSPEC_UQSHRNB, UNSPEC_UQSHRNT, UNSPEC_UQXTNB, UNSPEC_UQXTNT)
3216 (UNSPEC_USHLLB, UNSPEC_USHLLT, UNSPEC_USUBLB, UNSPEC_USUBLT)
3217 (UNSPEC_USUBWB, UNSPEC_USUBWT): New unspecs.
3218 (UNSPEC_SMULLB, UNSPEC_SMULLT, UNSPEC_UMULLB, UNSPEC_UMULLT)
3219 (UNSPEC_SMULHS, UNSPEC_SMULHRS, UNSPEC_UMULHS, UNSPEC_UMULHRS)
3220 (UNSPEC_RSHRNB, UNSPEC_RSHRNT, UNSPEC_SHRNB, UNSPEC_SHRNT): Move
3222 (VNARROW, Ventype): New mode attributes.
3223 (Vewtype): Handle VNx2DI. Fix typo in comment.
3224 (VDOUBLE): New mode attribute.
3225 (sve_lane_con): Handle VNx8HI.
3226 (SVE_INT_UNARY): Include ss_abs and ss_neg for TARGET_SVE2.
3227 (SVE_INT_BINARY): Likewise ss_plus, us_plus, ss_minus and us_minus.
3228 (sve_int_op, sve_int_op_rev): Handle the above codes.
3229 (sve_pred_int_rhs2_operand): Likewise.
3230 (MULLBT, SHRNB, SHRNT): Delete.
3231 (SVE_INT_SHIFT_IMM): New int iterator.
3232 (SVE_WHILE): Add UNSPEC_WHILEGE, UNSPEC_WHILEGT, UNSPEC_WHILEHI
3233 and UNSPEC_WHILEHS for TARGET_SVE2.
3234 (SVE2_U32_UNARY, SVE2_INT_UNARY_NARROWB, SVE2_INT_UNARY_NARROWT)
3235 (SVE2_INT_BINARY, SVE2_INT_BINARY_LANE, SVE2_INT_BINARY_LONG)
3236 (SVE2_INT_BINARY_LONG_LANE, SVE2_INT_BINARY_NARROWB)
3237 (SVE2_INT_BINARY_NARROWT, SVE2_INT_BINARY_PAIR, SVE2_FP_BINARY_PAIR)
3238 (SVE2_INT_BINARY_PAIR_LONG, SVE2_INT_BINARY_WIDE): New int iterators.
3239 (SVE2_INT_SHIFT_IMM_LONG, SVE2_INT_SHIFT_IMM_NARROWB): Likewise.
3240 (SVE2_INT_SHIFT_IMM_NARROWT, SVE2_INT_SHIFT_INSERT, SVE2_INT_CADD)
3241 (SVE2_INT_BITPERM, SVE2_INT_TERNARY, SVE2_INT_TERNARY_LANE): Likewise.
3242 (SVE2_FP_TERNARY_LONG, SVE2_FP_TERNARY_LONG_LANE, SVE2_INT_CMLA)
3243 (SVE2_INT_CDOT, SVE2_INT_ADD_BINARY_LONG, SVE2_INT_QADD_BINARY_LONG)
3244 (SVE2_INT_SUB_BINARY_LONG, SVE2_INT_QSUB_BINARY_LONG): Likewise.
3245 (SVE2_INT_ADD_BINARY_LONG_LANE, SVE2_INT_QADD_BINARY_LONG_LANE)
3246 (SVE2_INT_SUB_BINARY_LONG_LANE, SVE2_INT_QSUB_BINARY_LONG_LANE)
3247 (SVE2_COND_INT_UNARY_FP, SVE2_COND_FP_UNARY_LONG): Likewise.
3248 (SVE2_COND_FP_UNARY_NARROWB, SVE2_COND_INT_BINARY): Likewise.
3249 (SVE2_COND_INT_BINARY_NOREV, SVE2_COND_INT_BINARY_REV): Likewise.
3250 (SVE2_COND_INT_SHIFT, SVE2_MATCH, SVE2_PMULL): Likewise.
3251 (optab): Handle the new unspecs.
3252 (su, r): Remove entries for UNSPEC_SHRNB, UNSPEC_SHRNT, UNSPEC_RSHRNB
3254 (lr): Handle the new unspecs.
3256 (cmp_op, while_optab_cmp, sve_int_op): Handle the new unspecs.
3257 (sve_int_op_rev, sve_int_add_op, sve_int_qadd_op, sve_int_sub_op)
3258 (sve_int_qsub_op): New int attributes.
3259 (sve_fp_op, rot): Handle the new unspecs.
3260 * config/aarch64/aarch64-sve-builtins.h
3261 (function_resolver::require_matching_pointer_type): Declare.
3262 (function_resolver::resolve_unary): Add an optional boolean argument.
3263 (function_resolver::finish_opt_n_resolution): Add an optional
3264 type_suffix_index argument.
3265 (gimple_folder::redirect_call): Declare.
3266 (gimple_expander::prepare_gather_address_operands): Add an optional
3268 * config/aarch64/aarch64-sve-builtins.cc: Include
3269 aarch64-sve-builtins-sve2.h.
3270 (TYPES_b_unsigned, TYPES_b_integer, TYPES_bh_integer): New macros.
3271 (TYPES_bs_unsigned, TYPES_hs_signed, TYPES_hs_integer): Likewise.
3272 (TYPES_hd_unsigned, TYPES_hsd_signed): Likewise.
3273 (TYPES_hsd_integer): Use TYPES_hsd_signed.
3274 (TYPES_s_float_hsd_integer, TYPES_s_float_sd_integer): New macros.
3275 (TYPES_s_unsigned): Likewise.
3276 (TYPES_s_integer): Use TYPES_s_unsigned.
3277 (TYPES_sd_signed, TYPES_sd_unsigned): New macros.
3278 (TYPES_sd_integer): Use them.
3279 (TYPES_d_unsigned): New macro.
3280 (TYPES_d_integer): Use it.
3281 (TYPES_d_data, TYPES_cvt_long, TYPES_cvt_narrow_s): New macros.
3282 (TYPES_cvt_narrow): Likewise.
3283 (DEF_SVE_TYPES_ARRAY): Include the new types macros above.
3284 (preds_mx): New variable.
3285 (function_builder::add_overloaded_function): Allow the new feature
3286 set to be more restrictive than the original one.
3287 (function_resolver::infer_pointer_type): Remove qualifiers from
3288 the pointer type before printing it.
3289 (function_resolver::require_matching_pointer_type): New function.
3290 (function_resolver::resolve_sv_displacement): Handle functions
3291 that don't support 32-bit vector indices or svint32_t vector offsets.
3292 (function_resolver::finish_opt_n_resolution): Take the inferred type
3293 as a separate argument.
3294 (function_resolver::resolve_unary): Optionally treat all forms in
3295 the same way as normal merging functions.
3296 (gimple_folder::redirect_call): New function.
3297 (function_expander::prepare_gather_address_operands): Add an argument
3298 that says whether scaled forms are available. If they aren't,
3299 handle scaling of vector indices and don't add the extension and
3301 (function_expander::map_to_unspecs): If aarch64_sve isn't available,
3302 fall back to using cond_* instead.
3303 * config/aarch64/aarch64-sve-builtins-functions.h (rtx_code_function):
3304 Split out the member variables into...
3305 (rtx_code_function_base): ...this new base class.
3306 (rtx_code_function_rotated): Inherit rtx_code_function_base.
3307 (unspec_based_function): Split out the member variables into...
3308 (unspec_based_function_base): ...this new base class.
3309 (unspec_based_function_rotated): Inherit unspec_based_function_base.
3310 (unspec_based_function_exact_insn): New class.
3311 (unspec_based_add_function, unspec_based_add_lane_function)
3312 (unspec_based_lane_function, unspec_based_pred_function)
3313 (unspec_based_qadd_function, unspec_based_qadd_lane_function)
3314 (unspec_based_qsub_function, unspec_based_qsub_lane_function)
3315 (unspec_based_sub_function, unspec_based_sub_lane_function): New
3317 (unspec_based_fused_function): New class.
3318 (unspec_based_mla_function, unspec_based_mls_function): New typedefs.
3319 (unspec_based_fused_lane_function): New class.
3320 (unspec_based_mla_lane_function, unspec_based_mls_lane_function): New
3322 (CODE_FOR_MODE1): New macro.
3323 (fixed_insn_function): New class.
3324 (while_comparison): Likewise.
3325 * config/aarch64/aarch64-sve-builtins-shapes.h (binary_long_lane)
3326 (binary_long_opt_n, binary_narrowb_opt_n, binary_narrowt_opt_n)
3327 (binary_to_uint, binary_wide, binary_wide_opt_n, compare, compare_ptr)
3328 (load_ext_gather_index_restricted, load_ext_gather_offset_restricted)
3329 (load_gather_sv_restricted, shift_left_imm_long): Declare.
3330 (shift_left_imm_to_uint, shift_right_imm_narrowb): Likewise.
3331 (shift_right_imm_narrowt, shift_right_imm_narrowb_to_uint): Likewise.
3332 (shift_right_imm_narrowt_to_uint, store_scatter_index_restricted)
3333 (store_scatter_offset_restricted, tbl_tuple, ternary_long_lane)
3334 (ternary_long_opt_n, ternary_qq_lane_rotate, ternary_qq_rotate)
3335 (ternary_shift_left_imm, ternary_shift_right_imm, ternary_uint)
3336 (unary_convert_narrowt, unary_long, unary_narrowb, unary_narrowt)
3337 (unary_narrowb_to_uint, unary_narrowt_to_uint, unary_to_int): Likewise.
3338 * config/aarch64/aarch64-sve-builtins-shapes.cc (apply_predication):
3339 Also add an initial argument for unary_convert_narrowt, regardless
3340 of the predication type.
3341 (build_32_64): Allow loads and stores to specify MODE_none.
3342 (build_sv_index64, build_sv_uint_offset): New functions.
3343 (long_type_suffix): New function.
3344 (binary_imm_narrowb_base, binary_imm_narrowt_base): New classes.
3345 (binary_imm_long_base, load_gather_sv_base): Likewise.
3346 (shift_right_imm_narrow_wrapper, ternary_shift_imm_base): Likewise.
3347 (ternary_resize2_opt_n_base, ternary_resize2_lane_base): Likewise.
3348 (unary_narrowb_base, unary_narrowt_base): Likewise.
3349 (binary_long_lane_def, binary_long_lane): New shape.
3350 (binary_long_opt_n_def, binary_long_opt_n): Likewise.
3351 (binary_narrowb_opt_n_def, binary_narrowb_opt_n): Likewise.
3352 (binary_narrowt_opt_n_def, binary_narrowt_opt_n): Likewise.
3353 (binary_to_uint_def, binary_to_uint): Likewise.
3354 (binary_wide_def, binary_wide): Likewise.
3355 (binary_wide_opt_n_def, binary_wide_opt_n): Likewise.
3356 (compare_def, compare): Likewise.
3357 (compare_ptr_def, compare_ptr): Likewise.
3358 (load_ext_gather_index_restricted_def,
3359 load_ext_gather_index_restricted): Likewise.
3360 (load_ext_gather_offset_restricted_def,
3361 load_ext_gather_offset_restricted): Likewise.
3362 (load_gather_sv_def): Inherit from load_gather_sv_base.
3363 (load_gather_sv_restricted_def, load_gather_sv_restricted): New shape.
3364 (shift_left_imm_def, shift_left_imm): Likewise.
3365 (shift_left_imm_long_def, shift_left_imm_long): Likewise.
3366 (shift_left_imm_to_uint_def, shift_left_imm_to_uint): Likewise.
3367 (store_scatter_index_restricted_def,
3368 store_scatter_index_restricted): Likewise.
3369 (store_scatter_offset_restricted_def,
3370 store_scatter_offset_restricted): Likewise.
3371 (tbl_tuple_def, tbl_tuple): Likewise.
3372 (ternary_long_lane_def, ternary_long_lane): Likewise.
3373 (ternary_long_opt_n_def, ternary_long_opt_n): Likewise.
3374 (ternary_qq_lane_def): Inherit from ternary_resize2_lane_base.
3375 (ternary_qq_lane_rotate_def, ternary_qq_lane_rotate): New shape
3376 (ternary_qq_opt_n_def): Inherit from ternary_resize2_opt_n_base.
3377 (ternary_qq_rotate_def, ternary_qq_rotate): New shape.
3378 (ternary_shift_left_imm_def, ternary_shift_left_imm): Likewise.
3379 (ternary_shift_right_imm_def, ternary_shift_right_imm): Likewise.
3380 (ternary_uint_def, ternary_uint): Likewise.
3381 (unary_convert): Fix typo in comment.
3382 (unary_convert_narrowt_def, unary_convert_narrowt): New shape.
3383 (unary_long_def, unary_long): Likewise.
3384 (unary_narrowb_def, unary_narrowb): Likewise.
3385 (unary_narrowt_def, unary_narrowt): Likewise.
3386 (unary_narrowb_to_uint_def, unary_narrowb_to_uint): Likewise.
3387 (unary_narrowt_to_uint_def, unary_narrowt_to_uint): Likewise.
3388 (unary_to_int_def, unary_to_int): Likewise.
3389 * config/aarch64/aarch64-sve-builtins-base.cc (unspec_cmla)
3390 (unspec_fcmla, unspec_cond_fcmla, expand_mla_mls_lane): New functions.
3391 (svasrd_impl): Delete.
3392 (svcadd_impl::expand): Handle integer operations too.
3393 (svcmla_impl::expand, svcmla_lane::expand): Likewise, using the
3394 new functions to derive the unspec numbers.
3395 (svmla_svmls_lane_impl): Replace with...
3396 (svmla_lane_impl, svmls_lane_impl): ...these new classes. Handle
3397 integer operations too.
3398 (svwhile_impl): Rename to...
3399 (svwhilelx_impl): ...this and inherit from while_comparison.
3400 (svasrd): Use unspec_based_function.
3401 (svmla_lane): Use svmla_lane_impl.
3402 (svmls_lane): Use svmls_lane_impl.
3403 (svrecpe, svrsqrte): Handle unsigned integer operations too.
3404 (svwhilele, svwhilelt): Use svwhilelx_impl.
3405 * config/aarch64/aarch64-sve-builtins-sve2.h: New file.
3406 * config/aarch64/aarch64-sve-builtins-sve2.cc: Likewise.
3407 * config/aarch64/aarch64-sve-builtins-sve2.def: Likewise.
3408 * config/aarch64/aarch64-sve-builtins.def: Include
3409 aarch64-sve-builtins-sve2.def.
3411 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3413 * config/aarch64/aarch64-protos.h (aarch64_sve_arith_immediate_p)
3414 (aarch64_sve_sqadd_sqsub_immediate_p): Add a machine_mode argument.
3415 * config/aarch64/aarch64.c (aarch64_sve_arith_immediate_p)
3416 (aarch64_sve_sqadd_sqsub_immediate_p): Likewise. Handle scalar
3417 immediates as well as vector ones.
3418 * config/aarch64/predicates.md (aarch64_sve_arith_immediate)
3419 (aarch64_sve_sub_arith_immediate, aarch64_sve_qadd_immediate)
3420 (aarch64_sve_qsub_immediate): Update calls accordingly.
3422 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3424 * config/aarch64/aarch64-sve2.md: Add banner comments.
3425 (<su>mulh<r>s<mode>3): Move further up file.
3426 (<su>mull<bt><Vwide>, <r>shrnb<mode>, <r>shrnt<mode>)
3427 (*aarch64_sve2_sra<mode>): Move further down file.
3428 * config/aarch64/t-aarch64 (s-check-sve-md): Check aarch64-sve2.md too.
3430 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3432 * config/aarch64/iterators.md (SVE_WHILE): Add UNSPEC_WHILERW
3434 (while_optab_cmp): Handle them.
3435 * config/aarch64/aarch64-sve.md
3436 (*while_<while_optab_cmp><GPI:mode><PRED_ALL:mode>_ptest): Make public
3437 and add a "@" marker.
3438 * config/aarch64/aarch64-sve2.md (check_<raw_war>_ptrs<mode>): Use it
3439 instead of gen_aarch64_sve2_while_ptest.
3440 (@aarch64_sve2_while<cmp_op><GPI:mode><PRED_ALL:mode>_ptest): Delete.
3442 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3444 * config/aarch64/aarch64.md (UNSPEC_WHILE_LE): Rename to...
3445 (UNSPEC_WHILELE): ...this.
3446 (UNSPEC_WHILE_LO): Rename to...
3447 (UNSPEC_WHILELO): ...this.
3448 (UNSPEC_WHILE_LS): Rename to...
3449 (UNSPEC_WHILELS): ...this.
3450 (UNSPEC_WHILE_LT): Rename to...
3451 (UNSPEC_WHILELT): ...this.
3452 * config/aarch64/iterators.md (SVE_WHILE): Update accordingly.
3453 (cmp_op, while_optab_cmp): Likewise.
3454 * config/aarch64/aarch64.c (aarch64_sve_move_pred_via_while): Likewise.
3455 * config/aarch64/aarch64-sve-builtins-base.cc (svwhilele): Likewise.
3456 (svwhilelt): Likewise.
3458 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3460 * config/aarch64/aarch64-sve-builtins-shapes.h (unary_count): Delete.
3461 (unary_to_uint): Define.
3462 * config/aarch64/aarch64-sve-builtins-shapes.cc (unary_count_def)
3463 (unary_count): Rename to...
3464 (unary_to_uint_def, unary_to_uint): ...this.
3465 * config/aarch64/aarch64-sve-builtins-base.def: Update accordingly.
3467 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3469 * config/aarch64/aarch64-sve-builtins-functions.h
3470 (code_for_mode_function): New class.
3471 (CODE_FOR_MODE0, QUIET_CODE_FOR_MODE0): New macros.
3472 * config/aarch64/aarch64-sve-builtins-base.cc (svcompact_impl)
3473 (svext_impl, svmul_lane_impl, svsplice_impl, svtmad_impl): Delete.
3474 (svcompact, svext, svsplice): Use QUIET_CODE_FOR_MODE0.
3475 (svmul_lane, svtmad): Use CODE_FOR_MODE0.
3477 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3479 * config/aarch64/iterators.md (addsub): New code attribute.
3480 * config/aarch64/aarch64-simd.md (aarch64_<su_optab><optab><mode>):
3482 (aarch64_<su_optab>q<addsub><mode>): ...this, making the same change
3483 in the asm string and attributes. Fix indentation.
3484 * config/aarch64/aarch64-sve.md (@aarch64_<su_optab><optab><mode>):
3486 (@aarch64_sve_<optab><mode>): ...this.
3487 * config/aarch64/aarch64-sve-builtins.h
3488 (function_expander::expand_signed_unpred_op): Delete.
3489 * config/aarch64/aarch64-sve-builtins.cc
3490 (function_expander::expand_signed_unpred_op): Likewise.
3491 (function_expander::map_to_rtx_codes): If the optab isn't defined,
3492 try using code_for_aarch64_sve instead.
3493 * config/aarch64/aarch64-sve-builtins-base.cc (svqadd_impl): Delete.
3494 (svqsub_impl): Likewise.
3495 (svqadd, svqsub): Use rtx_code_function instead.
3497 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3499 * config/aarch64/iterators.md (SRHSUB, URHSUB): Delete.
3500 (HADDSUB, sur, addsub): Remove them.
3502 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3504 * tree-nrv.c (pass_return_slot::execute): Handle all internal
3505 functions the same way, rather than singling out those that
3506 aren't mapped directly to optabs.
3508 2020-01-09 Richard Sandiford <richard.sandiford@arm.com>
3510 * target.def (compatible_vector_types_p): New target hook.
3511 * hooks.h (hook_bool_const_tree_const_tree_true): Declare.
3512 * hooks.c (hook_bool_const_tree_const_tree_true): New function.
3513 * doc/tm.texi.in (TARGET_COMPATIBLE_VECTOR_TYPES_P): New hook.
3514 * doc/tm.texi: Regenerate.
3515 * gimple-expr.c: Include target.h.
3516 (useless_type_conversion_p): Use targetm.compatible_vector_types_p.
3517 * config/aarch64/aarch64.c (aarch64_compatible_vector_types_p): New
3519 (TARGET_COMPATIBLE_VECTOR_TYPES_P): Define.
3520 * config/aarch64/aarch64-sve-builtins.cc (gimple_folder::convert_pred):
3521 Use the original predicate if it already has a suitable type.
3523 2020-01-09 Martin Jambor <mjambor@suse.cz>
3525 * cgraph.h (cgraph_edge): Make remove, set_call_stmt, make_direct,
3526 resolve_speculation and redirect_call_stmt_to_callee static. Change
3527 return type of set_call_stmt to cgraph_edge *.
3528 * auto-profile.c (afdo_indirect_call): Adjust call to
3529 redirect_call_stmt_to_callee.
3530 * cgraph.c (cgraph_edge::set_call_stmt): Make return cgraph-edge *,
3531 make the this pointer explicit, adjust self-recursive calls and the
3532 call top make_direct. Return the resulting edge.
3533 (cgraph_edge::remove): Make this pointer explicit.
3534 (cgraph_edge::resolve_speculation): Likewise, adjust call to remove.
3535 (cgraph_edge::make_direct): Likewise, adjust call to
3536 resolve_speculation.
3537 (cgraph_edge::redirect_call_stmt_to_callee): Likewise, also adjust
3538 call to set_call_stmt.
3539 (cgraph_update_edges_for_call_stmt_node): Update call to
3540 set_call_stmt and remove.
3541 * cgraphclones.c (cgraph_node::set_call_stmt_including_clones):
3542 Renamed edge to master_edge. Adjusted calls to set_call_stmt.
3543 (cgraph_node::create_edge_including_clones): Moved "first" definition
3544 of edge to the block where it was used. Adjusted calls to
3546 (cgraph_node::remove_symbol_and_inline_clones): Adjust call to
3547 cgraph_edge::remove.
3548 * cgraphunit.c (walk_polymorphic_call_targets): Adjusted calls to
3549 make_direct and redirect_call_stmt_to_callee.
3550 * ipa-fnsummary.c (redirect_to_unreachable): Adjust calls to
3551 resolve_speculation and make_direct.
3552 * ipa-inline-transform.c (inline_transform): Adjust call to
3553 redirect_call_stmt_to_callee.
3554 (check_speculations_1):: Adjust call to resolve_speculation.
3555 * ipa-inline.c (resolve_noninline_speculation): Adjust call to
3556 resolve-speculation.
3557 (inline_small_functions): Adjust call to resolve_speculation.
3558 (ipa_inline): Likewise.
3559 * ipa-prop.c (ipa_make_edge_direct_to_target): Adjust call to
3561 * ipa-visibility.c (function_and_variable_visibility): Make iteration
3562 safe with regards to edge removal, adjust calls to
3563 redirect_call_stmt_to_callee.
3564 * ipa.c (walk_polymorphic_call_targets): Adjust calls to make_direct
3565 and redirect_call_stmt_to_callee.
3566 * multiple_target.c (create_dispatcher_calls): Adjust call to
3567 redirect_call_stmt_to_callee
3568 (redirect_to_specific_clone): Likewise.
3569 * tree-cfgcleanup.c (delete_unreachable_blocks_update_callgraph):
3570 Adjust calls to cgraph_edge::remove.
3571 * tree-inline.c (copy_bb): Adjust call to set_call_stmt.
3572 (redirect_all_calls): Adjust call to redirect_call_stmt_to_callee.
3573 (expand_call_inline): Adjust call to cgraph_edge::remove.
3575 2020-01-09 Martin Liska <mliska@suse.cz>
3577 * params.opt: Set Optimization for
3578 param_max_speculative_devirt_maydefs.
3580 2020-01-09 Martin Sebor <msebor@redhat.com>
3584 * builtins.c (compute_objsize): Avoid handling MEM_REFs of vector type.
3586 2020-01-09 Martin Liska <mliska@suse.cz>
3588 * auto-profile.c (auto_profile): Use opt_for_fn
3590 * ipa-cp.c (ipcp_lattice::add_value): Likewise.
3591 (propagate_vals_across_arith_jfunc): Likewise.
3592 (hint_time_bonus): Likewise.
3593 (incorporate_penalties): Likewise.
3594 (good_cloning_opportunity_p): Likewise.
3595 (perform_estimation_of_a_value): Likewise.
3596 (estimate_local_effects): Likewise.
3597 (ipcp_propagate_stage): Likewise.
3598 * ipa-fnsummary.c (decompose_param_expr): Likewise.
3599 (set_switch_stmt_execution_predicate): Likewise.
3600 (analyze_function_body): Likewise.
3601 * ipa-inline-analysis.c (offline_size): Likewise.
3602 * ipa-inline.c (early_inliner): Likewise.
3603 * ipa-prop.c (ipa_analyze_node): Likewise.
3604 (ipcp_transform_function): Likewise.
3605 * ipa-sra.c (process_scan_results): Likewise.
3606 (ipa_sra_summarize_function): Likewise.
3607 * params.opt: Rename ipcp-unit-growth to
3608 ipa-cp-unit-growth. Add Optimization for various
3609 IPA-related parameters.
3611 2020-01-09 Richard Biener <rguenther@suse.de>
3614 * gimplify.c (gimplify_expr): Deal with NOP definitions.
3616 2020-01-09 Richard Biener <rguenther@suse.de>
3618 PR tree-optimization/93040
3619 * gimple-ssa-store-merging.c (find_bswap_or_nop): Raise search limit.
3621 2020-01-09 Georg-Johann Lay <avr@gjlay.de>
3623 * common/config/avr/avr-common.c (avr_option_optimization_table)
3624 [OPT_LEVELS_1_PLUS]: Set -fsplit-wide-types-early.
3626 2020-01-09 Martin Liska <mliska@suse.cz>
3628 * cgraphclones.c (symbol_table::materialize_all_clones):
3629 Use cgraph_node::dump_name.
3631 2020-01-09 Jakub Jelinek <jakub@redhat.com>
3634 * config/riscv/riscv.c (riscv_print_operand_reloc): Use
3635 output_operand_lossage instead of gcc_unreachable.
3636 * doc/md.texi (riscv f constraint): Fix typo.
3639 * config/i386/i386.md (subv<mode>4): Use SWIDWI iterator instead of
3640 SWI. Use <general_hilo_operand> instead of <general_operand>. Use
3641 CONST_SCALAR_INT_P instead of CONST_INT_P.
3642 (*subv<mode>4_1): Rename to ...
3643 (subv<mode>4_1): ... this.
3644 (*subv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
3645 define_insn_and_split patterns.
3646 (*subv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
3649 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3651 * vec.c (class selftest::count_dtor): New class.
3652 (selftest::test_auto_delete_vec): New test.
3653 (selftest::vec_c_tests): Call it.
3654 * vec.h (class auto_delete_vec): New class template.
3655 (auto_delete_vec<T>::~auto_delete_vec): New dtor.
3657 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3659 * sbitmap.h (auto_sbitmap): Add operator const_sbitmap.
3661 2020-01-08 Jim Wilson <jimw@sifive.com>
3663 * config/riscv/riscv.c (riscv_legitimize_tls_address): Ifdef out
3664 use of TLS_MODEL_LOCAL_EXEC when not pic.
3666 2020-01-08 David Malcolm <dmalcolm@redhat.com>
3668 * hash-map-tests.c (selftest::test_map_of_strings_to_int): Fix
3671 2020-01-08 Jakub Jelinek <jakub@redhat.com>
3674 * config/i386/i386.md (*stack_protect_set_2_<mode> peephole2,
3675 *stack_protect_set_3 peephole2): Also check that the second
3676 insns source is general_operand.
3679 * config/i386/i386.md (addcarry<mode>_0): Use nonimmediate_operand
3680 predicate for output operand instead of register_operand.
3681 (addcarry<mode>, addcarry<mode>_1): Likewise. Add alternative with
3682 memory destination and non-memory operands[2].
3684 2020-01-08 Martin Liska <mliska@suse.cz>
3686 * cgraph.c (cgraph_node::dump): Use ::dump_name or
3687 ::dump_asm_name instead of (::name or ::asm_name).
3688 * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
3689 * cgraphunit.c (walk_polymorphic_call_targets): Likewise.
3690 (analyze_functions): Likewise.
3691 (expand_all_functions): Likewise.
3692 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
3693 (propagate_bits_across_jump_function): Likewise.
3694 (dump_profile_updates): Likewise.
3695 (ipcp_store_bits_results): Likewise.
3696 (ipcp_store_vr_results): Likewise.
3697 * ipa-devirt.c (dump_targets): Likewise.
3698 * ipa-fnsummary.c (analyze_function_body): Likewise.
3699 * ipa-hsa.c (check_warn_node_versionable): Likewise.
3700 (process_hsa_functions): Likewise.
3701 * ipa-icf.c (sem_item_optimizer::merge_classes): Likewise.
3702 (set_alias_uids): Likewise.
3703 * ipa-inline-transform.c (save_inline_function_body): Likewise.
3704 * ipa-inline.c (recursive_inlining): Likewise.
3705 (inline_to_all_callers_1): Likewise.
3706 (ipa_inline): Likewise.
3707 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
3708 (ipa_propagate_frequency): Likewise.
3709 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
3710 (remove_described_reference): Likewise.
3711 * ipa-pure-const.c (worse_state): Likewise.
3712 (check_retval_uses): Likewise.
3713 (analyze_function): Likewise.
3714 (propagate_pure_const): Likewise.
3715 (propagate_nothrow): Likewise.
3716 (dump_malloc_lattice): Likewise.
3717 (propagate_malloc): Likewise.
3718 (pass_local_pure_const::execute): Likewise.
3719 * ipa-visibility.c (optimize_weakref): Likewise.
3720 (function_and_variable_visibility): Likewise.
3721 * ipa.c (symbol_table::remove_unreachable_nodes): Likewise.
3722 (ipa_discover_variable_flags): Likewise.
3723 * lto-streamer-out.c (output_function): Likewise.
3724 (output_constructor): Likewise.
3725 * tree-inline.c (copy_bb): Likewise.
3726 * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
3727 * varpool.c (symbol_table::remove_unreferenced_decls): Likewise.
3729 2020-01-08 Richard Biener <rguenther@suse.de>
3732 * tree-eh.c (sink_clobbers): Update virtual operands for
3733 the first and last stmt only. Add a dry-run capability.
3734 (pass_lower_eh_dispatch::execute): Perform clobber sinking
3735 after CFG manipulations and in RPO order to catch all
3736 secondary opportunities reliably.
3738 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3741 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3743 2019-01-08 Richard Biener <rguenther@suse.de>
3746 * gimple-fold.c (rewrite_to_defined_overflow): Mark stmt modified.
3747 * tree-ssa-loop-im.c (move_computations_worker): Properly adjust
3748 virtual operand, also updating SSA use.
3749 * gimple-loop-interchange.cc (loop_cand::undo_simple_reduction):
3750 Update stmt after resetting virtual operand.
3751 (tree_loop_interchange::move_code_to_inner_loop): Likewise.
3752 * gimple-iterator.c (gsi_remove): When not removing the stmt
3753 permanently do not delink immediate uses or mark the stmt modified.
3755 2020-01-08 Martin Liska <mliska@suse.cz>
3757 * ipa-fnsummary.c (dump_ipa_call_summary): Use symtab_node::dump_name.
3758 (ipa_call_context::estimate_size_and_time): Likewise.
3759 (inline_analyze_function): Likewise.
3761 2020-01-08 Martin Liska <mliska@suse.cz>
3763 * cgraph.c (cgraph_node::dump): Use systematically
3766 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3768 Add -nodevicespecs option for avr.
3771 * config/avr/avr.opt (-nodevicespecs): New driver option.
3772 * config/avr/driver-avr.c (avr_devicespecs_file): Only issue
3773 "-specs=device-specs/..." if that option is not set.
3774 * doc/invoke.texi (AVR Options) <-nodevicespecs>: Document.
3776 2020-01-08 Georg-Johann Lay <avr@gjlay.de>
3778 Implement 64-bit double functions for avr.
3781 * config.gcc (tm_defines) [target=avr]: Support --with-libf7,
3782 --with-double-comparison.
3783 * doc/install.texi: Document them.
3784 * config/avr/avr-c.c (avr_cpu_cpp_builtins)
3785 <WITH_LIBF7_LIBGCC, WITH_LIBF7_MATH, WITH_LIBF7_MATH_SYMBOLS>
3786 <WITH_DOUBLE_COMPARISON>: New built-in defines.
3787 * doc/invoke.texi (AVR Built-in Macros): Document them.
3788 * config/avr/avr-protos.h (avr_float_lib_compare_returns_bool): New.
3789 * config/avr/avr.c (avr_float_lib_compare_returns_bool): New function.
3790 * config/avr/avr.h (FLOAT_LIB_COMPARE_RETURNS_BOOL): New macro.
3792 2020-01-08 Richard Earnshaw <rearnsha@arm.com>
3795 * config/arm/t-multilib (MULTILIB_MATCHES): Add rules to match
3796 armv7-a{+mp,+sec,+mp+sec} to appropriate armv7 multilib variants
3797 when only building rm-profile multilibs.
3799 2020-01-08 Feng Xue <fxue@os.amperecomputing.com>
3802 * ipa-cp.c (self_recursively_generated_p): Find matched aggregate
3803 lattice for a value to check.
3804 (propagate_vals_across_arith_jfunc): Add an assertion to ensure
3805 finite propagation in self-recursive scc.
3807 2020-01-08 Luo Xiong Hu <luoxhu@linux.ibm.com>
3809 * ipa-inline.c (caller_growth_limits): Restore the AND.
3811 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3813 * config/gcn/gcn-valu.md (VEC_1REG_INT_ALT): Delete iterator.
3814 (VEC_ALLREG_ALT): New iterator.
3815 (VEC_ALLREG_INT_MODE): New iterator.
3816 (VCMP_MODE): New iterator.
3817 (VCMP_MODE_INT): New iterator.
3818 (vec_cmpu<mode>di): Use VCMP_MODE_INT.
3819 (vec_cmp<u>v64qidi): New define_expand.
3820 (vec_cmp<mode>di_exec): Use VCMP_MODE.
3821 (vec_cmpu<mode>di_exec): New define_expand.
3822 (vec_cmp<u>v64qidi_exec): New define_expand.
3823 (vec_cmp<mode>di_dup): Use VCMP_MODE.
3824 (vec_cmp<mode>di_dup_exec): Use VCMP_MODE.
3825 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>): Rename ...
3826 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>): ... to this.
3827 (vcond<VEC_ALL1REG_MODE:mode><VEC_1REG_ALT:mode>_exec): Rename ...
3828 (vcond<VEC_ALLREG_MODE:mode><VEC_ALLREG_ALT:mode>_exec): ... to this.
3829 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>): Rename ...
3830 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>): ... to this.
3831 (vcondu<VEC_ALL1REG_MODE:mode><VEC_1REG_INT_ALT:mode>_exec): Rename ...
3832 (vcondu<VEC_ALLREG_MODE:mode><VEC_ALLREG_INT_MODE:mode>_exec): ... to
3834 * config/gcn/gcn.c (print_operand): Fix 8 and 16 bit suffixes.
3835 * config/gcn/gcn.md (expander): Add sign_extend and zero_extend.
3837 2020-01-07 Andrew Stubbs <ams@codesourcery.com>
3839 * config/gcn/constraints.md (DA): Update description and match.
3841 (Db): New constraint.
3842 * config/gcn/gcn-protos.h (gcn_inline_constant64_p): Add second
3844 * config/gcn/gcn.c (gcn_inline_constant64_p): Add 'mixed' parameter.
3845 Implement 'Db' mixed immediate type.
3846 * config/gcn/gcn-valu.md (addcv64si3<exec_vcc>): Rework constraints.
3847 (addcv64si3_dup<exec_vcc>): Delete.
3848 (subcv64si3<exec_vcc>): Rework constraints.
3849 (addv64di3): Rework constraints.
3850 (addv64di3_exec): Rework constraints.
3851 (subv64di3): Rework constraints.
3852 (addv64di3_dup): Delete.
3853 (addv64di3_dup_exec): Delete.
3854 (addv64di3_zext): Rework constraints.
3855 (addv64di3_zext_exec): Rework constraints.
3856 (addv64di3_zext_dup): Rework constraints.
3857 (addv64di3_zext_dup_exec): Rework constraints.
3858 (addv64di3_zext_dup2): Rework constraints.
3859 (addv64di3_zext_dup2_exec): Rework constraints.
3860 (addv64di3_sext_dup2): Rework constraints.
3861 (addv64di3_sext_dup2_exec): Rework constraints.
3863 2020-01-07 Andre Vieira <andre.simoesdiasvieira@arm.com>
3865 * doc/sourcebuild.texi (arm_little_endian, arm_nothumb): Documented
3866 existing target checks.
3868 2020-01-07 Richard Biener <rguenther@suse.de>
3870 * doc/install.texi: Bump minimal supported MPC version.
3872 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3874 * langhooks-def.h (lhd_simulate_enum_decl): Declare.
3875 (LANG_HOOKS_SIMULATE_ENUM_DECL): Use it.
3876 * langhooks.c: Include stor-layout.h.
3877 (lhd_simulate_enum_decl): New function.
3878 * config/aarch64/aarch64-sve-builtins.cc (init_builtins): Call
3879 handle_arm_sve_h for the LTO frontend.
3880 (register_vector_type): Cope with null returns from pushdecl.
3882 2020-01-07 Richard Sandiford <richard.sandiford@arm.com>
3884 * config/aarch64/aarch64-protos.h (aarch64_sve::svbool_type_p)
3885 (aarch64_sve::nvectors_if_data_type): Replace with...
3886 (aarch64_sve::builtin_type_p): ...this.
3887 * config/aarch64/aarch64-sve-builtins.cc: Include attribs.h.
3888 (find_vector_type): Delete.
3889 (add_sve_type_attribute): New function.
3890 (lookup_sve_type_attribute): Likewise.
3891 (register_builtin_types): Add an "SVE type" attribute to each type.
3892 (register_tuple_type): Likewise.
3893 (svbool_type_p, nvectors_if_data_type): Delete.
3894 (mangle_builtin_type): Use lookup_sve_type_attribute.
3895 (builtin_type_p): Likewise. Add an overload that returns the
3896 number of constituent vector and predicate registers.
3897 * config/aarch64/aarch64.c (aarch64_sve_argument_p): Delete.
3898 (aarch64_returns_value_in_sve_regs_p): Use aarch64_sve::builtin_type_p
3899 instead of aarch64_sve_argument_p.
3900 (aarch64_takes_arguments_in_sve_regs_p): Likewise.
3901 (aarch64_pass_by_reference): Likewise.
3902 (aarch64_function_value_1): Likewise.
3903 (aarch64_return_in_memory): Likewise.
3904 (aarch64_layout_arg): Likewise.
3906 2020-01-07 Jakub Jelinek <jakub@redhat.com>
3908 PR tree-optimization/93156
3909 * tree-ssa-ccp.c (bit_value_binop): For x * x note that the second
3910 least significant bit is always clear.
3912 PR tree-optimization/93118
3913 * match.pd ((x >> c) << c -> x & (-1<<c)): Add nop_convert?. Add new
3914 simplifier with two intermediate conversions.
3916 2020-01-07 Martin Liska <mliska@suse.cz>
3918 * params.opt: Add Optimization for various parameters.
3920 2020-01-07 Martin Liska <mliska@suse.cz>
3923 * doc/extend.texi: Explain cloning for target_clone
3926 2020-01-07 Martin Liska <mliska@suse.cz>
3928 PR tree-optimization/92860
3929 * common.opt: Make in Optimization option
3930 as it is affected by -O0, which is an Optimization
3932 * tree-inline.c (tree_inlinable_function_p):
3933 Use opt_for_fn for warn_inline.
3934 (expand_call_inline): Likewise.
3936 2020-01-07 Martin Liska <mliska@suse.cz>
3938 PR tree-optimization/92860
3939 * common.opt: Make flag_ree as optimization
3942 2020-01-07 Martin Liska <mliska@suse.cz>
3944 PR optimization/92860
3945 * params.opt: Mark param_min_crossjump_insns with Optimization
3948 2020-01-07 Luo Xiong Hu <luoxhu@linux.ibm.com>
3950 * ipa-inline-analysis.c (estimate_growth): Fix typo.
3951 * ipa-inline.c (caller_growth_limits): Use OR instead of AND.
3953 2020-01-06 Michael Meissner <meissner@linux.ibm.com>
3955 * config/rs6000/rs6000.c (hard_reg_and_mode_to_addr_mask): New
3956 helper function to return the valid addressing formats for a given
3957 hard register and mode.
3958 (rs6000_adjust_vec_address): Call hard_reg_and_mode_to_addr_mask.
3960 * config/rs6000/constraints.md (Q constraint): Update
3962 * doc/md.texi (RS/6000 constraints): Update 'Q' cosntraint
3965 * config/rs6000/vsx.md (vsx_extract_<mode>_var, VSX_D iterator):
3966 Use 'Q' for doing vector extract from memory.
3967 (vsx_extract_v4sf_var): Use 'Q' for doing vector extract from
3969 (vsx_extract_<mode>_var, VSX_EXTRACT_I iterator): Use 'Q' for
3970 doing vector extract from memory.
3971 (vsx_extract_<mode>_<VS_scalar>mode_var): Use 'Q' for doing vector
3972 extract from memory.
3974 * config/rs6000/rs6000.c (rs6000_adjust_vec_address): Add support
3975 for the offset being 34-bits when -mcpu=future is used.
3977 2020-01-06 John David Anglin <danglin@gcc.gnu.org>
3979 * config/pa/pa.md: Revert change to use ordered_comparison_operator
3980 instead of cmpib_comparison_operator in cmpib patterns.
3981 * config/pa/predicates.md (cmpib_comparison_operator): Revert removal
3982 of cmpib_comparison_operator. Revise comment.
3984 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3986 * tree-vect-slp.c (vect_build_slp_tree_1): Require all shifts
3987 in an IFN_DIV_POW2 node to be equal.
3989 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
3991 * tree-vect-stmts.c (vect_check_load_store_mask): Rename to...
3992 (vect_check_scalar_mask): ...this.
3993 (vectorizable_store, vectorizable_load): Update call accordingly.
3994 (vectorizable_call): Use vect_check_scalar_mask to check the mask
3995 argument in calls to conditional internal functions.
3997 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
3999 * config/gcn/gcn-valu.md (subv64di3): Use separate alternatives for
4000 '0' matching inputs.
4001 (subv64di3_exec): Likewise.
4003 2020-01-06 Bryan Stenson <bryan@siliconvortex.com>
4005 * config/mips/mips.c (vr4130_align_insns): Fix typo.
4006 * doc/md.texi (movstr): Likewise.
4008 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4010 * config/gcn/gcn-valu.md (vec_extract<mode><scalar_mode>): Add early
4013 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4015 * config/aarch64/t-aarch64 ($(srcdir)/config/aarch64/aarch64-tune.md):
4017 (s-aarch64-tune-md): ...this new stamp file. Pipe the new contents
4018 to a temporary file and use move-if-change to update the real
4019 file where necessary.
4021 2020-01-06 Richard Sandiford <richard.sandiford@arm.com>
4023 * config/aarch64/aarch64-sve.md (@aarch64_sel_dup<mode>): Use Upl
4024 rather than Upa for CPY /M.
4026 2020-01-06 Andrew Stubbs <ams@codesourcery.com>
4028 * config/gcn/gcn.c (gcn_inline_constant_p): Allow 64 as an inline
4031 2020-01-06 Martin Liska <mliska@suse.cz>
4033 PR tree-optimization/92860
4034 * params.opt: Mark param_max_combine_insns with Optimization
4037 2020-01-05 Jakub Jelinek <jakub@redhat.com>
4040 * config/i386/i386.md (SWIDWI): New mode iterator.
4041 (DWI, dwi): Add TImode variants.
4042 (addv<mode>4): Use SWIDWI iterator instead of SWI. Use
4043 <general_hilo_operand> instead of <general_operand>. Use
4044 CONST_SCALAR_INT_P instead of CONST_INT_P.
4045 (*addv<mode>4_1): Rename to ...
4046 (addv<mode>4_1): ... this.
4047 (QWI): New mode attribute.
4048 (*addv<dwi>4_doubleword, *addv<dwi>4_doubleword_1): New
4049 define_insn_and_split patterns.
4050 (*addv<mode>4_overflow_1, *addv<mode>4_overflow_2): New define_insn
4052 (uaddv<mode>4): Use SWIDWI iterator instead of SWI. Use
4053 <general_hilo_operand> instead of <general_operand>.
4054 (*addcarry<mode>_1): New define_insn.
4055 (*add<dwi>3_doubleword_cc_overflow_1): New define_insn_and_split.
4057 2020-01-03 Konstantin Kharlamov <Hi-Angel@yandex.ru>
4059 * gdbinit.in (pr, prl, pt, pct, pgg, pgq, pgs, pge, pmz, pdd, pbs, pbm):
4060 Use "call" instead of "set".
4062 2020-01-03 Martin Jambor <mjambor@suse.cz>
4065 * ipa-cp.c (print_all_lattices): Skip functions without info.
4067 2020-01-03 Jakub Jelinek <jakub@redhat.com>
4070 * config/i386/i386-options.c (ix86_simd_clone_adjust): If
4071 TARGET_PREFER_AVX128, use prefer-vector-width=256 for 'c' and 'd'
4072 simd clones. If TARGET_PREFER_AVX256, use prefer-vector-width=512
4073 for 'e' simd clones.
4076 * config/i386/i386.opt (x_prefer_vector_width_type): Remove TargetSave
4078 (mprefer-vector-width=): Add Save.
4079 * config/i386/i386-options.c (ix86_target_string): Add PVW argument, print
4080 -mprefer-vector-width= if non-zero. Fix up -mfpmath= comment.
4081 (ix86_debug_options, ix86_function_specific_print): Adjust
4082 ix86_target_string callers.
4083 (ix86_valid_target_attribute_inner_p): Handle prefer-vector-width=.
4084 (ix86_valid_target_attribute_tree): Likewise.
4085 * config/i386/i386-options.h (ix86_target_string): Add PVW argument.
4086 * config/i386/i386-expand.c (ix86_expand_builtin): Adjust
4087 ix86_target_string caller.
4090 * config/i386/i386.md (abs<mode>2): Use expand_simple_binop instead of
4091 emitting ASHIFTRT, XOR and MINUS by hand. Use gen_int_mode with QImode
4092 instead of gen_int_shift_amount + convert_modes.
4094 PR rtl-optimization/93088
4095 * loop-iv.c (find_single_def_src): Punt after looking through
4096 128 reg copies for regs with single definitions. Move definitions
4099 2020-01-02 Dennis Zhang <dennis.zhang@arm.com>
4101 * config/arm/arm-c.c (arm_cpu_builtins): Define
4102 __ARM_FEATURE_MATMUL_INT8, __ARM_FEATURE_BF16_VECTOR_ARITHMETIC,
4103 __ARM_FEATURE_BF16_SCALAR_ARITHMETIC, and
4104 __ARM_BF16_FORMAT_ALTERNATIVE when enabled.
4105 * config/arm/arm-cpus.in (armv8_6, i8mm, bf16): New features.
4106 * config/arm/arm-tables.opt: Regenerated.
4107 * config/arm/arm.c (arm_option_reconfigure_globals): Initialize
4108 arm_arch_i8mm and arm_arch_bf16 when enabled.
4109 * config/arm/arm.h (TARGET_I8MM): New macro.
4110 (TARGET_BF16_FP, TARGET_BF16_SIMD): Likewise.
4111 * config/arm/t-aprofile: Add matching rules for -march=armv8.6-a.
4112 * config/arm/t-arm-elf (all_v8_archs): Add armv8.6-a.
4113 * config/arm/t-multilib: Add matching rules for -march=armv8.6-a.
4114 (v8_6_a_simd_variants): New.
4115 (v8_*_a_simd_variants): Add i8mm and bf16.
4116 * doc/invoke.texi (armv8.6-a, i8mm, bf16): Document new options.
4118 2020-01-02 Jakub Jelinek <jakub@redhat.com>
4121 * predict.c (compute_function_frequency): Don't call
4122 warn_function_cold on functions that already have cold attribute.
4124 2020-01-01 John David Anglin <danglin@gcc.gnu.org>
4127 * config/pa/pa.c (pa_elf_select_rtx_section): New. Put references to
4128 COMDAT group function labels in .data.rel.ro.local section.
4129 * config/pa/pa32-linux.h (TARGET_ASM_SELECT_RTX_SECTION): Define.
4132 * config/pa/pa.md (scc): Use ordered_comparison_operator instead of
4133 comparison_operator in B and S integer comparisons. Likewise, use
4134 ordered_comparison_operator instead of cmpib_comparison_operator in
4136 * config/pa/predicates.md (cmpib_comparison_operator): Remove.
4138 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4140 Update copyright years.
4142 * gcc.c (process_command): Update copyright notice dates.
4143 * gcov-dump.c (print_version): Ditto.
4144 * gcov.c (print_version): Ditto.
4145 * gcov-tool.c (print_version): Ditto.
4146 * gengtype.c (create_file): Ditto.
4147 * doc/cpp.texi: Bump @copying's copyright year.
4148 * doc/cppinternals.texi: Ditto.
4149 * doc/gcc.texi: Ditto.
4150 * doc/gccint.texi: Ditto.
4151 * doc/gcov.texi: Ditto.
4152 * doc/install.texi: Ditto.
4153 * doc/invoke.texi: Ditto.
4155 2020-01-01 Jan Hubicka <hubicka@ucw.cz>
4157 * ipa.c (walk_polymorphic_call_targets): Fix updating of overall
4160 2020-01-01 Jakub Jelinek <jakub@redhat.com>
4162 PR tree-optimization/93098
4163 * match.pd (popcount): For shift amounts, use integer_onep
4164 or wi::to_widest () == cst instead of tree_to_uhwi () == cst
4165 tests. Make sure that precision is power of two larger than or equal
4166 to 16. Ensure shift is never negative. Use HOST_WIDE_INT_UC macro
4167 instead of ULL suffixed constants. Formatting fixes.
4169 Copyright (C) 2020 Free Software Foundation, Inc.
4171 Copying and distribution of this file, with or without modification,
4172 are permitted in any medium without royalty provided the copyright
4173 notice and this notice are preserved.