1 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
4 * config/i386/constraints.md (Bs): Replace
5 ix86_indirect_branch_register with
6 TARGET_INDIRECT_BRANCH_REGISTER.
8 * config/i386/i386.md (indirect_jump): Likewise.
10 (*sibcall_memory): Likewise.
11 (*sibcall_value_memory): Likewise.
12 Peepholes of indirect call and jump via memory: Likewise.
13 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
14 (*sibcall_value_GOT_32): Likewise.
15 * config/i386/i386.opt: Likewise.
16 * config/i386/predicates.md (indirect_branch_operand): Likewise.
17 (GOT_memory_operand): Likewise.
18 (call_insn_operand): Likewise.
19 (sibcall_insn_operand): Likewise.
20 (GOT32_symbol_operand): Likewise.
21 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
23 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
25 PR rtl-optimization/83496
26 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
27 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
28 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
29 redundant insn, if any.
30 (relax_delay_slots): Likewise.
31 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
33 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
35 PR tree-optimization/83965
36 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
37 that grouped statements are part of a reduction chain. Return
38 true if the statement is not marked as a reduction itself but
40 (vect_recog_dot_prod_pattern): Don't check whether the statement
41 is part of a group here.
42 (vect_recog_sad_pattern): Likewise.
43 (vect_recog_widen_sum_pattern): Likewise.
45 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
48 * final.c (rest_of_clean_state): Also look for calls inside sequences.
50 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
53 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
55 (ix86_output_indirect_function_return): New prototype.
56 (ix86_split_simple_return_pop_internal): Likewise.
57 * config/i386/i386.c (indirect_return_via_cx): New.
58 (indirect_return_via_cx_bnd): Likewise.
59 (indirect_thunk_name): Handle return va CX_REG.
60 (output_indirect_thunk_function): Create alias for
61 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
62 (ix86_output_indirect_jmp): Remove the bool argument.
63 (ix86_output_indirect_function_return): New function.
64 (ix86_split_simple_return_pop_internal): Likewise.
65 * config/i386/i386.md (*indirect_jump): Don't pass false
66 to ix86_output_indirect_jmp.
67 (*tablejump_1): Likewise.
68 (simple_return_pop_internal): Change it to define_insn_and_split.
69 Call ix86_split_simple_return_pop_internal to split it for
71 (simple_return_indirect_internal): Call
72 ix86_output_indirect_function_return instead of
73 ix86_output_indirect_jmp.
75 2018-02-26 Jakub Jelinek <jakub@redhat.com>
78 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
79 memset and value initialization afterwards.
81 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
83 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
85 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
88 * common/config/aarch64/aarch64-common.c
89 (aarch_option_optimization_table[]): Switch
90 off fomit-frame-pointer
92 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
93 Chung-Ju Wu <jasonwucj@gmail.com>
95 * config/nds32/nds32-multiple.md (load_multiple): Disallow
97 (store_multiple): Ditto.
99 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
101 * config.gcc: Add --with-cpu support for nds32 target.
102 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
103 * config/nds32/nds32.opt: Add -mcpu= option.
105 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
107 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
108 isel=yes): Warn for these deprecated options.
110 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
112 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
113 ISA_2_5_MASKS_EMBEDDED.
115 2018-02-23 Jakub Jelinek <jakub@redhat.com>
117 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
118 p->max as pointers rather than using iterative_hash_expr.
120 2018-02-23 Carl Love <cel@us.ibm.com>
122 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
123 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
125 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
126 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
127 P8V_BUILTIN_VEC_VUNSIGNED2.
129 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
132 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
133 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
135 * lra-constraints.c (curr_insn_transform): Set up
136 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
138 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
139 * lra-eliminations.c (spill_pseudos): Ditto.
140 (process_insn_for_elimination): Ditto.
141 * lra-lives.c (reg_early_clobber_p): Use the new macros.
142 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
143 LRA_NON_CLOBBERED_ALT.
145 2018-02-22 Martin Sebor <msebor@redhat.com>
147 PR tree-optimization/84480
148 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
149 to maybe_diag_stxncpy_trunc. Call it.
150 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
151 from gimple_fold_builtin_strcpy. Print inlining stack.
152 (handle_builtin_stxncpy): Print inlining stack.
153 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
155 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
158 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
159 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
160 and -fcheck-pointer-bounds are used together.
161 (indirect_thunk_prefix): New enum.
162 (indirect_thunk_need_prefix): New function.
163 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
164 "_nt" instead of "_bnd" for NOTRACK prefix.
165 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
166 (output_indirect_thunk_function): Likewise.
168 (ix86_code_end): Update output_indirect_thunk_function calls.
169 (ix86_output_indirect_branch_via_reg): Replace
170 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
171 (ix86_output_indirect_branch_via_push): Likewise.
172 (ix86_output_function_return): Likewise.
173 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
174 incompatible with -fcf-protection=branch and
175 -fcheck-pointer-bounds.
177 2018-02-22 Steve Ellcey <sellcey@cavium.com>
180 * config/aarch64/aarch64.c (aarch64_print_address_internal):
181 Change gcc_assert call to output_operand_lossage.
183 2018-02-22 Steve Ellcey <sellcey@cavium.com>
185 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
187 2018-02-22 DJ Delorie <dj@redhat.com>
188 Sebastian Perta <sebastian.perta@renesas.com>
189 Oleg Endo <olegendo@gcc.gnu.org>
191 * config/rx/rx.c (rx_rtx_costs): New function.
192 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
194 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
196 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
198 2018-02-22 Martin Liska <mliska@suse.cz>
201 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
202 Add "native" as a possible value.
204 2018-02-22 Martin Liska <mliska@suse.cz>
207 * config/i386/i386.c (ix86_option_override_internal):
208 Add "native" as a possible value for -march and -mtune.
210 2018-02-22 Jakub Jelinek <jakub@redhat.com>
213 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
214 to all type variants.
216 PR tree-optimization/84503
217 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
218 width as info->bitpos + info->bitsize - start.
219 (merged_store_group::merge_overlapping): Simplify width computation.
220 (check_no_overlap): New function.
221 (imm_store_chain_info::try_coalesce_bswap): Compute expected
222 start + width and last_order of the group, fail if check_no_overlap
224 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
225 to group if check_no_overlap fails.
227 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
229 * config/rs6000/altivec.md: Delete contraint arguments to
230 define_expand, define_split, and define_peephole2, and in
231 define_insn_and_split if always unused.
232 * config/rs6000/darwin.md: Ditto.
233 * config/rs6000/dfp.md: Ditto.
234 * config/rs6000/rs6000.md: Ditto.
235 * config/rs6000/sync.md: Ditto.
236 * config/rs6000/vector.md: Ditto.
237 * config/rs6000/vsx.md: Ditto.
239 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
241 * config/rs6000/altivec.md: Write output control strings as braced
242 blocks instead of double-quoted strings.
243 * config/rs6000/darwin.md: Ditto.
244 * config/rs6000/rs6000.md: Ditto.
245 * config/rs6000/vector.md: Ditto.
246 * config/rs6000/vsx.md: Ditto.
248 2018-02-21 Jason Merrill <jason@redhat.com>
250 PR c++/84314 - ICE with templates and fastcall attribute.
251 * attribs.c (build_type_attribute_qual_variant): Remove assert.
253 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
255 * ipa-cp.c (determine_versionability): Fix comment typos.
257 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
260 * ipa-cp.c (determine_versionability): Do not version functions caling
263 2018-02-21 Martin Liska <mliska@suse.cz>
266 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
267 Add "native" as a possible value.
268 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
269 the macro when native cpu detection is available.
271 2018-02-21 Martin Liska <mliska@suse.cz>
274 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
275 Add "native" as a possible value.
276 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
277 when native cpu detection is available.
279 2018-02-21 Jakub Jelinek <jakub@redhat.com>
280 Martin Sebor <msebor@redhat.com>
282 PR tree-optimization/84478
283 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
285 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
286 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
287 support which is conservatively correct, for 2 only stay conservative
288 for maxlen. Formatting and comment capitalization fixes. Add STRICT
289 argument to the 2 argument get_range_strlen, adjust 6 arg
290 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
292 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
293 (gimple_fold_builtin_strlen): Pass true as last argument to
296 2018-02-20 Martin Sebor <msebor@redhat.com>
299 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
300 (builtin_memref::set_base_and_offset): Same. Handle inner references.
301 (builtin_memref::builtin_memref): Factor out parts into
302 set_base_and_offset and call it.
304 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
307 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
308 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
309 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
310 search at the associated MODE_INT.
312 2018-02-20 Jeff Law <law@redhat.com>
315 PR tree-optimization/81592
317 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
318 for range data rather than using global data.
319 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
320 range data rather than using global data.
321 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
322 pass it to children as needed.
323 (struct directive::fmtresult): Similarly.
324 (struct directive::set_width): Similarly.
325 (struct directive::set_precision): Similarly.
326 (format_integer, format_directive, parse_directive): Similarly.
327 (format_none): Accept unnamed vr_values parameter.
328 (format_percent, format_floating, format_character): Similarly.
329 (format_string, format_plain): Similarly.
330 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
331 the EVRP range analyzer for range data rather than using global data.
332 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
333 gimple-ssa-evrp-analyze.h
334 (class sprintf_dom_walker): Add after_dom_children member function.
335 Add evrp_range_analyzer member.
336 (sprintf_dom_walker::before_dom_children): Call into the EVRP
337 range analyzer as needed.
338 (sprintf_dom_walker::after_dom_children): New member function.
339 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
341 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
342 (evrp_range_analyzer::pop_to_marker): Likewise.
344 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
346 PR tree-optimization/84419
347 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
348 with the required type if its current type is compatible but
351 2018-02-20 Jakub Jelinek <jakub@redhat.com>
354 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
357 2018-02-20 Martin Liska <mliska@suse.cz>
360 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
361 possible values if we don't have a hint.
363 2018-02-20 Martin Liska <mliska@suse.cz>
367 * final.c (shorten_branches): Build align_tab array with one
369 * opts.c (finish_options): Add alignment option limit check.
370 (MAX_CODE_ALIGN): Likewise.
371 (MAX_CODE_ALIGN_VALUE): Likewise.
372 * doc/invoke.texi: Document maximum allowed option value for
373 all -falign-* options.
375 2018-02-19 Jakub Jelinek <jakub@redhat.com>
378 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
379 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
380 * var-tracking.c (emit_note_insn_var_location): Remove all references
381 to NOTE_INSN_CALL_ARG_LOCATION.
382 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
383 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
384 Use copy_rtx_if_shared.
385 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
386 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
387 (dwarf2out_var_location): Remove handling of
388 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
390 * final.c (final_scan_insn): Remove all references to
391 NOTE_INSN_CALL_ARG_LOCATION.
392 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
393 before dumping final insns.
394 * except.c (emit_note_eh_region_end): Remove all references to
395 NOTE_INSN_CALL_ARG_LOCATION.
396 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
397 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
398 * config/arc/arc.c (hwloop_optimize): Likewise.
399 * config/arm/arm.c (create_fix_barrier): Likewise.
400 * config/s390/s390.c (s390_chunkify_start): Likewise.
401 * config/sh/sh.c (find_barrier): Likewise.
402 * config/i386/i386.c (rest_of_insert_endbranch,
403 ix86_seh_fixup_eh_fallthru): Likewise.
404 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
405 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
406 * config/frv/frv.c (frv_function_prologue): Likewise.
407 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
409 (note_outside_basic_block_p): Remove all references to
410 NOTE_INSN_CALL_ARG_LOCATION.
411 * gengtype.c (adjust_field_rtx_def): Likewise.
412 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
414 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
415 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
418 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
421 PR tree-optimization/84452
422 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
423 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
426 2018-02-19 Martin Liska <mliska@suse.cz>
429 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
431 2018-02-19 Martin Liska <mliska@suse.cz>
432 Richard Sandiford <richard.sandiford@linaro.org>
434 PR tree-optimization/82491
435 * gimple-fold.c (get_base_constructor): Make earlier bail out
438 2018-02-19 Carl Love <cel@us.ibm.com>
440 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
441 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
443 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
446 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
448 * config/rl78/rl78.md (movdf): New define expand.
450 2018-02-19 Martin Liska <mliska@suse.cz>
453 * doc/invoke.texi: Fix typo.
454 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
456 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
458 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
459 handle rs6000_single_float and rs6000_double_float specially for
462 2018-02-16 Jeff Law <law@redhat.com>
464 * config/rx/rx.c (add_pop_cfi_notes): New function.;
467 2018-02-16 Jakub Jelinek <jakub@redhat.com>
470 * ipa-inline.c (inline_small_functions): Fix a typo.
472 2018-02-16 Nathan Sidwell <nathan@acm.org>
474 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
476 2018-02-16 Carl Love <cel@us.ibm.com>
478 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
479 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
480 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
481 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
482 expansion to P8V_BUILTIN_VEC_FLOAT2.
484 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
486 PR rtl-optimization/70023
487 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
488 src_regno into account.
490 2018-02-16 Carl Love <cel@us.ibm.com>
492 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
493 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
494 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
495 * config/rs6000/rs6000.c: Remove case statements for
496 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
497 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
498 and P9V_BUILTIN_VEC_VINSERT4B.
499 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
500 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
501 * config/rs6000/vsx.md:
502 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
505 2018-02-16 Carl Love <cel@us.ibm.com>
507 * config/rs6000/altivec.h: Add builtin names vec_extract4b
509 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
511 * config/rs6000/rs6000-c.c: Add the definitions for
512 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
513 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
514 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
515 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
516 definition for insert4b and define insn *insert3b_internal.
517 * doc/extend.texi: Add documentation for vec_extract4b.
519 2018-02-16 Nathan Sidwell <nathan@acm.org>
521 * doc/extend.texi (Backwards Compatibility): Mention friend
522 injection. Note for-scope is deprecated.
523 * doc/invoke.texi (-ffriend-injection): Deprecate.
525 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
527 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
528 that moved to I2, also allow destinations that are a paradoxical
529 subreg (instead of a normal reg).
531 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
534 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
537 2018-02-16 Richard Biener <rguenther@suse.de>
539 PR tree-optimization/84037
540 PR tree-optimization/84016
542 * config/i386/i386.c (ix86_builtin_vectorization_cost):
543 Adjust vec_construct for the fact we need additional higher latency
544 128bit inserts for AVX256 and AVX512 vector builds.
545 (ix86_add_stmt_cost): Scale vector construction cost for
548 2018-02-16 Richard Biener <rguenther@suse.de>
550 PR tree-optimization/84417
551 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
552 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
553 (non_rewritable_lvalue_p): Likewise, use poly-ints.
555 2018-02-16 Martin Liska <mliska@suse.cz>
558 * internal-fn.def (ASAN_CHECK): Set proper flags.
559 (ASAN_MARK): Likewise.
561 2018-02-16 Julia Koval <julia.koval@intel.com>
563 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
566 2018-02-16 Jakub Jelinek <jakub@redhat.com>
569 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
570 Use ++iter rather than iter++ for std::list iterators.
571 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
572 defer deleting them until all nodes in the forest are processed. Do
573 free even leaf nodes. Change to_process into auto_vec.
576 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
577 * vec.h (vec_default_construct): Use memset instead of placement new
578 if BROKEN_VALUE_INITIALIZATION is defined.
579 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
580 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
583 PR rtl-optimization/83723
584 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
585 * lra.c (lra_substitute_pseudo): Likewise. If true, use
586 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
588 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
590 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
592 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
594 PR rtl-optimization/81443
595 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
596 from inner REGs to paradoxical SUBREGs.
598 2018-02-16 Richard Biener <rguenther@suse.de>
600 PR tree-optimization/84399
601 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
602 For operands we can analyze at their definition make sure we can
603 analyze them at each use as well.
605 2018-02-16 Richard Biener <rguenther@suse.de>
607 PR tree-optimization/84190
608 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
609 volatile accesses if the decl isn't volatile.
611 2018-02-15 Jason Merrill <jason@redhat.com>
613 PR c++/84314 - ICE with templates and fastcall attribute.
614 * attribs.c (build_type_attribute_qual_variant): Don't clobber
615 TYPE_CANONICAL on an existing type.
617 2018-02-15 Jakub Jelinek <jakub@redhat.com>
619 PR tree-optimization/84383
620 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
621 dstoff nor call operand_equal_p if dstbase is NULL.
623 PR tree-optimization/84334
624 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
625 also a CONSTANT_CLASS_P, punt.
627 2018-02-14 Jim Wilson <jimw@sifive.com>
629 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
630 first SMALL_OPERAND check. New local min_second_step. Move assert
631 to where locals are set. Add TARGET_RVC support.
632 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
634 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
636 * doc/invoke.texi: Correct -Wformat-overflow code sample.
638 2018-02-14 Martin Sebor <msebor@redhat.com>
640 PR tree-optimization/83698
641 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
642 arrays constrain the offset range to their bounds.
643 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
644 (builtin_access::overlap): Avoid setting the size of overlap if it's
646 (maybe_diag_overlap): Also consider arrays when deciding what values
647 of offsets to include in diagnostics.
649 2018-02-14 Martin Sebor <msebor@redhat.com>
652 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
653 that correspond to the kind of a declaration.
655 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
658 * config/pa/pa.md: Load address of PIC label using the linkage table
659 if the label is nonlocal.
661 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
663 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
664 warning message if user requests -maltivec=be.
665 * doc/invoke.texi: Document deprecation of -maltivec=be.
667 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
670 * config/rs6000/rs6000-c.c: Update definitions for
671 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
672 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
674 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
677 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
678 add _get_ssp intrinsics. Remove argument from
679 __builtin_ia32_rdssp[d|q].
680 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
681 * config/i386/i386-builtin.def: Remove argument from
682 __builtin_ia32_rdssp[d|q].
683 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
684 ix86_expand_special_args_builtin for _rdssp[d|q].
685 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
686 Clear register before usage.
687 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
688 Add documentation for new _get_ssp and _inc_ssp intrinsics.
690 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
692 PR tree-optimization/84357
693 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
694 operand 1 of an ARRAY_REF too.
696 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
699 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
700 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
702 (set_of_reg): New struct.
703 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
704 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
705 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
707 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
708 Split into bitclr, bitset, bitinvert patterns if appropriate.
709 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
710 use rx_fuse_in_memory_bitop.
711 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
712 to named insn, correct maximum insn length.
714 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
717 * machmode.def: Define a complex mode for PARTIAL_INT.
718 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
720 * doc/rtl.texi: Document CSPImode.
721 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
723 (msp430_hard_regno_nregs_with_padding): Likewise.
725 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
728 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
730 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
732 PR rtl-optimization/84169
733 * combine.c (try_combine): New variable split_i2i3. Set it to true if
734 we generated a parallel as new i3 and we split that to new i2 and i3
735 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
736 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
737 those to i2, not i1. Partially rewrite this scan code.
739 2018-02-13 Jakub Jelinek <jakub@redhat.com>
742 * stor-layout.c (place_field): For variable length fields, adjust
743 offset_align afterwards not just based on the field's alignment,
744 but also on the size.
747 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
748 of exps and logs in the use_exp2 case.
750 2018-02-13 Jeff Law <law@redhat.com>
752 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
755 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
758 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
762 * common.opt (gas-loc-support, gas-locview-support): New.
763 (ginline-points, ginternal-reset-location-views): New.
764 * doc/invoke.texi: Document them. Use @itemx where intended.
765 (gvariable-location-views): Adjust.
766 * target.def (reset_location_view): New.
767 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
768 (TARGET_RESET_LOCATION_VIEW): New.
769 * doc/tm.texi: Rebuilt.
770 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
771 (dwarf2out_default_as_locview_support): New.
772 (output_asm_line_debug_info): Use option variables.
773 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
774 (output_loc_list): Likewise.
775 (add_high_low_attributes): Check option variables.
776 Don't output entry view attribute in strict mode.
777 (gen_inlined_subroutine_die): Check option variables.
778 (dwarf2out_inline_entry): Likewise.
779 (init_sections_and_labels): Likewise.
780 (dwarf2out_early_finish): Likewise.
781 (maybe_reset_location_view): New, from...
782 (dwarf2out_var_location): ... here. Call it.
783 * debug.h (dwarf2out_default_as_loc_support): Declare.
784 (dwarf2out_default_as_locview_support): Declare.
785 * hooks.c (hook_int_rtx_insn_0): New.
786 * hooks.h (hook_int_rtx_insn_0): Declare.
787 * toplev.c (process_options): Take -gas-loc-support and
788 -gas-locview-support from dwarf2out. Enable
789 -gvariable-location-views by default only with locview
790 assembler support. Enable -ginternal-reset-location-views by
791 default only if the target defines the corresponding hook.
792 Enable -ginline-points by default if location views are
793 enabled; force it disabled if statement frontiers are
795 * tree-inline.c (expand_call_inline): Check option variables.
796 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
798 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
800 PR tree-optimization/84321
801 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
802 handling. Also check whether the anti-range contains any values
803 that satisfy the mask; switch to a VR_RANGE if not.
805 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
808 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
810 2018-02-13 Martin Jambor <mjambor@suse.cz>
813 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
814 of call statements, also set location of a load to a temporary.
816 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
818 * config/rl78/rl78.c (add_vector_labels): New function.
819 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
820 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
821 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
822 which checks that no arguments are passed.
823 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
824 * doc/extend.texi: Documentation for the new attribute.
826 2018-02-13 Andreas Schwab <schwab@suse.de>
828 * config/riscv/linux.h (CPP_SPEC): Define.
830 2018-02-13 Jakub Jelinek <jakub@redhat.com>
833 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
834 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
835 OPTION_MASK_ISA_AES as first argument to def_builtin_const
836 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
837 instead of OPTION_MASK_ISA_PCLMUL as first argument to
838 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
839 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
840 temporarily for AES and PCLMUL builtins.
842 PR tree-optimization/84339
843 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
844 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
848 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
849 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
850 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
852 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
854 * omp-simd-clone.h: New file.
855 * omp-simd-clone.c: Include omp-simd-clone.h.
856 (expand_simd_clones): No longer static.
857 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
858 cgraph.h and omp-simd-clone.h.
859 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
860 (vect_recog_widen_shift_pattern): Formatting fix.
861 (vect_pattern_recog_1): Don't check optab for calls.
864 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
865 operands[2] into a REG before using gen_lowpart on it.
867 2018-02-12 Jeff Law <law@redhat.com>
870 * config/sh/sh.c (find_barrier): Consider a sibling call
873 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
874 successfully back substituting a reg.
876 2018-02-12 Richard Biener <rguenther@suse.de>
878 PR tree-optimization/84037
879 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
880 parameter, move visited init to caller.
881 (vect_slp_analyze_operations): Separate cost from validity
882 check, initialize visited once for all instances.
883 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
885 * tree-vect-stmts.c (vect_model_simple_cost): Make early
887 (vect_model_promotion_demotion_cost): Likewise.
888 (vectorizable_bswap): Guard cost modeling with !slp_node
889 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
891 (vectorizable_call): Likewise.
892 (vectorizable_conversion): Likewise.
893 (vectorizable_assignment): Likewise.
894 (vectorizable_shift): Likewise.
895 (vectorizable_operation): Likewise.
896 (vectorizable_store): Likewise.
897 (vectorizable_load): Likewise.
898 (vectorizable_condition): Likewise.
899 (vectorizable_comparison): Likewise.
901 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
904 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
905 (ASAN_MARK): Fix fnspec to account for return value, change pointer
906 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
908 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
911 * params.def (inline-min-speedup): Increase from 8 to 15.
912 (max-inline-insns-auto): Decrease from 40 to 30.
913 * ipa-split.c (consider_split): Add some buffer for function to
914 be considered inlining candidate.
915 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
918 2018-02-12 Richard Biener <rguenther@suse.de>
920 PR tree-optimization/84037
921 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
922 matched stmts if we cannot swap the non-matched ones.
924 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
926 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
927 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
928 _mm_maskz_scalef_round_ss): New intrinsics.
929 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
930 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
931 __builtin_ia32_scalefss_round): Remove.
932 (__builtin_ia32_scalefsd_mask_round,
933 __builtin_ia32_scalefss_mask_round): New intrinsics.
934 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
935 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
936 ((match_operand:VF_128 2 "<round_nimm_predicate>"
937 "<round_constraint>")): Changed to ...
938 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
939 "<round_scalar_constraint>")): ... this.
940 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
941 %0, %1, %2<round_op3>}"): Changed to ...
942 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
943 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
944 %2<round_scalar_mask_op3>}"): ... this.
945 * config/i386/subst.md (round_scalar_nimm_predicate): New.
947 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
949 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
950 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
951 (_mm_maskz_sqrt_round_ss): New intrinsics.
952 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
953 (__builtin_ia32_sqrtsd_mask_round)
954 (__builtin_ia32_sqrtss_mask_round): New builtins.
955 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
956 (__builtin_ia32_sqrtss_round): Remove.
957 (__builtin_ia32_sqrtsd_mask_round)
958 (__builtin_ia32_sqrtss_mask_round): New builtins.
959 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
960 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
961 ((match_operand:VF_128 1 "vector_operand"
962 "xBm,<round_constraint>")): Changed to ...
963 ((match_operand:VF_128 1 "vector_operand"
964 "xBm,<round_scalar_constraint>")): ... this.
965 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
966 %0, %2, %<iptr>1<round_op3>}): Changed to ...
967 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
968 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
969 %<iptr>1<round_scalar_mask_op3>}): ... this.
970 ((set_attr "prefix" "<round_prefix>")): Changed to ...
971 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
973 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
976 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
977 Cast vec_cmpeq result to correct type.
978 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
979 Cast vec_cmpgt result to correct type.
981 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
983 * final.c (final_scan_insn_1): Renamed from...
984 (final_scan_insn): ... this. New wrapper, to recover
985 seen from the outermost call in recursive ones.
986 * config/sparc/sparc.c (output_return): Drop seen from call.
987 (output_sibcall): Likewise.
988 * config/visium/visium.c (output_branch): Likewise.
990 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
992 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
995 2018-02-10 Alan Modra <amodra@gmail.com>
998 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
999 Specify LR as an input.
1001 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1004 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1005 remove_member_access_dummy_vars): New functions.
1006 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1007 lower_omp_1, execute_lower_omp): Use them.
1009 PR rtl-optimization/84308
1010 * shrink-wrap.c (spread_components): Release todo vector.
1012 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1014 PR rtl-optimization/57193
1015 * ira-color.c (struct allocno_color_data): Add member
1016 conflict_allocno_hard_prefs.
1017 (update_conflict_allocno_hard_prefs): New.
1018 (bucket_allocno_compare_func): Add a preference based on
1019 conflict_allocno_hard_prefs.
1020 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1021 (color_allocnos): Remove a dead code. Initiate
1022 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1024 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1027 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1028 constraint from =wa to wa. Avoid a subreg on the output operand,
1029 instead use a pseudo and subreg it in a move.
1030 (p9_xxbrd_<mode>): Changed to ...
1031 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1032 (p9_xxbrd_v2df): New expander.
1033 (p9_xxbrw_<mode>): Changed to ...
1034 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1035 (p9_xxbrw_v4sf): New expander.
1037 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1039 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1041 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1044 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1045 multiply in 32-bit mode.
1046 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1047 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1050 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1052 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1053 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1054 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1055 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1057 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1060 * dwarf2out.c (is_trivial_indirect_ref): New function.
1061 (dwarf2out_late_global_decl): Do not generate a location
1062 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1063 and that are not defined in the current unit.
1065 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1067 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1068 instead of a libcall for UNORDERED.
1070 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1073 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1074 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1076 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1079 * config/s390/s390.c (s390_set_current_function): Invoke
1080 s390_indirect_branch_settings also if fndecl didn't change.
1082 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1084 * config/rs6000/rs6000.md (blockage): Set length to zero.
1086 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1088 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1090 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1093 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1094 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1098 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1099 PARALLEL incoming that failed vt_get_decl_and_offset check.
1102 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1103 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1105 (get_variable_section): For decls in named .bss* sections pass true as
1106 second argument to bss_initializer_p.
1108 2018-02-09 Marek Polacek <polacek@redhat.com>
1109 Jakub Jelinek <jakub@redhat.com>
1112 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1113 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1114 Sync some changes from cxx_fold_indirect_ref.
1116 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1118 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1120 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1121 (BLOCK_INLINE_ENTRY_LABEL): New.
1122 (dwarf2out_var_location): Disregard inline entry markers.
1123 (inline_entry_data): New struct.
1124 (inline_entry_data_hasher): New hashtable type.
1125 (inline_entry_data_hasher::hash): New.
1126 (inline_entry_data_hasher::equal): New.
1127 (inline_entry_data_table): New variable.
1128 (add_high_low_attributes): Add DW_AT_entry_pc and
1129 DW_AT_GNU_entry_view attributes if a pending entry is found
1130 in inline_entry_data_table. Add old entry_pc attribute only
1131 if debug nonbinding markers are disabled.
1132 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1133 markers are enabled.
1134 (block_within_block_p, dwarf2out_inline_entry): New.
1135 (dwarf2out_finish): Check that no entries remained in
1136 inline_entry_data_table.
1137 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1138 (final_scan_insn, notice_source_line): Likewise.
1139 (rest_of_clean_state): Skip inline entry markers.
1140 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1142 * gimple.c (gimple_build_debug_inline_entry): New.
1143 * gimple.h (enum gimple_debug_subcode): Add
1144 GIMPLE_DEBUG_INLINE_ENTRY.
1145 (gimple_build_debug_inline_entry): Declare.
1146 (gimple_debug_inline_entry_p): New.
1147 (gimple_debug_nonbind_marker_p): Adjust.
1148 * insn-notes.def (INLINE_ENTRY): New.
1149 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1150 inline entry marker notes.
1151 (print_insn): Likewise.
1152 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1153 (INSN_DEBUG_MARKER_KIND): Likewise.
1154 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1155 * tree-inline.c (expand_call_inline): Build and insert
1156 debug_inline_entry stmt.
1157 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1158 inline entry blocks early, if nonbind markers are enabled.
1159 (dump_scope_block): Dump fragment info.
1160 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1161 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1162 (gimple_build_debug_inline_entry): New.
1163 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1164 Enable/disable inline entry points too.
1165 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1166 (DEBUG_INSN): Describe inline entry markers.
1168 * common.opt (gvariable-location-views): New.
1169 (gvariable-location-views=incompat5): New.
1170 * config.in: Rebuilt.
1171 * configure: Rebuilt.
1172 * configure.ac: Test assembler for view support.
1173 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1174 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1175 * dwarf2out.c (var_loc_view): New typedef.
1176 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1177 (dwarf2out_locviews_in_attribute): New.
1178 (dwarf2out_locviews_in_loclist): New.
1179 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1180 (enum dw_line_info_opcode): Add LI_adv_address.
1181 (struct dw_line_info_table): Add view.
1182 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1183 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1184 (zero_view_p): New variable.
1185 (ZERO_VIEW_P): New macro.
1186 (output_asm_line_debug_info): New.
1187 (struct var_loc_node): Add view.
1188 (add_AT_view_list, AT_loc_list): New.
1189 (add_var_loc_to_decl): Add view param. Test it against last.
1190 (new_loc_list): Add view params. Record them.
1191 (AT_loc_list_ptr): Handle loc and view lists.
1192 (view_list_to_loc_list_val_node): New.
1193 (print_dw_val): Handle dw_val_class_view_list.
1194 (size_of_die): Likewise.
1195 (value_format): Likewise.
1196 (loc_list_has_views): New.
1197 (gen_llsym): Set vl_symbol too.
1198 (maybe_gen_llsym, skip_loc_list_entry): New.
1199 (dwarf2out_maybe_output_loclist_view_pair): New.
1200 (output_loc_list): Output view list or entries too.
1201 (output_view_list_offset): New.
1202 (output_die): Handle dw_val_class_view_list.
1203 (output_dwarf_version): New.
1204 (output_compilation_unit_header): Use it.
1205 (output_skeleton_debug_sections): Likewise.
1206 (output_rnglists, output_line_info): Likewise.
1207 (output_pubnames, output_aranges): Update version comments.
1208 (output_one_line_info_table): Output view numbers in asm comments.
1209 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1210 Call maybe_gen_llsym.
1211 (loc_list_from_tree_1): Adjust.
1212 (add_AT_location_description): Create view list attribute if
1213 needed, check it's absent otherwise.
1214 (convert_cfa_to_fb_loc_list): Adjust.
1215 (maybe_emit_file): Call output_asm_line_debug_info for test.
1216 (dwarf2out_var_location): Reset views as needed. Precompute
1217 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1218 attribute. Set view.
1219 (new_line_info_table): Reset next view.
1220 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1221 (dwarf2out_source_line): Likewise. Output view resets and labels to
1222 the assembler, or select appropriate line info opcodes.
1223 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1224 (optimize_string_length): Catch it. Adjust.
1225 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1226 dw_val_class_view_list, and remove it if no longer needed.
1227 (hash_loc_list): Hash view numbers.
1228 (loc_list_hasher::equal): Compare them.
1229 (optimize_location_lists): Check whether a view list symbol is
1230 needed, and whether the locview attribute is present, and
1231 whether they match. Remove the locview attribute if no longer
1233 (index_location_lists): Call skip_loc_list_entry for test.
1234 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1235 Use output_dwarf_version.
1236 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1237 (struct dw_val_node): Add val_view_list.
1238 * final.c (SEEN_NEXT_VIEW): New.
1239 (set_next_view_needed): New.
1240 (clear_next_view_needed): New.
1241 (maybe_output_next_view): New.
1242 (final_start_function): Rename to...
1243 (final_start_function_1): ... this. Take pointer to FIRST,
1244 add SEEN parameter. Emit param bindings in the initial view.
1245 (final_start_function): Reintroduce SEEN-less interface.
1246 (final): Rename to...
1247 (final_1): ... this. Take SEEN parameter. Output final pending
1248 next view at the end.
1249 (final): Reintroduce seen-less interface.
1250 (final_scan_insn): Output pending next view before switching
1251 sections or ending a block. Mark the next view as needed when
1252 outputting variable locations. Notify debug backend of section
1253 changes, and of location view changes.
1254 (rest_of_handle_final): Adjust.
1255 * toplev.c (process_options): Autodetect value for debug variable
1256 location views option. Warn on incompat5 without -gdwarf-5.
1257 * doc/invoke.texi (gvariable-location-views): New.
1258 (gvariable-location-views=incompat5): New.
1259 (gno-variable-location-views): New.
1261 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1263 PR tree-optimization/84136
1264 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1265 that the result of find_edge is non-NULL.
1267 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1270 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1271 storing integer register in SImode. Fix cost of 256 and 512
1272 byte aligned SSE register store.
1274 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1276 * config/i386/i386.c (ix86_multiplication_cost): Fix
1277 multiplication cost for TARGET_AVX512DQ.
1279 2018-02-08 Marek Polacek <polacek@redhat.com>
1281 PR tree-optimization/84238
1282 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1285 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1287 PR tree-optimization/84265
1288 * tree-vect-stmts.c (vectorizable_store): Don't treat
1289 VMAT_CONTIGUOUS accesses as grouped.
1290 (vectorizable_load): Likewise.
1292 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1294 PR tree-optimization/81635
1295 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1296 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1297 (test_round_for_mask): New functions.
1298 (wide_int_cc_tests): Call test_round_for_mask.
1299 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1300 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1301 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1302 range returned by get_range_info.
1304 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1307 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1308 * symtab.c: Include builtins.h
1309 (symtab_node::output_to_lto_symbol_table_p): Move here
1310 from lto-streamer-out.c:output_symbol_p.
1311 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1312 (output_symbol_p): Move all logic to symtab.c
1313 (produce_symtab): Update.
1315 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1317 * config/s390/s390-opts.h (enum indirect_branch): Define.
1318 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1319 (s390_indirect_branch_via_thunk)
1320 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1321 (enum s390_indirect_branch_type): Define.
1322 * config/s390/s390.c (struct s390_frame_layout, struct
1323 machine_function): Remove.
1324 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1325 (indirect_branch_table_label_no, indirect_branch_table_name):
1327 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1328 (enum s390_indirect_branch_option): Define.
1329 (s390_return_addr_from_memory): New function.
1330 (s390_handle_string_attribute): New function.
1331 (s390_attribute_table): Add new attribute handler.
1332 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1333 (s390_indirect_branch_via_thunk): New function.
1334 (s390_indirect_branch_via_inline_thunk): New function.
1335 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1336 sibling call optimization for non z10 compiles.
1337 (s390_emit_call): Force indirect branch target to be a single
1338 register. Add r1 clobber for non-z10 compiles.
1339 (s390_emit_epilogue): Emit return jump via return_use expander.
1340 (s390_reorg): Handle JUMP_INSNs as execute targets.
1341 (s390_option_override_internal): Perform validity checks for the
1342 new command line options.
1343 (s390_indirect_branch_attrvalue): New function.
1344 (s390_indirect_branch_settings): New function.
1345 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1346 (s390_output_indirect_thunk_function): New function.
1347 (s390_code_end): Implement target hook.
1348 (s390_case_values_threshold): Implement target hook.
1349 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1351 * config/s390/s390.h (struct s390_frame_layout)
1352 (struct machine_function): Move here from s390.c.
1353 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1354 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1355 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1356 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1357 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1358 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1359 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1360 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1361 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1362 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1363 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1364 (mnemonic attribute): Add values which aren't recognized
1366 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1367 pattern for branch conversion. Fix mnemonic attribute.
1368 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1369 indirect branch via thunk if requested.
1370 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1371 ("*indirect_jump"): Disable for branch conversion using out of
1373 ("indirect_jump_via_thunk<mode>_z10")
1374 ("indirect_jump_via_thunk<mode>")
1375 ("indirect_jump_via_inlinethunk<mode>_z10")
1376 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1377 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1378 ("casesi_jump_via_inlinethunk<mode>_z10")
1379 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1380 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1381 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1382 ("*indirect2_jump"): Disable for branch conversion.
1383 ("casesi_jump"): Turn into expander and expand patterns for branch
1385 ("return_use"): New expander.
1386 ("*return"): Emit return via thunk and rename it to ...
1387 ("*return<mode>"): ... this one.
1388 * config/s390/s390.opt: Add new options and and enum for the
1391 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1393 * lra-constraints.c (match_reload): Unconditionally use
1394 gen_lowpart_SUBREG, rather than selecting between that
1395 and equivalent gen_rtx_SUBREG code.
1397 2018-02-08 Richard Biener <rguenther@suse.de>
1399 PR tree-optimization/84233
1400 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1401 changed flag instead of boguously re-using phi_inserted.
1403 2018-02-08 Martin Jambor <mjambor@suse.cz>
1405 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1406 static local variables.
1408 2018-02-08 Richard Biener <rguenther@suse.de>
1410 PR tree-optimization/84278
1411 * tree-vect-stmts.c (vectorizable_store): When looking for
1412 smaller vector types to perform grouped strided loads/stores
1413 make sure the mode is supported by the target.
1414 (vectorizable_load): Likewise.
1416 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1418 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1419 Increase LDP/STP opportunities by adding adjacent callee-saves.
1421 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1423 PR rtl-optimization/84068
1424 PR rtl-optimization/83459
1425 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1427 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1429 PR tree-optimization/84224
1430 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1431 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1434 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1437 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1438 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1439 count, remove one USE.
1441 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1443 * doc/install.texi (Configuration): Document the
1444 --with-long-double-format={ibm,ieee} PowerPC configuration
1448 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1449 Convert from define_expand to be define_insn_and_split. Rework
1450 float/double/_Float128 conversions to QI/HI/SImode to work with
1451 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1452 conversions to QI/HImode types did a store and then a load to
1453 truncate the value. For conversions to VSX registers, don't split
1454 the insn, instead emit the code directly. Use the code iterator
1455 any_fix to combine signed and unsigned conversions.
1456 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1457 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1458 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1459 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1460 (fix_<mode>di2_hw): Likewise.
1461 (fixuns_<mode>di2_hw): Likewise.
1462 (fix_<mode>si2_hw): Likewise.
1463 (fixuns_<mode>si2_hw): Likewise.
1464 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1465 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1466 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1467 fix<uns>_trunc<SFDF:mode>si2_p8.
1468 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1470 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1471 (fix<uns>_<mode>_mem): Likewise.
1472 (fctiw<u>z_<mode>_mem): Likewise.
1473 (fix<uns>_<mode>_mem): Likewise.
1474 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1475 the register allocator from doing a direct move to the GPRs to do
1476 a store, and instead use the ISA 3.0 store byte/half-word from
1477 vector register instruction. For IEEE 128-bit floating point,
1478 also optimize stores of 32-bit ints.
1479 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1481 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1483 * genextract.c (push_pathstr_operand): New function to support
1485 (walk_rtx): Call push_pathstr_operand.
1486 (print_path): Support [a-zA-Z].
1488 2018-02-07 Richard Biener <rguenther@suse.de>
1490 PR tree-optimization/84037
1491 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
1492 (cse_and_gimplify_to_preheader): Declare.
1493 (vect_get_place_in_interleaving_chain): Likewise.
1494 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1496 (_loop_vec_info::~_loop_vec_info): Delete it.
1497 (cse_and_gimplify_to_preheader): New function.
1498 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
1499 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
1500 (vectorizable_load): Likewise. For grouped stores always base
1501 the IV on the first element.
1502 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
1503 condition before gimplifying.
1505 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1507 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
1508 *DIV_EXPR and *MOD_EXPR.
1510 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
1513 * config/i386/i386.c (ix86_option_override_internal): Mask out
1514 the CF_SET bit when checking -fcf-protection.
1516 2018-02-07 Tom de Vries <tom@codesourcery.com>
1519 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
1522 2018-02-07 Richard Biener <rguenther@suse.de>
1524 PR tree-optimization/84204
1525 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1528 PR tree-optimization/84205
1529 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
1530 special-case isl_ast_op_zdiv_r.
1532 PR tree-optimization/84223
1533 * graphite-scop-detection.c (gather_bbs::before_dom_children):
1534 Only add conditions from within the region.
1535 (gather_bbs::after_dom_children): Adjust.
1537 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
1540 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
1541 * config/avr/avr.md: Only post-reload split REG-REG moves if
1542 either register is REGERAL_REG_P.
1544 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1546 PR tree-optimization/84235
1547 * tree-ssa-scopedtables.c
1548 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
1549 if the subtraction is performed in floating point type where NaNs are
1550 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
1551 build 1. Formatting fix.
1553 2018-02-06 Jakub Jelinek <jakub@redhat.com>
1556 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
1557 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
1558 and skip it regardless of bb boundaries. Use CALL_P macro,
1559 don't test INSN_P (insn) together with CALL_P or JUMP_P check
1560 unnecessarily, formatting fix.
1562 2018-02-06 Michael Collison <michael.collison@arm.com>
1564 * config/arm/thumb2.md:
1565 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
1566 (*thumb_mov_notscc): Ditto.
1568 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
1571 * config/rs6000/rs6000.md (su code attribute): Use "u" for
1572 unsigned_fix, not "s".
1574 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1576 * configure.ac (gcc_fn_eh_frame_ro): New function.
1577 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
1578 correct .eh_frame permissions.
1579 * configure: Regenerate.
1581 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
1583 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
1586 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1588 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1589 Display warning message for -mno-speculate-indirect-jumps.
1591 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
1593 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
1595 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
1597 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
1599 PR tree-optimization/84225
1600 * tree-eh.c (find_trapping_overflow): Only call
1601 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
1603 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1606 * config/i386/i386.c: Reimplement the check of possible options
1607 -mibt/-mshstk conbination. Change error messages.
1608 * doc/invoke.texi: Fix a typo: remove extra '='.
1610 2018-02-06 Marek Polacek <polacek@redhat.com>
1612 PR tree-optimization/84228
1613 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
1615 2018-02-06 Tamar Christina <tamar.christina@arm.com>
1618 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
1619 emitted arch directives.
1620 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
1621 __ARM_FEATURE_COPROC before changing architectures.
1623 2018-02-06 Richard Biener <rguenther@suse.de>
1625 * config/i386/i386.c (print_reg): Fix typo.
1626 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
1628 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
1630 * configure: Regenerate.
1632 2018-02-05 Martin Sebor <msebor@redhat.com>
1634 PR tree-optimization/83369
1635 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
1638 2018-02-05 Martin Liska <mliska@suse.cz>
1640 * doc/invoke.texi: Cherry-pick upstream r323995.
1642 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
1644 * ira.c (ira_init_register_move_cost): Adjust comment.
1646 2018-02-05 Martin Liska <mliska@suse.cz>
1648 PR gcov-profile/84137
1649 * doc/gcov.texi: Fix typo in documentation.
1651 2018-02-05 Martin Liska <mliska@suse.cz>
1653 PR gcov-profile/83879
1654 * doc/gcov.texi: Document necessity of --dynamic-list-data when
1655 using dlopen functionality.
1657 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
1659 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
1660 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
1661 _mm_maskz_range_ss, _mm_mask_range_round_ss,
1662 _mm_maskz_range_round_ss): New intrinsics.
1663 (__builtin_ia32_rangesd128_round)
1664 (__builtin_ia32_rangess128_round): Remove.
1665 (__builtin_ia32_rangesd128_mask_round,
1666 __builtin_ia32_rangess128_mask_round): New builtins.
1667 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
1668 __builtin_ia32_rangess128_round): Remove.
1669 (__builtin_ia32_rangesd128_mask_round,
1670 __builtin_ia32_rangess128_mask_round): New builtins.
1671 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
1672 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
1673 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
1674 "<round_saeonly_constraint>")): Changed to ...
1675 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
1676 "<round_saeonly_scalar_constraint>")): ... this.
1677 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
1678 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
1679 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
1680 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
1681 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
1683 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
1685 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
1687 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
1688 Remove all values except native, 8540 and 8548.
1690 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
1692 * config/i386/i386.c (ix86_output_function_return): Pass
1693 INVALID_REGNUM, instead of -1, as invalid register number to
1694 indirect_thunk_name and output_indirect_thunk.
1696 2018-02-02 Julia Koval <julia.koval@intel.com>
1698 * config.gcc: Add -march=icelake.
1699 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
1700 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
1701 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
1702 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
1703 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
1704 (processor_target_table): Add icelake.
1705 (ix86_option_override_internal): Handle new PTAs.
1706 (get_builtin_code_for_version): Handle icelake.
1707 (M_INTEL_COREI7_ICELAKE): New.
1708 (fold_builtin_cpu): Handle icelake.
1709 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
1710 * doc/invoke.texi: Add -march=icelake.
1712 2018-02-02 Julia Koval <julia.koval@intel.com>
1714 * config/i386/i386.c (ix86_option_override_internal): Change flags type
1715 to wide_int_bitmask.
1716 * wide-int-bitmask.h: New.
1718 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1721 * config/i386/i386.md: Replace Pmode with word_mode in
1722 builtin_setjmp_setup and builtin_longjmp to support x32.
1724 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
1728 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
1730 (rs6000_supported_cpu_names): New static variable.
1731 (linux_cpu_translation_table): Likewise.
1732 (elf_platform) <cpu>: Define new static variable and use it.
1733 Translate kernel AT_PLATFORM name to canonical name if needed.
1734 Error if platform name is unknown.
1736 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
1739 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
1741 2018-02-01 Jeff Law <law@redhat.com>
1744 * config/i386/i386.c (release_scratch_register_on_entry): Add new
1745 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
1746 the scratch if RELEASE_VIA_POP is false.
1747 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
1748 If we have to save a temporary register, decrement SIZE appropriately.
1749 Pass new arguments to release_scratch_register_on_entry.
1750 (ix86_adjust_stack_and_probe): Likewise.
1751 (ix86_emit_probe_stack_range): Pass new arguments to
1752 release_scratch_register_on_entry.
1754 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
1756 PR rtl-optimization/84157
1757 * combine.c (change_zero_ext): Use REG_P predicate in
1758 front of HARD_REGISTER_P predicate.
1760 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
1762 * config/avr/avr.c (avr_option_override): Move disabling of
1763 -fdelete-null-pointer-checks to...
1764 * common/config/avr/avr-common.c (avr_option_optimization_table):
1767 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1769 PR tree-optimization/81635
1770 * tree-data-ref.c (split_constant_offset_1): For types that
1771 wrap on overflow, try to use range info to prove that wrapping
1774 2018-02-01 Renlin Li <renlin.li@arm.com>
1777 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
1779 (aarch64_register_move_cost): Likewise.
1780 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
1782 (REG_CLASS_NAMES): Likewise.
1783 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
1784 TAILCALL_ADDR_REGS. Remove IP registers.
1785 * config/aarch64/aarch64.md (Ucs): Update register constraint.
1787 2018-02-01 Richard Biener <rguenther@suse.de>
1789 * domwalk.h (dom_walker::dom_walker): Add additional constructor
1790 for specifying RPO order and allow NULL for that.
1791 * domwalk.c (dom_walker::dom_walker): Likewise.
1792 (dom_walker::walk): Handle NULL RPO order.
1793 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
1795 (rewrite_update_dom_walker): Likewise.
1796 (mark_def_dom_walker): Likewise.
1798 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1800 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
1801 (aarch64_maybe_expand_sve_subreg_move): Declare.
1802 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
1803 * config/aarch64/predicates.md (aarch64_any_register_operand): New
1805 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
1806 that are semantically a reverse operation.
1807 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
1808 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
1809 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
1811 (aarch64_can_change_mode_class): For big-endian, forbid changes
1812 between two SVE modes if they have different element sizes.
1814 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1816 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
1817 the TImode handling for big-endian targets.
1819 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1821 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
1822 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
1824 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
1825 Remove BSWAP handing for big-endian targets and use the form of
1826 LD1RQ appropariate for the mode.
1828 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1830 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
1831 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
1834 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1837 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
1838 check for operands that need to go through aarch64_sve_reload_be.
1840 2018-02-01 Jakub Jelinek <jakub@redhat.com>
1842 PR tree-optimization/81661
1843 PR tree-optimization/84117
1844 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
1845 * tree-eh.c: Include gimplify.h.
1846 (find_trapping_overflow, replace_trapping_overflow,
1847 rewrite_to_non_trapping_overflow): New functions.
1848 * tree-vect-loop.c: Include tree-eh.h.
1849 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
1850 * tree-data-ref.c: Include tree-eh.h.
1851 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
1853 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
1855 PR rtl-optimization/84123
1856 * combine.c (change_zero_ext): Check if hard register satisfies
1857 can_change_dest_mode before calling gen_lowpart_SUBREG.
1859 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
1862 * ira.c (ira_init_register_move_cost): Remove assert.
1864 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
1866 PR rtl-optimization/84071
1867 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
1868 * doc/tm.texi: Regenerate.
1870 2018-01-31 Richard Biener <rguenther@suse.de>
1872 PR tree-optimization/84132
1873 * tree-data-ref.c (analyze_miv_subscript): Properly
1874 check whether evolution_function_is_affine_multivariate_p
1875 before calling gcd_of_steps_may_divide_p.
1877 2018-01-31 Julia Koval <julia.koval@intel.com>
1880 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
1881 * config/i386/i386.md (rdpid_rex64) New.
1882 (rdpid): Make 32bit only.
1884 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
1887 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
1888 an IDENTIFIER_NODE for FUNCTION_TYPE's.
1890 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
1893 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
1895 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
1897 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
1899 PR rtl-optimization/84071
1900 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
1901 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
1903 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
1905 * config/arc/arc.c (arc_handle_aux_attribute): New function.
1906 (arc_attribute_table): Add 'aux' attribute.
1907 (arc_in_small_data_p): Consider aux like variables.
1908 (arc_is_aux_reg_p): New function.
1909 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
1910 (arc_get_aux_arg): New function.
1911 (prepare_move_operands): Handle aux-register access.
1912 (arc_handle_aux_attribute): New function.
1913 * doc/extend.texi (ARC Variable attributes): Add subsection.
1915 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
1917 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
1918 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
1919 (arc_attribute_table): Add 'uncached' attribute.
1920 (arc_print_operand): Print '.di' flag for uncached memory
1922 (arc_in_small_data_p): Do not consider for small data the uncached
1924 (arc_is_uncached_mem_p): New function.
1925 * config/arc/predicates.md (compact_store_memory_operand): Check
1926 for uncached memory accesses.
1927 (nonvol_nonimm_operand): Likewise.
1928 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
1930 2018-01-31 Jakub Jelinek <jakub@redhat.com>
1933 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
1934 falign-loops=): Add Optimization flag.
1936 2018-01-30 Jeff Law <law@redhat.com>
1939 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
1940 INT_REGISTERS_SAVED. Check it prior to calling
1941 get_scratch_register_on_entry.
1942 (ix86_adjust_stack_and_probe): Similarly.
1943 (ix86_emit_probe_stack_range): Similarly.
1944 (ix86_expand_prologue): Corresponding changes.
1946 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1949 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
1950 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
1952 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
1955 * lra-constraints.c (curr_insn_transform): Process AND in the
1958 2018-01-30 Jakub Jelinek <jakub@redhat.com>
1960 PR rtl-optimization/83986
1961 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
1962 dependence against last_pending_memory_flush in addition to
1965 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
1967 PR tree-optimization/81611
1968 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
1971 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1974 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
1977 2018-01-30 Richard Biener <rguenther@suse.de>
1978 Jakub Jelinek <jakub@redhat.com>
1980 PR tree-optimization/84111
1981 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
1982 inner loops added during recursion, as they don't have up-to-date
1985 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
1988 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
1989 (can_inline_edge_by_limits_p): ... here.
1990 (can_early_inline_edge_p, check_callers,
1991 update_caller_keys, update_callee_keys, recursive_inlining,
1992 add_new_edges_to_heap, speculation_useful_p,
1993 inline_small_functions,
1994 inline_small_functions, flatten_function,
1995 inline_to_all_callers_1): Update.
1997 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
1999 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2002 2018-01-30 Richard Biener <rguenther@suse.de>
2004 PR tree-optimization/83008
2005 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2006 invariant and constant vector uses in stmts when they need
2009 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2012 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2013 * configure: Regenerate.
2015 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2017 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2019 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2020 Use gen_rtx_REG rather than gen_lowpart.
2022 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2024 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2025 rather than 0 when creating partial subregs.
2027 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2029 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2032 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2035 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2036 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2037 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2038 flags. This restores the settings used before the 2017-07-24.
2039 Turning off pre increment/decrement/modify allows IVOPTS to
2040 optimize DF/SF loops where the index is an int.
2042 2018-01-29 Richard Biener <rguenther@suse.de>
2043 Kelvin Nilsen <kelvin@gcc.gnu.org>
2046 * tree-vect-stmts.c (vectorizable_call): Don't call
2047 targetm.vectorize_builtin_md_vectorized_function if callee is
2050 2018-01-22 Carl Love <cel@us.ibm.com>
2052 * doc/extend.tex: Fix typo in second arg in
2053 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2055 2018-01-29 Richard Biener <rguenther@suse.de>
2057 PR tree-optimization/84086
2058 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2059 (flush_ssaname_freelist): When SSA names were released reset
2060 the SCEV hash table.
2062 2018-01-29 Richard Biener <rguenther@suse.de>
2064 PR tree-optimization/84057
2065 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2066 removed paths when removing edges.
2068 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2070 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2071 -mfunction-return=@var{choice}.
2073 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2076 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2078 (layout::print_source_line): Likewise.
2079 (test_get_line_width_without_trailing_whitespace): Add test cases.
2081 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2084 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2087 2018-01-26 Jim Wilson <jimw@sifive.com>
2089 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2091 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2094 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2096 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2097 and CMP + SUB-immediate -> SUBS.
2099 2018-01-26 Martin Sebor <msebor@redhat.com>
2101 PR tree-optimization/83896
2102 * tree-ssa-strlen.c (get_string_len): Rename...
2103 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2104 Avoid assuming length is constant.
2105 (handle_char_store): Use HOST_WIDE_INT for string length.
2107 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2110 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2111 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2113 2018-01-26 Richard Biener <rguenther@suse.de>
2115 PR rtl-optimization/84003
2116 * dse.c (record_store): Only record redundant stores when
2117 the earlier store aliases at least all accesses the later one does.
2119 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2121 PR rtl-optimization/83985
2122 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2123 REG_CFA_RESTORE insns.
2124 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2125 REG_CFA_RESTORE insns here.
2128 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2129 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2131 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2133 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2134 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2135 (arc_init): Likewise.
2136 (arc_override_options): Likewise.
2137 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2139 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2141 * config/arc/arc.h (TARGET_DBNZ): Define.
2142 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2143 properly set the tune attribute.
2144 (dbnz): Use TARGET_DBNZ guard.
2145 * config/arc/arc.opt (mtune): Add core3 option.
2147 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2149 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2150 recognize new pic like addresses.
2151 (arc_delegitimize_address): Clean up.
2153 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2155 * config/arc/arc-arches.def: Option mrf16 valid for all
2157 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2158 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2159 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2160 * config/arc/arc-tables.opt: Regenerate.
2161 * config/arc/arc.c (arc_conditional_register_usage): Handle
2162 reduced register file case.
2163 (arc_file_start): Set must have build attributes.
2164 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2166 * config/arc/arc.opt (mrf16): Add new option.
2167 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2168 * config/arc/genmultilib.awk: Handle new mrf16 option.
2169 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2170 * config/arc/t-multilib: Regenerate.
2171 * doc/invoke.texi (ARC Options): Document mrf16 option.
2173 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2175 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2176 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2177 (arc_attribute_table): Add 'secure_call' attribute.
2178 (arc_print_operand): Print secure call operand.
2179 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2181 (arc_is_secure_call_p): New function. * config/arc/arc.md
2182 (call_i): Add support for sjli instruction.
2183 (call_value_i): Likewise.
2184 * config/arc/constraints.md (Csc): New constraint.
2186 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2187 John Eric Martin <John.Martin@emmicro-us.com>
2189 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2190 * config/arc/arc.c (_arc_jli_section): New struct.
2191 (arc_jli_section): New type.
2192 (rc_jli_sections): New static variable.
2193 (arc_handle_jli_attribute): New function.
2194 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2195 (arc_file_end): New function.
2196 (TARGET_ASM_FILE_END): Define.
2197 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2198 (arc_add_jli_section): New function.
2199 (jli_call_scan): Likewise.
2200 (arc_reorg): Call jli_call_scan.
2201 (arc_output_addsi): Remove 'S' from printing asm operand.
2202 (arc_is_jli_call_p): New function.
2203 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2205 (movhi_insn): Likewise.
2206 (movsi_insn): Likewise.
2207 (movsi_set_cc_insn): Likewise.
2208 (loadqi_update): Likewise.
2209 (load_zeroextendqisi_update): Likewise.
2210 (load_signextendqisi_update): Likewise.
2211 (loadhi_update): Likewise.
2212 (load_zeroextendhisi_update): Likewise.
2213 (load_signextendhisi_update): Likewise.
2214 (loadsi_update): Likewise.
2215 (loadsf_update): Likewise.
2216 (movsicc_insn): Likewise.
2217 (bset_insn): Likewise.
2218 (bxor_insn): Likewise.
2219 (bclr_insn): Likewise.
2220 (bmsk_insn): Likewise.
2221 (bicsi3_insn): Likewise.
2222 (cmpsi_cc_c_insn): Likewise.
2223 (movsi_ne): Likewise.
2224 (movsi_cond_exec): Likewise.
2225 (clrsbsi2): Likewise.
2236 (call_i): Remove 'S' asm letter, add jli instruction.
2237 (call_value_i): Likewise.
2238 * config/arc/arc.op (mjli-always): New option.
2239 * config/arc/constraints.md (Cji): New constraint.
2240 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2242 (subsf3_fpx): Likewise.
2243 (mulsf3_fpx): Likewise.
2244 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2246 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2248 * doc/invoke.texi (ARC): Document mjli-always option.
2250 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2252 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2253 avoid addition with 0 and use incw and decw where possible.
2255 2018-01-26 Richard Biener <rguenther@suse.de>
2257 PR tree-optimization/81082
2258 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2259 association if it requires casting to unsigned.
2260 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2261 from fold_plusminus_mult_expr to catch important cases late when
2262 range info is available.
2264 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2266 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2267 * configure.ac (hidden_linkonce): New test.
2268 * configure: Regenerate.
2269 * config.in: Regenerate.
2271 2018-01-26 Julia Koval <julia.koval@intel.com>
2273 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2274 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2275 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2276 _mm_mask_bitshuffle_epi64_mask): Fix type.
2277 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2278 USI_FTYPE_V4DI_V4DI_USI): Remove.
2279 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2280 __builtin_ia32_vpshufbitqmb256_mask,
2281 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2282 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2283 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2285 2018-01-26 Alan Modra <amodra@gmail.com>
2288 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2289 UNSPEC_VBPERMQ. Sort other unspecs.
2291 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2293 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2295 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2298 * predict.c (drop_profile): Do not push/pop cfun; update also
2300 (handle_missing_profiles): Fix logic looking for zero profiles.
2302 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2305 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2306 on functions with #pragma omp declare simd or functions with simd
2308 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2309 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2310 Remove trailing \n from warning_at calls.
2312 2018-01-25 Tom de Vries <tom@codesourcery.com>
2315 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2316 for neutered workers.
2318 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2321 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2322 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2324 2018-01-24 Jeff Law <law@redhat.com>
2327 * i386.c (get_probe_interval): Move to earlier point.
2328 (ix86_compute_frame_layout): If -fstack-clash-protection and
2329 the frame is larger than the probe interval, then use pushes
2330 to save registers rather than reg->mem moves.
2331 (ix86_expand_prologue): Remove conditional for int_registers_saved
2334 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2337 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2338 min/max for never referenced object.
2340 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2343 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2345 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2346 attributes from DECL_ATTRIBUTES (decl) without affecting
2347 DECL_ATTRIBUTES (current_function_decl).
2348 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2349 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2351 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2353 PR tree-optimization/83979
2354 * fold-const.c (fold_comparison): Use constant_boolean_node
2355 instead of boolean_{true,false}_node.
2357 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2359 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2362 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2364 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2365 Simplify the clause that sets the length attribute.
2366 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2367 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2368 clause that sets the length attribute.
2369 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2371 2018-01-24 Tom de Vries <tom@codesourcery.com>
2374 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2375 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2376 Add strict parameter.
2377 (prevent_branch_around_nothing): Insert dummy insn between branch to
2378 label and label with no ptx insn inbetween.
2379 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2381 2018-01-24 Tom de Vries <tom@codesourcery.com>
2384 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2385 for neutered threads in warp.
2386 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2388 2018-01-24 Richard Biener <rguenther@suse.de>
2390 PR tree-optimization/83176
2391 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2394 2018-01-24 Richard Biener <rguenther@suse.de>
2396 PR tree-optimization/82819
2397 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2398 code generating pluses that are no-ops in the target precision.
2400 2018-01-24 Richard Biener <rguenther@suse.de>
2403 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2405 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2407 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2408 to merge probabilities.
2409 * predict.c (probably_never_executed): Also mark as cold functions
2410 with global 0 profile and guessed local profile.
2411 * profile-count.c (profile_probability::combine_with_count): New
2413 * profile-count.h (profile_probability::operator*,
2414 profile_probability::operator*=, profile_probability::operator/,
2415 profile_probability::operator/=): Reduce precision to adjusted
2416 and set value to guessed on contradictory divisions.
2417 (profile_probability::combine_with_freq): Remove.
2418 (profile_probability::combine_wiht_count): Declare.
2419 (profile_count::force_nonzero):: Set to adjusted.
2420 (profile_count::probability_in):: Set quality to adjusted.
2421 * tree-ssa-tail-merge.c (replace_block_by): Use
2424 2018-01-23 Andrew Waterman <andrew@sifive.com>
2425 Jim Wilson <jimw@sifive.com>
2427 * config/riscv/riscv.c (riscv_stack_boundary): New.
2428 (riscv_option_override): Set riscv_stack_boundary. Handle
2429 riscv_preferred_stack_boundary_arg.
2430 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2431 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2432 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2433 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2434 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2435 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2437 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2440 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2441 of struct ix86_frame.
2442 (ix86_expand_epilogue): Likewise. Add a local variable for
2443 the reg_save_offset field in struct ix86_frame.
2445 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2447 PR tree-optimization/82604
2448 * tree-loop-distribution.c (enum partition_kind): New enum item
2449 PKIND_PARTIAL_MEMSET.
2450 (partition_builtin_p): Support above new enum item.
2451 (generate_code_for_partition): Ditto.
2452 (compute_access_range): Differentiate cases that equality can be
2453 proven at all loops, the innermost loops or no loops.
2454 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2455 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2456 (finalize_partitions, distribute_loop): Don't fuse partition of
2457 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2458 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2461 2018-01-23 Martin Liska <mliska@suse.cz>
2463 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2464 order to ignore the predictor.
2465 (PRED_POLYMORPHIC_CALL): Likewise.
2466 (PRED_RECURSIVE_CALL): Likewise.
2468 2018-01-23 Martin Liska <mliska@suse.cz>
2470 * tree-profile.c (tree_profiling): Print function header to
2471 aware reader which function we are working on.
2472 * value-prof.c (gimple_find_values_to_profile): Do not print
2473 not interesting value histograms.
2475 2018-01-23 Martin Liska <mliska@suse.cz>
2477 * profile-count.h (enum profile_quality): Add
2478 profile_uninitialized as the first value. Do not number values
2479 as they are zero based.
2480 (profile_count::verify): Update sanity check.
2481 (profile_probability::verify): Likewise.
2483 2018-01-23 Nathan Sidwell <nathan@acm.org>
2485 * doc/invoke.texi (ffor-scope): Deprecate.
2487 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2489 PR tree-optimization/83510
2490 * domwalk.c (set_all_edges_as_executable): New function.
2491 (dom_walker::dom_walker): Convert bool param
2492 "skip_unreachable_blocks" to enum reachability. Move setup of
2493 edge flags to set_all_edges_as_executable and only do it when
2494 reachability is REACHABLE_BLOCKS.
2495 * domwalk.h (enum dom_walker::reachability): New enum.
2496 (dom_walker::dom_walker): Convert bool param
2497 "skip_unreachable_blocks" to enum reachability.
2498 (set_all_edges_as_executable): New decl.
2499 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
2500 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
2502 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
2503 but converting true to REACHABLE_BLOCKS.
2504 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
2506 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
2507 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
2508 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
2510 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
2511 if check_all_array_refs will be called.
2513 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2515 * tree.c (selftest::test_location_wrappers): Add more test
2518 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2520 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
2521 (selftest::test_bit_in_range): Likewise.
2523 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2526 * doc/sourcebuild.texi (vect_float): Say that the selector
2527 only describes the situation when -funsafe-math-optimizations is on.
2528 (vect_float_strict): Document.
2530 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2532 PR tree-optimization/83965
2533 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2534 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
2535 instead of checking only for a reduction.
2536 (vect_recog_widen_sum_pattern): Likewise.
2538 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2540 * predict.c (probably_never_executed): Only use precise profile info.
2541 (compute_function_frequency): Skip after inlining hack since we now
2542 have quality checking.
2544 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2546 * profile-count.h (profile_probability::very_unlikely,
2547 profile_probability::unlikely, profile_probability::even): Set
2548 precision to guessed.
2550 2018-01-23 Richard Biener <rguenther@suse.de>
2552 PR tree-optimization/83963
2553 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
2554 Properly terminate dominator walk when crossing the exit edge not
2555 when visiting its source block.
2557 2018-01-23 Jakub Jelinek <jakub@redhat.com>
2560 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
2561 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
2563 2018-01-22 Jakub Jelinek <jakub@redhat.com>
2565 PR tree-optimization/83957
2566 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
2567 semicolon after for body surrounded by braces.
2569 PR tree-optimization/83081
2570 * profile-count.h (profile_probability::split): New method.
2571 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
2572 Use profile_probability::split.
2573 (do_compare_rtx_and_jump): Fix adjustment of probabilities
2574 when splitting a single conditional jump into 2.
2576 2018-01-22 David Malcolm <dmalcolm@redhat.com>
2578 PR tree-optimization/69452
2579 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
2582 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2584 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
2585 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
2586 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
2588 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2590 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
2591 * config/rl78/rl78.md (movdi): New define_expand.
2592 * config/rl78/rl78.c (rl78_split_movdi): New function.
2594 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
2597 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
2599 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
2600 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
2601 128-bit to produce an UNSPEC move to get the double word with the
2602 signbit and then a shift directly to do signbit.
2603 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
2604 implementation with a new version that just does either a direct
2605 move or a regular move. Move memory interface to separate insns.
2606 Move insns so they are next to the expander.
2607 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
2608 with signbit move. Split big and little endian case.
2609 (signbit<mode>2_dm_mem_le): Likewise.
2610 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
2611 (signbit<mode>2_dm2): Likewise.
2613 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2615 * config/rl78/rl78.md (anddi3): New define_expand.
2617 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2619 * config/rl78/rl78.md (umindi3): New define_expand.
2621 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2623 * config/rl78/rl78.md (smindi3): New define_expand.
2625 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2627 * config/rl78/rl78.md (smaxdi3): New define_expand.
2629 2018-01-22 Carl Love <cel@us.ibm.com>
2631 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
2632 LVX_V1TI): Add macro expansion.
2633 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
2634 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
2635 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
2636 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
2637 Change check to determine if the instruction is a byte reversing
2638 entry. Fix typo in comment.
2639 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
2640 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
2641 Add def_builtin calls for new builtins.
2642 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
2643 Add define_insn expansion.
2645 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2647 * config/rl78/rl78.md (umaxdi3): New define_expand.
2649 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2651 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
2652 for non-QImode registers.
2654 2018-01-22 Richard Biener <rguenther@suse.de>
2656 PR tree-optimization/83963
2657 * graphite-scop-detection.c (scop_detection::get_sese): Delay
2658 including the loop exit block.
2659 (scop_detection::merge_sese): Likewise.
2660 (scop_detection::add_scop): Do it here instead.
2662 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2664 * doc/sourcebuild.texi (arm_softfloat): Document.
2666 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
2669 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
2670 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
2671 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
2673 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2674 David Edelsohn <dje.gcc@gmail.com>
2677 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2678 Change "crset eq" to "crset 2".
2679 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2680 (*call_indirect_aix<mode>_nospec): Likewise.
2681 (*call_value_indirect_aix<mode>_nospec): Likewise.
2682 (*call_indirect_elfv2<mode>_nospec): Likewise.
2683 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
2684 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
2685 change assembly output from . to $.
2686 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2687 (indirect_jump<mode>_nospec): Change assembly output from . to $.
2688 (*tablejump<mode>_internal1_nospec): Likewise.
2690 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
2693 * config/sh/sh_optimize_sett_clrt.cc:
2694 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
2696 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
2698 PR tree-optimization/83940
2699 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
2700 offset_dt to vect_constant_def rather than vect_unknown_def_type.
2701 (vect_check_load_store_mask): Add a mask_dt_out parameter and
2702 use it to pass back the definition type.
2703 (vect_check_store_rhs): Likewise rhs_dt_out.
2704 (vect_build_gather_load_calls): Add a mask_dt argument and use
2705 it instead of a call to vect_is_simple_use.
2706 (vectorizable_store): Update calls to vect_check_load_store_mask
2707 and vect_check_store_rhs. Use the dt returned by the latter instead
2708 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
2709 instead of calls to vect_is_simple_use. Pass the scalar rather
2710 than the vector operand to vect_is_simple_use when handling
2711 second and subsequent copies of an rhs value.
2712 (vectorizable_load): Update calls to vect_check_load_store_mask
2713 and vect_build_gather_load_calls. Use the cached mask_dt and
2714 gs_info.offset_dt instead of calls to vect_is_simple_use.
2716 2018-01-20 Jakub Jelinek <jakub@redhat.com>
2719 * tree-emutls.c: Include gimplify.h.
2720 (lower_emutls_2): New function.
2721 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
2722 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
2723 it before further processing.
2726 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
2727 UINTVAL (trueop1) instead of INTVAL (op1).
2729 2018-01-19 Jakub Jelinek <jakub@redhat.com>
2733 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
2734 INCOMING_FRAME_SP_OFFSET if not defined.
2735 (scan_trace): Add ENTRY argument. If true and
2736 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
2737 emit a note to adjust the CFA offset.
2738 (create_cfi_notes): Adjust scan_trace callers.
2739 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
2740 INCOMING_FRAME_SP_OFFSET in the CIE.
2741 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
2742 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
2744 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
2745 * doc/tm.texi: Regenerated.
2747 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2749 PR rtl-optimization/83147
2750 * lra-constraints.c (remove_inheritance_pseudos): Use
2751 lra_substitute_pseudo_within_insn.
2753 2018-01-19 Tom de Vries <tom@codesourcery.com>
2754 Cesar Philippidis <cesar@codesourcery.com>
2757 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
2759 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
2762 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
2763 spaces for function labels.
2765 2018-01-19 Martin Liska <mliska@suse.cz>
2767 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
2768 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
2769 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
2770 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
2771 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
2772 (PRED_CONST_RETURN): Change from 69 to 65.
2773 (PRED_NULL_RETURN): Change from 91 to 71.
2774 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
2775 (PRED_LOOP_GUARD): Change from 66 to 73.
2777 2018-01-19 Martin Liska <mliska@suse.cz>
2779 * predict.c (predict_insn_def): Add new assert.
2780 (struct branch_predictor): Change type to signed integer.
2781 (test_prediction_value_range): Amend test to cover
2783 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
2784 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
2785 (PRED_LOOP_ITERATIONS_MAX): Likewise.
2786 (PRED_LOOP_IV_COMPARE): Likewise.
2787 * predict.h (PROB_UNINITIALIZED): Define new constant.
2789 2018-01-19 Martin Liska <mliska@suse.cz>
2791 * predict.c (dump_prediction): Add new format for
2792 analyze_brprob.py script which is enabled with -details
2794 * profile-count.h (precise_p): New function.
2796 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
2798 PR tree-optimization/83922
2799 * tree-vect-loop.c (vect_verify_full_masking): Return false if
2800 there are no statements that need masking.
2801 (vect_active_double_reduction_p): New function.
2802 (vect_analyze_loop_operations): Use it when handling phis that
2803 are not in the loop header.
2805 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
2807 PR tree-optimization/83914
2808 * tree-vect-loop.c (vectorizable_induction): Don't convert
2809 init_expr or apply the peeling adjustment for inductions
2810 that are nested within the vectorized loop.
2812 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2814 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
2817 2018-01-18 Jakub Jelinek <jakub@redhat.com>
2821 * function.h (gimplify_parameters): Add gimple_seq * argument.
2822 * function.c: Include gimple.h and options.h.
2823 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
2824 for the added local temporaries if needed.
2825 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
2826 if there are any parameter cleanups, wrap whole body into a
2827 try/finally with the cleanups.
2829 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
2832 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
2833 Use GET_MODE_CLASS for scalar floating point.
2835 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
2839 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
2840 Fix call of call_cgraph_insertion_hooks.
2842 2018-01-18 Martin Sebor <msebor@redhat.com>
2844 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
2846 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
2849 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
2852 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
2855 * common.opt: (-ffile-prefix-map): New option.
2856 * opts.c (common_handle_option): Defer it.
2857 * opts-global.c (handle_common_deferred_options): Handle it.
2858 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
2859 * file-prefix-map.h: New file.
2860 (remap_debug_filename, add_debug_prefix_map): ...here.
2861 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
2862 * final.c (debug_prefix_map, add_debug_prefix_map
2863 remap_debug_filename): Move to...
2864 * file-prefix-map.c: New file.
2865 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
2866 generalize, get rid of alloca(), use strrchr() instead of strchr().
2867 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
2868 Implement in terms of add_prefix_map().
2869 (remap_macro_filename, remap_debug_filename): Implement in term of
2871 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
2872 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
2873 * dbxout.c: Include file-prefix-map.h.
2874 * varasm.c: Likewise.
2875 * vmsdbgout.c: Likewise.
2876 * xcoffout.c: Likewise.
2877 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
2878 * doc/cppopts.texi (-fmacro-prefix-map): Document.
2879 * doc/invoke.texi (-ffile-prefix-map): Document.
2880 (-fdebug-prefix-map): Update description.
2882 2018-01-18 Martin Liska <mliska@suse.cz>
2884 * config/i386/i386.c (indirect_thunk_name): Document that also
2886 (output_indirect_thunk): Document why both instructions
2887 (pause and lfence) are generated.
2889 2018-01-18 Richard Biener <rguenther@suse.de>
2891 PR tree-optimization/83887
2892 * graphite-scop-detection.c
2893 (scop_detection::get_nearest_dom_with_single_entry): Remove.
2894 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
2895 (scop_detection::merge_sese): Re-implement with a flood-fill
2896 algorithm that properly finds a SESE region if it exists.
2898 2018-01-18 Jakub Jelinek <jakub@redhat.com>
2901 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
2902 pointer_diff optimizations use view_convert instead of convert.
2904 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2906 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2907 Generate different code for -mno-speculate-indirect-jumps.
2908 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2909 (*call_indirect_aix<mode>): Disable for
2910 -mno-speculate-indirect-jumps.
2911 (*call_indirect_aix<mode>_nospec): New define_insn.
2912 (*call_value_indirect_aix<mode>): Disable for
2913 -mno-speculate-indirect-jumps.
2914 (*call_value_indirect_aix<mode>_nospec): New define_insn.
2915 (*sibcall_nonlocal_sysv<mode>): Generate different code for
2916 -mno-speculate-indirect-jumps.
2917 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2919 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2921 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
2922 long double type, set the flags for noting the default long double
2923 type, even if we don't pass or return a long double type.
2925 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
2928 * ipa-inline.c (flatten_function): Do not overwrite final inlining
2931 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
2933 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
2934 support for merge[hl].
2935 (fold_mergehl_helper): New helper function.
2936 (tree-vector-builder.h): New #include for tree_vector_builder usage.
2937 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
2938 (altivec_vmrglw_direct): Add xxmrglw insn.
2940 2018-01-17 Andrew Waterman <andrew@sifive.com>
2942 * config/riscv/riscv.c (riscv_conditional_register_usage): If
2943 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
2945 2018-01-17 David Malcolm <dmalcolm@redhat.com>
2948 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
2949 call the lto_location_cache before reading the
2950 DECL_SOURCE_LOCATION of the types.
2952 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2953 Richard Sandiford <richard.sandiford@linaro.org>
2955 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
2956 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
2957 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
2958 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
2959 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
2961 * config/aarch64/constraints.md (aarch64_movti_operand):
2963 * config/aarch64/predicates.md (Uti): Add new constraint.
2965 2018-01-17 Carl Love <cel@us.ibm.com>
2967 * config/rs6000/vsx.md (define_expand xl_len_r,
2968 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
2969 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
2971 (define_expand, define_insn): Move the shift left from the
2972 define_insn to the define_expand for lxvl and stxvl instructions.
2973 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
2974 and XL_LEN_R definitions to PURE.
2976 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
2978 * config/i386/i386.c (indirect_thunk_name): Declare regno
2979 as unsigned int. Compare regno with INVALID_REGNUM.
2980 (output_indirect_thunk): Ditto.
2981 (output_indirect_thunk_function): Ditto.
2982 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
2983 in the call to output_indirect_thunk_function.
2985 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
2988 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
2989 rather than the size of inner_type to determine the stack slot size
2990 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
2992 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
2995 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
2998 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3000 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3001 endian Linux systems to optionally enable multilibs for selecting
3002 the long double type if the user configured an explicit type.
3003 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3004 have no long double multilibs if not defined.
3005 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3006 warn if the user used -mabi={ieee,ibm}longdouble and we built
3007 multilibs for long double.
3008 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3009 appropriate multilib option.
3010 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3012 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3013 for building long double multilibs.
3014 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3016 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3018 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3021 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3023 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3026 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3029 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3032 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3034 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3035 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3037 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3039 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3040 different rtl trees depending on TARGET_64BIT.
3041 (rs6000_gen_lvx): Likewise.
3043 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3045 * config/visium/visium.md (nop): Tweak comment.
3046 (hazard_nop): Likewise.
3048 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3050 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3051 -mspeculate-indirect-jumps.
3052 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3053 for -mno-speculate-indirect-jumps.
3054 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3055 (*call_value_indirect_elfv2<mode>): Disable for
3056 -mno-speculate-indirect-jumps.
3057 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3058 (indirect_jump): Emit different RTL for
3059 -mno-speculate-indirect-jumps.
3060 (*indirect_jump<mode>): Disable for
3061 -mno-speculate-indirect-jumps.
3062 (*indirect_jump<mode>_nospec): New define_insn.
3063 (tablejump): Emit different RTL for
3064 -mno-speculate-indirect-jumps.
3065 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3066 (tablejumpsi_nospec): New define_expand.
3067 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3068 (tablejumpdi_nospec): New define_expand.
3069 (*tablejump<mode>_internal1): Disable for
3070 -mno-speculate-indirect-jumps.
3071 (*tablejump<mode>_internal1_nospec): New define_insn.
3072 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3075 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3077 * caller-save.c (insert_save): Drop unnecessary parameter. All
3080 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3081 Richard Biener <rguenth@suse.de>
3084 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3085 return early, inline manually is_gimple_sizepos. Make sure if we
3086 call gimplify_expr we don't end up with a gimple constant.
3087 * tree.c (variably_modified_type_p): Don't return true for
3088 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3089 * gimplify.h (is_gimple_sizepos): Remove.
3091 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3093 PR tree-optimization/83857
3094 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3095 vectorizable_live_operation for pure SLP statements.
3096 (vectorizable_live_operation): Handle PHIs.
3098 2018-01-16 Richard Biener <rguenther@suse.de>
3100 PR tree-optimization/83867
3101 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3102 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3104 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3107 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3108 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3109 If off is not INTEGER_CST, issue a may not be aligned warning
3110 rather than isn't aligned. Use isn%'t rather than isn't.
3111 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3113 <case MULT_EXPR>: Improve the case when bottom and one of the
3114 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3115 operand, in that case check if the other operand is multiple of
3116 bottom divided by the INTEGER_CST operand.
3118 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3121 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3122 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3123 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3124 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3125 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3126 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3127 (pa_function_arg_size): New function.
3129 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3131 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3132 in a separate statement.
3134 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3136 PR tree-optimization/83847
3137 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3138 group gathers and scatters.
3140 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3142 PR rtl-optimization/86620
3143 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3145 PR rtl-optimization/83213
3146 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3147 to last if both are JUMP_INSNs.
3149 PR tree-optimization/83843
3150 * gimple-ssa-store-merging.c
3151 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3152 store_immediate_info for bswap/nop orig_stores.
3154 2018-01-15 Andrew Waterman <andrew@sifive.com>
3156 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3158 <UDIV>: Increase cost if !TARGET_DIV.
3160 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3162 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3163 (define_attr "cr_logical_3op"): New.
3164 (cceq_ior_compare): Adjust.
3165 (cceq_ior_compare_complement): Adjust.
3166 (*cceq_rev_compare): Adjust.
3167 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3168 (is_cracked_insn): Adjust.
3169 (insn_must_be_first_in_group): Adjust.
3170 * config/rs6000/40x.md: Adjust.
3171 * config/rs6000/440.md: Adjust.
3172 * config/rs6000/476.md: Adjust.
3173 * config/rs6000/601.md: Adjust.
3174 * config/rs6000/603.md: Adjust.
3175 * config/rs6000/6xx.md: Adjust.
3176 * config/rs6000/7450.md: Adjust.
3177 * config/rs6000/7xx.md: Adjust.
3178 * config/rs6000/8540.md: Adjust.
3179 * config/rs6000/cell.md: Adjust.
3180 * config/rs6000/e300c2c3.md: Adjust.
3181 * config/rs6000/e500mc.md: Adjust.
3182 * config/rs6000/e500mc64.md: Adjust.
3183 * config/rs6000/e5500.md: Adjust.
3184 * config/rs6000/e6500.md: Adjust.
3185 * config/rs6000/mpc.md: Adjust.
3186 * config/rs6000/power4.md: Adjust.
3187 * config/rs6000/power5.md: Adjust.
3188 * config/rs6000/power6.md: Adjust.
3189 * config/rs6000/power7.md: Adjust.
3190 * config/rs6000/power8.md: Adjust.
3191 * config/rs6000/power9.md: Adjust.
3192 * config/rs6000/rs64.md: Adjust.
3193 * config/rs6000/titan.md: Adjust.
3195 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3197 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3198 ix86_indirect_branch_register logic.
3200 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3202 * config/i386/constraints.md (Bs): Update
3203 ix86_indirect_branch_register check. Don't check
3204 ix86_indirect_branch_register with GOT_memory_operand.
3206 * config/i386/predicates.md (GOT_memory_operand): Don't check
3207 ix86_indirect_branch_register here.
3208 (GOT32_symbol_operand): Likewise.
3210 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3212 * config/i386/predicates.md (constant_call_address_operand):
3213 Rewrite ix86_indirect_branch_register logic.
3214 (sibcall_insn_operand): Likewise.
3216 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3218 * config/i386/constraints.md (Bs): Replace
3219 ix86_indirect_branch_thunk_register with
3220 ix86_indirect_branch_register.
3222 * config/i386/i386.md (indirect_jump): Likewise.
3223 (tablejump): Likewise.
3224 (*sibcall_memory): Likewise.
3225 (*sibcall_value_memory): Likewise.
3226 Peepholes of indirect call and jump via memory: Likewise.
3227 * config/i386/i386.opt: Likewise.
3228 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3229 (GOT_memory_operand): Likewise.
3230 (call_insn_operand): Likewise.
3231 (sibcall_insn_operand): Likewise.
3232 (GOT32_symbol_operand): Likewise.
3234 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3237 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3238 type rather than type addr's type points to.
3239 (expand_omp_atomic_mutex): Likewise.
3240 (expand_omp_atomic): Likewise.
3242 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3245 * config/i386/i386.c (output_indirect_thunk_function): Use
3246 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3247 for __x86_return_thunk.
3249 2018-01-15 Richard Biener <rguenther@suse.de>
3252 * expmed.c (extract_bit_field_1): Fix typo.
3254 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3257 * config/arm/iterators.md (VF): New mode iterator.
3258 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3259 Remove integer-related logic from pattern.
3260 (neon_vabd<mode>_3): Likewise.
3262 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3265 * common.opt (fstrict-overflow): No longer an alias.
3266 (fwrapv-pointer): New option.
3267 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3268 also for pointer types based on flag_wrapv_pointer.
3269 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3270 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3271 opts->x_flag_wrapv got set.
3272 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3273 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3274 POINTER_TYPE_OVERFLOW_UNDEFINED.
3275 * match.pd: Likewise in address comparison pattern.
3276 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3278 2018-01-15 Richard Biener <rguenther@suse.de>
3281 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3282 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3283 Reset type names to their identifier if their TYPE_DECL doesn't
3284 have linkage (and thus is used for ODR and devirt).
3285 (save_debug_info_for_decl): Remove.
3286 (save_debug_info_for_type): Likewise.
3287 (add_tree_to_fld_list): Adjust.
3288 * tree-pretty-print.c (dump_generic_node): Make dumping of
3289 type names more robust.
3291 2018-01-15 Richard Biener <rguenther@suse.de>
3293 * BASE-VER: Bump to 8.0.1.
3295 2018-01-14 Martin Sebor <msebor@redhat.com>
3298 * builtins.c (check_access): Avoid warning when the no-warning bit
3301 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3303 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3304 * ira-color (allocno_hard_regs_compare): Likewise.
3306 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3309 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3310 Use .pushsection/.popsection.
3312 2018-01-14 Martin Sebor <msebor@redhat.com>
3315 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3317 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3319 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3320 entry from extra_headers.
3321 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3322 extra_headers, make the list bitwise identical to the i?86-*-* one.
3324 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3326 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3327 -mcmodel=large with -mindirect-branch=thunk,
3328 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3329 -mfunction-return=thunk-extern.
3330 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3331 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3332 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3334 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3336 * config/i386/i386.c (print_reg): Print the name of the full
3337 integer register without '%'.
3338 (ix86_print_operand): Handle 'V'.
3339 * doc/extend.texi: Document 'V' modifier.
3341 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3343 * config/i386/constraints.md (Bs): Disallow memory operand for
3344 -mindirect-branch-register.
3346 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3347 (GOT_memory_operand): Likewise.
3348 (call_insn_operand): Likewise.
3349 (sibcall_insn_operand): Likewise.
3350 (GOT32_symbol_operand): Likewise.
3351 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3352 for -mindirect-branch-register.
3353 (tablejump): Likewise.
3354 (*sibcall_memory): Likewise.
3355 (*sibcall_value_memory): Likewise.
3356 Disallow peepholes of indirect call and jump via memory for
3357 -mindirect-branch-register.
3358 (*call_pop): Replace m with Bw.
3359 (*call_value_pop): Likewise.
3360 (*sibcall_pop_memory): Replace m with Bs.
3361 * config/i386/i386.opt (mindirect-branch-register): New option.
3362 * doc/invoke.texi: Document -mindirect-branch-register option.
3364 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3366 * config/i386/i386-protos.h (ix86_output_function_return): New.
3367 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3368 set function_return_type.
3369 (indirect_thunk_name): Add ret_p to indicate thunk for function
3371 (output_indirect_thunk_function): Pass false to
3372 indirect_thunk_name.
3373 (ix86_output_indirect_branch_via_reg): Likewise.
3374 (ix86_output_indirect_branch_via_push): Likewise.
3375 (output_indirect_thunk_function): Create alias for function
3376 return thunk if regno < 0.
3377 (ix86_output_function_return): New function.
3378 (ix86_handle_fndecl_attribute): Handle function_return.
3379 (ix86_attribute_table): Add function_return.
3380 * config/i386/i386.h (machine_function): Add
3381 function_return_type.
3382 * config/i386/i386.md (simple_return_internal): Use
3383 ix86_output_function_return.
3384 (simple_return_internal_long): Likewise.
3385 * config/i386/i386.opt (mfunction-return=): New option.
3386 (indirect_branch): Mention -mfunction-return=.
3387 * doc/extend.texi: Document function_return function attribute.
3388 * doc/invoke.texi: Document -mfunction-return= option.
3390 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3392 * config/i386/i386-opts.h (indirect_branch): New.
3393 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3394 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3395 with local indirect jump when converting indirect call and jump.
3396 (ix86_set_indirect_branch_type): New.
3397 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3398 (indirectlabelno): New.
3399 (indirect_thunk_needed): Likewise.
3400 (indirect_thunk_bnd_needed): Likewise.
3401 (indirect_thunks_used): Likewise.
3402 (indirect_thunks_bnd_used): Likewise.
3403 (INDIRECT_LABEL): Likewise.
3404 (indirect_thunk_name): Likewise.
3405 (output_indirect_thunk): Likewise.
3406 (output_indirect_thunk_function): Likewise.
3407 (ix86_output_indirect_branch_via_reg): Likewise.
3408 (ix86_output_indirect_branch_via_push): Likewise.
3409 (ix86_output_indirect_branch): Likewise.
3410 (ix86_output_indirect_jmp): Likewise.
3411 (ix86_code_end): Call output_indirect_thunk_function if needed.
3412 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3414 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3415 (ix86_attribute_table): Add indirect_branch.
3416 * config/i386/i386.h (machine_function): Add indirect_branch_type
3417 and has_local_indirect_jump.
3418 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3420 (tablejump): Likewise.
3421 (*indirect_jump): Use ix86_output_indirect_jmp.
3422 (*tablejump_1): Likewise.
3423 (simple_return_indirect_internal): Likewise.
3424 * config/i386/i386.opt (mindirect-branch=): New option.
3425 (indirect_branch): New.
3428 (thunk-inline): Likewise.
3429 (thunk-extern): Likewise.
3430 * doc/extend.texi: Document indirect_branch function attribute.
3431 * doc/invoke.texi: Document -mindirect-branch= option.
3433 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3436 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3438 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3440 * ipa-inline.c (want_inline_small_function_p): Return false if
3441 inlining has already failed with CIF_FINAL_ERROR.
3442 (update_caller_keys): Call want_inline_small_function_p before
3444 (update_callee_keys): Likewise.
3446 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3448 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3450 (rs6000_quadword_masked_address_p): Likewise.
3451 (quad_aligned_load_p): Likewise.
3452 (quad_aligned_store_p): Likewise.
3453 (const_load_sequence_p): Add comment to describe the outer-most loop.
3454 (mimic_memory_attributes_and_flags): New function.
3455 (rs6000_gen_stvx): Likewise.
3456 (replace_swapped_aligned_store): Likewise.
3457 (rs6000_gen_lvx): Likewise.
3458 (replace_swapped_aligned_load): Likewise.
3459 (replace_swapped_load_constant): Capitalize argument name in
3460 comment describing this function.
3461 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3462 and stores that access quad-word aligned addresses and replace
3463 with stvx or lvx instructions when appropriate.
3464 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3465 New function prototype.
3466 (rs6000_quadword_masked_address_p): Likewise.
3467 (rs6000_gen_lvx): Likewise.
3468 (rs6000_gen_stvx): Likewise.
3469 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3470 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3471 when memory address is aligned.
3472 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3473 this split to select lvx instruction when memory address is aligned.
3474 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3475 instruction when memory address is aligned.
3476 (*vsx_le_perm_load_v16qi): Likewise.
3477 (four unnamed splitters): Modify to select the stvx instruction
3478 when memory is aligned.
3480 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3482 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3483 which appears in the queue multiple times.
3485 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3486 Alan Hayward <alan.hayward@arm.com>
3487 David Sherwood <david.sherwood@arm.com>
3489 * tree-vectorizer.h (vec_lower_bound): New structure.
3490 (_loop_vec_info): Add check_nonzero and lower_bounds.
3491 (LOOP_VINFO_CHECK_NONZERO): New macro.
3492 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
3493 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
3494 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
3495 fields. Make seg_len the distance travelled, not including the
3497 (dr_direction_indicator): Declare.
3498 (dr_zero_step_indicator): Likewise.
3499 (dr_known_forward_stride_p): Likewise.
3500 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
3502 (runtime_alias_check_p): Allow runtime alias checks with
3504 (operator ==): Compare access_size and align.
3505 (prune_runtime_alias_test_list): Rework for new distinction between
3506 the access_size and seg_len.
3507 (create_intersect_range_checks_index): Likewise. Cope with polynomial
3509 (get_segment_min_max): New function.
3510 (create_intersect_range_checks): Use it.
3511 (dr_step_indicator): New function.
3512 (dr_direction_indicator): Likewise.
3513 (dr_zero_step_indicator): Likewise.
3514 (dr_known_forward_stride_p): Likewise.
3515 * tree-loop-distribution.c (data_ref_segment_size): Return
3516 DR_STEP * (niters - 1).
3517 (compute_alias_check_pairs): Update call to the dr_with_seg_len
3519 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
3520 (vect_preserves_scalar_order_p): New function, split out from...
3521 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
3522 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
3523 (vect_vfa_access_size): New function.
3524 (vect_vfa_align): Likewise.
3525 (vect_compile_time_alias): Take access_size_a and access_b arguments.
3526 (dump_lower_bound): New function.
3527 (vect_check_lower_bound): Likewise.
3528 (vect_small_gap_p): Likewise.
3529 (vectorizable_with_step_bound_p): Likewise.
3530 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
3531 depencies if the vectorization factor is 1. Convert the checks
3532 for nonzero steps into checks on the bounds of DR_STEP. Try using
3533 a bunds check for variable steps if the minimum required step is
3534 relatively small. Update calls to the dr_with_seg_len
3535 constructor and to vect_compile_time_alias.
3536 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
3538 (vect_loop_versioning): Call it.
3539 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
3541 (vect_estimate_min_profitable_iters): Account for any bounds checks.
3543 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3544 Alan Hayward <alan.hayward@arm.com>
3545 David Sherwood <david.sherwood@arm.com>
3547 * doc/sourcebuild.texi (vect_scatter_store): Document.
3548 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
3550 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
3552 * genopinit.c (main): Add supports_vec_scatter_store and
3553 supports_vec_scatter_store_cached to target_optabs.
3554 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
3555 IFN_MASK_SCATTER_STORE.
3556 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
3558 * internal-fn.h (internal_store_fn_p): Declare.
3559 (internal_fn_stored_value_index): Likewise.
3560 * internal-fn.c (scatter_store_direct): New macro.
3561 (expand_scatter_store_optab_fn): New function.
3562 (direct_scatter_store_optab_supported_p): New macro.
3563 (internal_store_fn_p): New function.
3564 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
3565 IFN_MASK_SCATTER_STORE.
3566 (internal_fn_mask_index): Likewise.
3567 (internal_fn_stored_value_index): New function.
3568 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
3570 * optabs-query.h (supports_vec_scatter_store_p): Declare.
3571 * optabs-query.c (supports_vec_scatter_store_p): New function.
3572 * tree-vectorizer.h (vect_get_store_rhs): Declare.
3573 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
3574 true for scatter stores.
3575 (vect_gather_scatter_fn_p): Handle scatter stores too.
3576 (vect_check_gather_scatter): Consider using scatter stores if
3577 supports_vec_scatter_store_p.
3578 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
3580 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3581 internal_fn_stored_value_index.
3582 (check_load_store_masking): Handle scatter stores too.
3583 (vect_get_store_rhs): Make public.
3584 (vectorizable_call): Use internal_store_fn_p.
3585 (vectorizable_store): Handle scatter store internal functions.
3586 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
3587 when deciding whether the end of the group has been reached.
3588 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
3589 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
3590 (mask_scatter_store<mode>): New insns.
3592 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3593 Alan Hayward <alan.hayward@arm.com>
3594 David Sherwood <david.sherwood@arm.com>
3596 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
3597 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
3598 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
3600 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
3601 Use vect_truncate_gather_scatter_offset if we can't treat the
3602 operation as a normal gather load or scatter store.
3603 (get_group_load_store_type): Take the gather_scatter_info
3604 as argument. Try using a gather load or scatter store for
3605 single-element groups.
3606 (get_load_store_type): Update calls to get_group_load_store_type
3607 and vect_use_strided_gather_scatters_p.
3609 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3610 Alan Hayward <alan.hayward@arm.com>
3611 David Sherwood <david.sherwood@arm.com>
3613 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
3614 optional tree argument.
3615 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
3617 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
3618 but continue to use the current value as a fallback.
3619 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
3620 to compare the updates.
3621 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
3622 (get_load_store_type): Use it when handling a strided access.
3623 (vect_get_strided_load_store_ops): New function.
3624 (vect_get_data_ptr_increment): Likewise.
3625 (vectorizable_load): Handle strided gather loads. Always pass
3626 a step to vect_create_data_ref_ptr and bump_vector_ptr.
3628 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3629 Alan Hayward <alan.hayward@arm.com>
3630 David Sherwood <david.sherwood@arm.com>
3632 * doc/md.texi (gather_load@var{m}): Document.
3633 (mask_gather_load@var{m}): Likewise.
3634 * genopinit.c (main): Add supports_vec_gather_load and
3635 supports_vec_gather_load_cached to target_optabs.
3636 * optabs-tree.c (init_tree_optimization_optabs): Use
3637 ggc_cleared_alloc to allocate target_optabs.
3638 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
3639 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
3641 * internal-fn.h (internal_load_fn_p): Declare.
3642 (internal_gather_scatter_fn_p): Likewise.
3643 (internal_fn_mask_index): Likewise.
3644 (internal_gather_scatter_fn_supported_p): Likewise.
3645 * internal-fn.c (gather_load_direct): New macro.
3646 (expand_gather_load_optab_fn): New function.
3647 (direct_gather_load_optab_supported_p): New macro.
3648 (direct_internal_fn_optab): New function.
3649 (internal_load_fn_p): Likewise.
3650 (internal_gather_scatter_fn_p): Likewise.
3651 (internal_fn_mask_index): Likewise.
3652 (internal_gather_scatter_fn_supported_p): Likewise.
3653 * optabs-query.c (supports_at_least_one_mode_p): New function.
3654 (supports_vec_gather_load_p): Likewise.
3655 * optabs-query.h (supports_vec_gather_load_p): Declare.
3656 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
3657 and memory_type field.
3658 (NUM_PATTERNS): Bump to 15.
3659 * tree-vect-data-refs.c: Include internal-fn.h.
3660 (vect_gather_scatter_fn_p): New function.
3661 (vect_describe_gather_scatter_call): Likewise.
3662 (vect_check_gather_scatter): Try using internal functions for
3663 gather loads. Recognize existing calls to a gather load function.
3664 (vect_analyze_data_refs): Consider using gather loads if
3665 supports_vec_gather_load_p.
3666 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
3667 (vect_get_gather_scatter_offset_type): Likewise.
3668 (vect_convert_mask_for_vectype): Likewise.
3669 (vect_add_conversion_to_patterm): Likewise.
3670 (vect_try_gather_scatter_pattern): Likewise.
3671 (vect_recog_gather_scatter_pattern): New pattern recognizer.
3672 (vect_vect_recog_func_ptrs): Add it.
3673 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3674 internal_fn_mask_index and internal_gather_scatter_fn_p.
3675 (check_load_store_masking): Take the gather_scatter_info as an
3676 argument and handle gather loads.
3677 (vect_get_gather_scatter_ops): New function.
3678 (vectorizable_call): Check internal_load_fn_p.
3679 (vectorizable_load): Likewise. Handle gather load internal
3681 (vectorizable_store): Update call to check_load_store_masking.
3682 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
3683 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
3684 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
3685 (aarch64_gather_scale_operand_d): New predicates.
3686 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
3687 (mask_gather_load<mode>): New insns.
3689 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3690 Alan Hayward <alan.hayward@arm.com>
3691 David Sherwood <david.sherwood@arm.com>
3693 * optabs.def (fold_left_plus_optab): New optab.
3694 * doc/md.texi (fold_left_plus_@var{m}): Document.
3695 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
3696 * internal-fn.c (fold_left_direct): Define.
3697 (expand_fold_left_optab_fn): Likewise.
3698 (direct_fold_left_optab_supported_p): Likewise.
3699 * fold-const-call.c (fold_const_fold_left): New function.
3700 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
3701 * tree-parloops.c (valid_reduction_p): New function.
3702 (gather_scalar_reductions): Use it.
3703 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
3704 (vect_finish_replace_stmt): Declare.
3705 * tree-vect-loop.c (fold_left_reduction_fn): New function.
3706 (needs_fold_left_reduction_p): New function, split out from...
3707 (vect_is_simple_reduction): ...here. Accept reductions that
3708 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
3709 (vect_force_simple_reduction): Also store the reduction type in
3710 the assignment's STMT_VINFO_REDUC_TYPE.
3711 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
3712 (merge_with_identity): New function.
3713 (vect_expand_fold_left): Likewise.
3714 (vectorize_fold_left_reduction): Likewise.
3715 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
3716 scalar phi in place for it. Check for target support and reject
3717 cases that would reassociate the operation. Defer the transform
3718 phase to vectorize_fold_left_reduction.
3719 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
3720 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
3721 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
3723 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3725 * tree-if-conv.c (predicate_mem_writes): Remove redundant
3726 call to ifc_temp_var.
3728 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3729 Alan Hayward <alan.hayward@arm.com>
3730 David Sherwood <david.sherwood@arm.com>
3732 * target.def (legitimize_address_displacement): Take the original
3733 offset as a poly_int.
3734 * targhooks.h (default_legitimize_address_displacement): Update
3736 * targhooks.c (default_legitimize_address_displacement): Likewise.
3737 * doc/tm.texi: Regenerate.
3738 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
3739 as an argument, moving assert of ad->disp == ad->disp_term to...
3740 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
3741 Try calling targetm.legitimize_address_displacement before expanding
3742 the address rather than afterwards, and adjust for the new interface.
3743 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
3744 Match the new hook interface. Handle SVE addresses.
3745 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
3748 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3750 * Makefile.in (OBJS): Add early-remat.o.
3751 * target.def (select_early_remat_modes): New hook.
3752 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
3753 * doc/tm.texi: Regenerate.
3754 * targhooks.h (default_select_early_remat_modes): Declare.
3755 * targhooks.c (default_select_early_remat_modes): New function.
3756 * timevar.def (TV_EARLY_REMAT): New timevar.
3757 * passes.def (pass_early_remat): New pass.
3758 * tree-pass.h (make_pass_early_remat): Declare.
3759 * early-remat.c: New file.
3760 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
3762 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
3764 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3765 Alan Hayward <alan.hayward@arm.com>
3766 David Sherwood <david.sherwood@arm.com>
3768 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
3769 vfm1 with a bound_epilog parameter.
3770 (vect_do_peeling): Update calls accordingly, and move the prologue
3771 call earlier in the function. Treat the base bound_epilog as 0 for
3772 fully-masked loops and retain vf - 1 for other loops. Add 1 to
3773 this base when peeling for gaps.
3774 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
3775 with fully-masked loops.
3776 (vect_estimate_min_profitable_iters): Handle the single peeled
3777 iteration in that case.
3779 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3780 Alan Hayward <alan.hayward@arm.com>
3781 David Sherwood <david.sherwood@arm.com>
3783 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
3784 single-element interleaving even if the size is not a power of 2.
3785 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
3786 accesses for single-element interleaving if the group size is
3789 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3790 Alan Hayward <alan.hayward@arm.com>
3791 David Sherwood <david.sherwood@arm.com>
3793 * doc/md.texi (fold_extract_last_@var{m}): Document.
3794 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
3795 * optabs.def (fold_extract_last_optab): New optab.
3796 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
3797 * internal-fn.c (fold_extract_direct): New macro.
3798 (expand_fold_extract_optab_fn): Likewise.
3799 (direct_fold_extract_optab_supported_p): Likewise.
3800 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
3801 * tree-vect-loop.c (vect_model_reduction_cost): Handle
3802 EXTRACT_LAST_REDUCTION.
3803 (get_initial_def_for_reduction): Do not create an initial vector
3804 for EXTRACT_LAST_REDUCTION reductions.
3805 (vectorizable_reduction): Leave the scalar phi in place for
3806 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
3807 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
3808 epilogue code for EXTRACT_LAST_REDUCTION and defer the
3809 transform phase to vectorizable_condition.
3810 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
3812 (vect_finish_stmt_generation): ...here.
3813 (vect_finish_replace_stmt): New function.
3814 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
3815 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
3817 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
3819 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3820 Alan Hayward <alan.hayward@arm.com>
3821 David Sherwood <david.sherwood@arm.com>
3823 * doc/md.texi (extract_last_@var{m}): Document.
3824 * optabs.def (extract_last_optab): New optab.
3825 * internal-fn.def (EXTRACT_LAST): New internal function.
3826 * internal-fn.c (cond_unary_direct): New macro.
3827 (expand_cond_unary_optab_fn): Likewise.
3828 (direct_cond_unary_optab_supported_p): Likewise.
3829 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
3830 loops using EXTRACT_LAST.
3831 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
3832 (extract_last_<mode>): ...this optab.
3833 (vec_extract<mode><Vel>): Update accordingly.
3835 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3836 Alan Hayward <alan.hayward@arm.com>
3837 David Sherwood <david.sherwood@arm.com>
3839 * target.def (empty_mask_is_expensive): New hook.
3840 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
3841 * doc/tm.texi: Regenerate.
3842 * targhooks.h (default_empty_mask_is_expensive): Declare.
3843 * targhooks.c (default_empty_mask_is_expensive): New function.
3844 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
3845 if the target says that empty masks are expensive.
3846 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
3848 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
3850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3851 Alan Hayward <alan.hayward@arm.com>
3852 David Sherwood <david.sherwood@arm.com>
3854 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
3855 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
3856 (vect_use_loop_mask_for_alignment_p): New function.
3857 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
3858 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
3859 niters_skip argument. Make sure that the first niters_skip elements
3860 of the first iteration are inactive.
3861 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
3862 Update call to vect_set_loop_masks_directly.
3863 (get_misalign_in_elems): New function, split out from...
3864 (vect_gen_prolog_loop_niters): ...here.
3865 (vect_update_init_of_dr): Take a code argument that specifies whether
3866 the adjustment should be added or subtracted.
3867 (vect_update_init_of_drs): Likewise.
3868 (vect_prepare_for_masked_peels): New function.
3869 (vect_do_peeling): Skip prologue peeling if we're using a mask
3870 instead. Update call to vect_update_inits_of_drs.
3871 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3873 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
3874 alignment. Do not include the number of peeled iterations in
3875 the minimum threshold in that case.
3876 (vectorizable_induction): Adjust the start value down by
3877 LOOP_VINFO_MASK_SKIP_NITERS iterations.
3878 (vect_transform_loop): Call vect_prepare_for_masked_peels.
3879 Take the number of skipped iterations into account when calculating
3881 * tree-vect-stmts.c (vect_gen_while_not): New function.
3883 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3884 Alan Hayward <alan.hayward@arm.com>
3885 David Sherwood <david.sherwood@arm.com>
3887 * doc/sourcebuild.texi (vect_fully_masked): Document.
3888 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
3890 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
3892 (vect_analyze_loop_2): ...here. Don't check the vectorization
3893 factor against the number of loop iterations if the loop is
3896 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3897 Alan Hayward <alan.hayward@arm.com>
3898 David Sherwood <david.sherwood@arm.com>
3900 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
3901 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
3902 (dump_groups): Update accordingly.
3903 (iv_use::mem_type): New member variable.
3904 (address_p): New function.
3905 (record_use): Add a mem_type argument and initialize the new
3907 (record_group_use): Add a mem_type argument. Use address_p.
3908 Remove obsolete null checks of base_object. Update call to record_use.
3909 (find_interesting_uses_op): Update call to record_group_use.
3910 (find_interesting_uses_cond): Likewise.
3911 (find_interesting_uses_address): Likewise.
3912 (get_mem_type_for_internal_fn): New function.
3913 (find_address_like_use): Likewise.
3914 (find_interesting_uses_stmt): Try find_address_like_use before
3915 calling find_interesting_uses_op.
3916 (addr_offset_valid_p): Use the iv mem_type field as the type
3917 of the addressed memory.
3918 (add_autoinc_candidates): Likewise.
3919 (get_address_cost): Likewise.
3920 (split_small_address_groups_p): Use address_p.
3921 (split_address_groups): Likewise.
3922 (add_iv_candidate_for_use): Likewise.
3923 (autoinc_possible_for_pair): Likewise.
3924 (rewrite_groups): Likewise.
3925 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
3926 (determine_group_iv_cost): Update after split of USE_ADDRESS.
3927 (get_alias_ptr_type_for_ptr_address): New function.
3928 (rewrite_use_address): Rewrite address uses in calls that were
3929 identified by find_address_like_use.
3931 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3932 Alan Hayward <alan.hayward@arm.com>
3933 David Sherwood <david.sherwood@arm.com>
3935 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
3937 * gimple-expr.h (is_gimple_addressable: Likewise.
3938 * gimple-expr.c (is_gimple_address): Likewise.
3939 * internal-fn.c (expand_call_mem_ref): New function.
3940 (expand_mask_load_optab_fn): Use it.
3941 (expand_mask_store_optab_fn): Likewise.
3943 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3944 Alan Hayward <alan.hayward@arm.com>
3945 David Sherwood <david.sherwood@arm.com>
3947 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
3948 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
3949 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
3950 (cond_umax@var{mode}): Document.
3951 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
3952 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
3953 (cond_umin_optab, cond_umax_optab): New optabs.
3954 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
3955 (COND_IOR, COND_XOR): New internal functions.
3956 * internal-fn.h (get_conditional_internal_fn): Declare.
3957 * internal-fn.c (cond_binary_direct): New macro.
3958 (expand_cond_binary_optab_fn): Likewise.
3959 (direct_cond_binary_optab_supported_p): Likewise.
3960 (get_conditional_internal_fn): New function.
3961 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
3962 Cope with reduction statements that are vectorized as calls rather
3964 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
3965 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
3966 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
3967 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3968 (UNSPEC_COND_EOR): New unspecs.
3969 (optab): Add mappings for them.
3970 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
3971 (sve_int_op, sve_fp_op): New int attributes.
3973 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3974 Alan Hayward <alan.hayward@arm.com>
3975 David Sherwood <david.sherwood@arm.com>
3977 * optabs.def (while_ult_optab): New optab.
3978 * doc/md.texi (while_ult@var{m}@var{n}): Document.
3979 * internal-fn.def (WHILE_ULT): New internal function.
3980 * internal-fn.h (direct_internal_fn_supported_p): New override
3981 that takes two types as argument.
3982 * internal-fn.c (while_direct): New macro.
3983 (expand_while_optab_fn): New function.
3984 (convert_optab_supported_p): Likewise.
3985 (direct_while_optab_supported_p): New macro.
3986 * wide-int.h (wi::udiv_ceil): New function.
3987 * tree-vectorizer.h (rgroup_masks): New structure.
3988 (vec_loop_masks): New typedef.
3989 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
3991 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
3992 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
3993 (vect_max_vf): New function.
3994 (slpeel_make_loop_iterate_ntimes): Delete.
3995 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
3996 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
3997 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
3998 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
3999 internal-fn.h, stor-layout.h and optabs-query.h.
4000 (vect_set_loop_mask): New function.
4001 (add_preheader_seq): Likewise.
4002 (add_header_seq): Likewise.
4003 (interleave_supported_p): Likewise.
4004 (vect_maybe_permute_loop_masks): Likewise.
4005 (vect_set_loop_masks_directly): Likewise.
4006 (vect_set_loop_condition_masked): Likewise.
4007 (vect_set_loop_condition_unmasked): New function, split out from
4008 slpeel_make_loop_iterate_ntimes.
4009 (slpeel_make_loop_iterate_ntimes): Rename to..
4010 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4011 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4012 (vect_do_peeling): Update call accordingly.
4013 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4015 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4016 mask_compare_type, can_fully_mask_p and fully_masked_p.
4017 (release_vec_loop_masks): New function.
4018 (_loop_vec_info): Use it to free the loop masks.
4019 (can_produce_all_loop_masks_p): New function.
4020 (vect_get_max_nscalars_per_iter): Likewise.
4021 (vect_verify_full_masking): Likewise.
4022 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4023 retries, and free the mask rgroups before retrying. Check loop-wide
4024 reasons for disallowing fully-masked loops. Make the final decision
4025 about whether use a fully-masked loop or not.
4026 (vect_estimate_min_profitable_iters): Do not assume that peeling
4027 for the number of iterations will be needed for fully-masked loops.
4028 (vectorizable_reduction): Disable fully-masked loops.
4029 (vectorizable_live_operation): Likewise.
4030 (vect_halve_mask_nunits): New function.
4031 (vect_double_mask_nunits): Likewise.
4032 (vect_record_loop_mask): Likewise.
4033 (vect_get_loop_mask): Likewise.
4034 (vect_transform_loop): Handle the case in which the final loop
4035 iteration might handle a partial vector. Call vect_set_loop_condition
4036 instead of slpeel_make_loop_iterate_ntimes.
4037 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4038 (check_load_store_masking): New function.
4039 (prepare_load_store_mask): Likewise.
4040 (vectorizable_store): Handle fully-masked loops.
4041 (vectorizable_load): Likewise.
4042 (supportable_widening_operation): Use vect_halve_mask_nunits for
4044 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4045 (vect_gen_while): New function.
4046 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4047 (aarch64_uqdec<mode>): New insn.
4049 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4050 Alan Hayward <alan.hayward@arm.com>
4051 David Sherwood <david.sherwood@arm.com>
4053 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4054 (reduc_xor_scal_optab): New optabs.
4055 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4056 (reduc_xor_scal_@var{m}): Document.
4057 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4058 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4060 * fold-const-call.c (fold_const_call): Handle them.
4061 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4062 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4063 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4064 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4065 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4066 (UNSPEC_XORV): New unspecs.
4067 (optab): Add entries for them.
4068 (BITWISEV): New int iterator.
4069 (bit_reduc_op): New int attributes.
4071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4072 Alan Hayward <alan.hayward@arm.com>
4073 David Sherwood <david.sherwood@arm.com>
4075 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4076 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4077 * optabs.def (vec_shl_insert_optab): New optab.
4078 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4079 (duplicate_and_interleave): Likewise.
4080 * tree-vect-loop.c: Include internal-fn.h.
4081 (neutral_op_for_slp_reduction): New function, split out from
4082 get_initial_defs_for_reduction.
4083 (get_initial_def_for_reduction): Handle option 2 for variable-length
4084 vectors by loading the neutral value into a vector and then shifting
4085 the initial value into element 0.
4086 (get_initial_defs_for_reduction): Replace the code argument with
4087 the neutral value calculated by neutral_op_for_slp_reduction.
4088 Use gimple_build_vector for constant-length vectors.
4089 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4090 but the first group_size elements have a neutral value.
4091 Use duplicate_and_interleave otherwise.
4092 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4093 Update call to get_initial_defs_for_reduction. Handle SLP
4094 reductions for variable-length vectors by creating one vector
4095 result for each scalar result, with the elements associated
4096 with other scalar results stubbed out with the neutral value.
4097 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4098 Require IFN_VEC_SHL_INSERT for double reductions on
4099 variable-length vectors, or SLP reductions that have
4100 a neutral value. Require can_duplicate_and_interleave_p
4101 support for variable-length unchained SLP reductions if there
4102 is no neutral value, such as for MIN/MAX reductions. Also require
4103 the number of vector elements to be a multiple of the number of
4104 SLP statements when doing variable-length unchained SLP reductions.
4105 Update call to vect_create_epilog_for_reduction.
4106 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4107 and remove initial values.
4108 (duplicate_and_interleave): Make public.
4109 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4110 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4112 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4113 Alan Hayward <alan.hayward@arm.com>
4114 David Sherwood <david.sherwood@arm.com>
4116 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4117 (can_duplicate_and_interleave_p): New function.
4118 (vect_get_and_check_slp_defs): Take the vector of statements
4119 rather than just the current one. Remove excess parentheses.
4120 Restriction rejectinon of vect_constant_def and vect_external_def
4121 for variable-length vectors to boolean types, or types for which
4122 can_duplicate_and_interleave_p is false.
4123 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4124 (duplicate_and_interleave): New function.
4125 (vect_get_constant_vectors): Use gimple_build_vector for
4126 constant-length vectors and suitable variable-length constant
4127 vectors. Use duplicate_and_interleave for other variable-length
4128 vectors. Don't defer the update when inserting new statements.
4130 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4131 Alan Hayward <alan.hayward@arm.com>
4132 David Sherwood <david.sherwood@arm.com>
4134 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4135 min_profitable_iters doesn't go negative.
4137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4138 Alan Hayward <alan.hayward@arm.com>
4139 David Sherwood <david.sherwood@arm.com>
4141 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4142 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4143 * optabs.def (vec_mask_load_lanes_optab): New optab.
4144 (vec_mask_store_lanes_optab): Likewise.
4145 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4146 (MASK_STORE_LANES): Likewise.
4147 * internal-fn.c (mask_load_lanes_direct): New macro.
4148 (mask_store_lanes_direct): Likewise.
4149 (expand_mask_load_optab_fn): Handle masked operations.
4150 (expand_mask_load_lanes_optab_fn): New macro.
4151 (expand_mask_store_optab_fn): Handle masked operations.
4152 (expand_mask_store_lanes_optab_fn): New macro.
4153 (direct_mask_load_lanes_optab_supported_p): Likewise.
4154 (direct_mask_store_lanes_optab_supported_p): Likewise.
4155 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4157 (vect_load_lanes_supported): Likewise.
4158 * tree-vect-data-refs.c (strip_conversion): New function.
4159 (can_group_stmts_p): Likewise.
4160 (vect_analyze_data_ref_accesses): Use it instead of checking
4161 for a pair of assignments.
4162 (vect_store_lanes_supported): Take a masked_p parameter.
4163 (vect_load_lanes_supported): Likewise.
4164 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4165 vect_store_lanes_supported and vect_load_lanes_supported.
4166 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4167 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4168 parameter. Don't allow gaps for masked accesses.
4169 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4170 and vect_load_lanes_supported.
4171 (get_load_store_type): Take a masked_p parameter and update
4172 call to get_group_load_store_type.
4173 (vectorizable_store): Update call to get_load_store_type.
4174 Handle IFN_MASK_STORE_LANES.
4175 (vectorizable_load): Update call to get_load_store_type.
4176 Handle IFN_MASK_LOAD_LANES.
4178 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4179 Alan Hayward <alan.hayward@arm.com>
4180 David Sherwood <david.sherwood@arm.com>
4182 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4184 * config/aarch64/aarch64-protos.h
4185 (aarch64_sve_struct_memory_operand_p): Declare.
4186 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4187 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4188 (VPRED, vpred): Handle SVE structure modes.
4189 * config/aarch64/constraints.md (Utx): New constraint.
4190 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4191 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4192 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4193 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4194 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4195 structure modes. Split into pieces after RA.
4196 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4197 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4199 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4200 SVE structure modes.
4201 (aarch64_classify_address): Likewise.
4202 (sizetochar): Move earlier in file.
4203 (aarch64_print_operand): Handle SVE register lists.
4204 (aarch64_array_mode): New function.
4205 (aarch64_sve_struct_memory_operand_p): Likewise.
4206 (TARGET_ARRAY_MODE): Redefine.
4208 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4209 Alan Hayward <alan.hayward@arm.com>
4210 David Sherwood <david.sherwood@arm.com>
4212 * target.def (array_mode): New target hook.
4213 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4214 * doc/tm.texi: Regenerate.
4215 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4216 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4217 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4219 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4222 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4223 Alan Hayward <alan.hayward@arm.com>
4224 David Sherwood <david.sherwood@arm.com>
4226 * fold-const.c (fold_binary_loc): Check the argument types
4227 rather than the result type when testing for a vector operation.
4229 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4231 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4232 * doc/tm.texi: Regenerate.
4234 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4235 Alan Hayward <alan.hayward@arm.com>
4236 David Sherwood <david.sherwood@arm.com>
4238 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4239 (sve): Document new AArch64 extension.
4240 * doc/md.texi (w): Extend the description of the AArch64
4241 constraint to include SVE vectors.
4242 (Upl, Upa): Document new AArch64 predicate constraints.
4243 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4245 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4246 (msve-vector-bits=): New option.
4247 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4248 SVE when these are disabled.
4249 (sve): New extension.
4250 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4251 modes. Adjust their number of units based on aarch64_sve_vg.
4252 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4253 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4254 aarch64_addr_query_type.
4255 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4256 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4257 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4258 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4259 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4260 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4261 (aarch64_simd_imm_zero_p): Delete.
4262 (aarch64_check_zero_based_sve_index_immediate): Declare.
4263 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4264 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4265 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4266 (aarch64_sve_float_mul_immediate_p): Likewise.
4267 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4269 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4270 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4271 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4272 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4273 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4274 (aarch64_regmode_natural_size): Likewise.
4275 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4276 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4278 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4279 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4280 for VG and the SVE predicate registers.
4281 (V_ALIASES): Add a "z"-prefixed alias.
4282 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4283 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4284 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4285 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4286 (REG_CLASS_NAMES): Add entries for them.
4287 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4288 and the predicate registers.
4289 (aarch64_sve_vg): Declare.
4290 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4291 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4292 (REGMODE_NATURAL_SIZE): Define.
4293 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4295 * config/aarch64/aarch64.c: Include cfgrtl.h.
4296 (simd_immediate_info): Add a constructor for series vectors,
4297 and an associated step field.
4298 (aarch64_sve_vg): New variable.
4299 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4300 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4301 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4302 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4303 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4304 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4305 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4306 (aarch64_get_mask_mode): New functions.
4307 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4308 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4309 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4310 predicate modes and predicate registers. Explicitly restrict
4311 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4312 to store a vector mode if it is recognized by
4313 aarch64_classify_vector_mode.
4314 (aarch64_regmode_natural_size): New function.
4315 (aarch64_hard_regno_caller_save_mode): Return the original mode
4317 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4318 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4319 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4320 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4322 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4323 does not overlap dest if the function is frame-related. Handle
4325 (aarch64_split_add_offset): New function.
4326 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4327 them aarch64_add_offset.
4328 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4329 and update call to aarch64_sub_sp.
4330 (aarch64_add_cfa_expression): New function.
4331 (aarch64_expand_prologue): Pass extra temporary registers to the
4332 functions above. Handle the case in which we need to emit new
4333 DW_CFA_expressions for registers that were originally saved
4334 relative to the stack pointer, but now have to be expressed
4335 relative to the frame pointer.
4336 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4338 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4339 IP0 and IP1 values for SVE frames.
4340 (aarch64_expand_vec_series): New function.
4341 (aarch64_expand_sve_widened_duplicate): Likewise.
4342 (aarch64_expand_sve_const_vector): Likewise.
4343 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4344 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4345 into the register, rather than emitting a SET directly.
4346 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4347 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4348 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4349 (offset_9bit_signed_scaled_p): New functions.
4350 (aarch64_replicate_bitmask_imm): New function.
4351 (aarch64_bitmask_imm): Use it.
4352 (aarch64_cannot_force_const_mem): Reject expressions involving
4353 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4354 (aarch64_classify_index): Handle SVE indices, by requiring
4355 a plain register index with a scale that matches the element size.
4356 (aarch64_classify_address): Handle SVE addresses. Assert that
4357 the mode of the address is VOIDmode or an integer mode.
4358 Update call to aarch64_classify_symbol.
4359 (aarch64_classify_symbolic_expression): Update call to
4360 aarch64_classify_symbol.
4361 (aarch64_const_vec_all_in_range_p): New function.
4362 (aarch64_print_vector_float_operand): Likewise.
4363 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4364 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4365 and the FP immediates 1.0 and 0.5.
4366 (aarch64_print_address_internal): Handle SVE addresses.
4367 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4368 (aarch64_regno_regclass): Handle predicate registers.
4369 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4371 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4372 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4373 (aarch64_convert_sve_vector_bits): New function.
4374 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4375 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4377 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4378 Handle SVE vector and predicate modes. Accept VL-based constants
4379 that need only one temporary register, and VL offsets that require
4380 no temporary registers.
4381 (aarch64_conditional_register_usage): Mark the predicate registers
4382 as fixed if SVE isn't available.
4383 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4384 Return true for SVE vector and predicate modes.
4385 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4386 rather than an unsigned int. Handle SVE modes.
4387 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4389 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4391 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4392 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4393 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4394 (aarch64_sve_float_mul_immediate_p): New functions.
4395 (aarch64_sve_valid_immediate): New function.
4396 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4397 Explicitly reject structure modes. Check for INDEX constants.
4398 Handle PTRUE and PFALSE constants.
4399 (aarch64_check_zero_based_sve_index_immediate): New function.
4400 (aarch64_simd_imm_zero_p): Delete.
4401 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4402 vector modes. Accept constants in the range of CNT[BHWD].
4403 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4404 ask for an Advanced SIMD mode.
4405 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4406 (aarch64_simd_vector_alignment): Handle SVE predicates.
4407 (aarch64_vectorize_preferred_vector_alignment): New function.
4408 (aarch64_simd_vector_alignment_reachable): Use it instead of
4410 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4411 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4413 (MAX_VECT_LEN): Delete.
4414 (expand_vec_perm_d): Add a vec_flags field.
4415 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4416 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4417 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4419 (aarch64_evpc_rev): Rename to...
4420 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4421 (aarch64_evpc_rev_global): New function.
4422 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4423 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4425 (aarch64_evpc_sve_tbl): New function.
4426 (aarch64_expand_vec_perm_const_1): Update after rename of
4427 aarch64_evpc_rev. Handle SVE permutes too, trying
4428 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4429 than aarch64_evpc_tbl.
4430 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4431 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4432 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4433 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4434 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4435 (aarch64_expand_sve_vcond): New functions.
4436 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4437 of aarch64_vector_mode_p.
4438 (aarch64_dwarf_poly_indeterminate_value): New function.
4439 (aarch64_compute_pressure_classes): Likewise.
4440 (aarch64_can_change_mode_class): Likewise.
4441 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4442 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4443 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4444 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4445 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4446 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4447 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4448 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4450 (Dn, Dl, Dr): Accept const as well as const_vector.
4451 (Dz): Likewise. Compare against CONST0_RTX.
4452 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4453 of "vector" where appropriate.
4454 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4455 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4456 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4457 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4458 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4459 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4460 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4461 (v_int_equiv): Extend to SVE modes.
4462 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4464 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4465 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4466 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4467 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4468 (SVE_COND_FP_CMP): New int iterators.
4469 (perm_hilo): Handle the new unpack unspecs.
4470 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4472 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4473 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4474 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4475 (aarch64_equality_operator, aarch64_constant_vector_operand)
4476 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4477 (aarch64_sve_nonimmediate_operand): Likewise.
4478 (aarch64_sve_general_operand): Likewise.
4479 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4480 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4481 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4482 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4483 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4484 (aarch64_sve_float_arith_immediate): Likewise.
4485 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4486 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4487 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4488 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4489 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4490 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4491 (aarch64_sve_float_arith_operand): Likewise.
4492 (aarch64_sve_float_arith_with_sub_operand): Likewise.
4493 (aarch64_sve_float_mul_operand): Likewise.
4494 (aarch64_sve_vec_perm_operand): Likewise.
4495 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
4496 (aarch64_mov_operand): Accept const_poly_int and const_vector.
4497 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
4498 as well as const_vector.
4499 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
4500 in file. Use CONST0_RTX and CONSTM1_RTX.
4501 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
4502 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
4503 Use aarch64_simd_imm_zero.
4504 * config/aarch64/aarch64-sve.md: New file.
4505 * config/aarch64/aarch64.md: Include it.
4506 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
4507 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
4508 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
4509 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
4510 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
4511 (sve): New attribute.
4512 (enabled): Disable instructions with the sve attribute unless
4514 (movqi, movhi): Pass CONST_POLY_INT operaneds through
4515 aarch64_expand_mov_immediate.
4516 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
4517 CNT[BHSD] immediates.
4518 (movti): Split CONST_POLY_INT moves into two halves.
4519 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
4520 Split additions that need a temporary here if the destination
4521 is the stack pointer.
4522 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
4523 (*add<mode>3_poly_1): New instruction.
4524 (set_clobber_cc): New expander.
4526 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4528 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
4529 parameter and use it instead of GET_MODE_SIZE (innermode). Use
4530 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
4531 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
4532 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
4533 Change innermode from fixed_mode_size to machine_mode.
4534 (simplify_subreg): Update call accordingly. Handle a constant-sized
4535 subreg of a variable-length CONST_VECTOR.
4537 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4538 Alan Hayward <alan.hayward@arm.com>
4539 David Sherwood <david.sherwood@arm.com>
4541 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
4542 (add_offset_to_base): New function, split out from...
4543 (create_mem_ref): ...here. When handling a scale other than 1,
4544 check first whether the address is valid without the offset.
4545 Add it into the base if so, leaving the index and scale as-is.
4547 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4550 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
4551 fold_for_warn before checking if arg2 is INTEGER_CST.
4553 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
4555 * config/rs6000/predicates.md (load_multiple_operation): Delete.
4556 (store_multiple_operation): Delete.
4557 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
4558 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
4559 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
4560 guarded by TARGET_STRING.
4561 (rs6000_output_load_multiple): Delete.
4562 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
4563 OPTION_MASK_STRING / TARGET_STRING handling.
4564 (print_operand) <'N', 'O'>: Add comment that these are unused now.
4565 (const rs6000_opt_masks) <"string">: Change mask to 0.
4566 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
4567 (MASK_STRING): Delete.
4568 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
4570 (load_multiple): Delete.
4577 (store_multiple): Delete.
4584 (movmemsi_8reg): Delete.
4585 (corresponding unnamed define_insn): Delete.
4586 (movmemsi_6reg): Delete.
4587 (corresponding unnamed define_insn): Delete.
4588 (movmemsi_4reg): Delete.
4589 (corresponding unnamed define_insn): Delete.
4590 (movmemsi_2reg): Delete.
4591 (corresponding unnamed define_insn): Delete.
4592 (movmemsi_1reg): Delete.
4593 (corresponding unnamed define_insn): Delete.
4594 * config/rs6000/rs6000.opt (mno-string): New.
4595 (mstring): Replace by deprecation warning stub.
4596 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
4598 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4600 * regrename.c (regrename_do_replace): If replacing the same
4601 reg multiple times, try to reuse last created gen_raw_REG.
4604 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
4605 main to workaround a bug in GDB.
4607 2018-01-12 Tom de Vries <tom@codesourcery.com>
4610 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
4612 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
4614 PR rtl-optimization/80481
4615 * ira-color.c (get_cap_member): New function.
4616 (allocnos_conflict_by_live_ranges_p): Use it.
4617 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
4618 (setup_slot_coalesced_allocno_live_ranges): Ditto.
4620 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
4623 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
4624 (*saddl_se_1): Ditto.
4626 (*ssubl_se_1): Ditto.
4628 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4630 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
4631 rather than wi::to_widest for DR_INITs.
4632 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
4633 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
4634 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
4636 (vect_analyze_group_access_1): Note that here.
4638 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4640 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
4641 polynomial type sizes.
4643 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4645 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
4646 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
4647 (gimple_add_tmp_var): Likewise.
4649 2018-01-12 Martin Liska <mliska@suse.cz>
4651 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
4652 (gimple_alloc_sizes): Likewise.
4653 (dump_gimple_statistics): Use PRIu64 in printf format.
4654 * gimple.h: Change uint64_t to int.
4656 2018-01-12 Martin Liska <mliska@suse.cz>
4658 * tree-core.h: Use uint64_t instead of int.
4659 * tree.c (tree_node_counts): Likewise.
4660 (tree_node_sizes): Likewise.
4661 (dump_tree_statistics): Use PRIu64 in printf format.
4663 2018-01-12 Martin Liska <mliska@suse.cz>
4665 * Makefile.in: As qsort_chk is implemented in vec.c, add
4666 vec.o to linkage of gencfn-macros.
4667 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
4668 passing the info to record_node_allocation_statistics.
4669 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
4671 * ggc-common.c (struct ggc_usage): Add operator== and use
4672 it in operator< and compare function.
4673 * mem-stats.h (struct mem_usage): Likewise.
4674 * vec.c (struct vec_usage): Remove operator< and compare
4675 function. Can be simply inherited.
4677 2018-01-12 Martin Jambor <mjambor@suse.cz>
4680 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
4681 * tree-ssa-math-opts.c: Include domwalk.h.
4682 (convert_mult_to_fma_1): New function.
4683 (fma_transformation_info): New type.
4684 (fma_deferring_state): Likewise.
4685 (cancel_fma_deferring): New function.
4686 (result_of_phi): Likewise.
4687 (last_fma_candidate_feeds_initial_phi): Likewise.
4688 (convert_mult_to_fma): Added deferring logic, split actual
4689 transformation to convert_mult_to_fma_1.
4690 (math_opts_dom_walker): New type.
4691 (math_opts_dom_walker::after_dom_children): New method, body moved
4692 here from pass_optimize_widening_mul::execute, added deferring logic
4694 (pass_optimize_widening_mul::execute): Moved most of code to
4695 math_opts_dom_walker::after_dom_children.
4696 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
4697 * config/i386/i386.c (ix86_option_override_internal): Added
4698 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
4700 2018-01-12 Richard Biener <rguenther@suse.de>
4703 * dwarf2out.c (gen_variable_die): Do not reset old_die for
4704 inline instance vars.
4706 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
4709 * config/rx/rx.c (rx_is_restricted_memory_address):
4712 2018-01-12 Richard Biener <rguenther@suse.de>
4714 PR tree-optimization/80846
4715 * target.def (split_reduction): New target hook.
4716 * targhooks.c (default_split_reduction): New function.
4717 * targhooks.h (default_split_reduction): Declare.
4718 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
4719 target requests first reduce vectors by combining low and high
4721 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
4722 (get_vectype_for_scalar_type_and_size): Export.
4723 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
4724 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
4725 * doc/tm.texi: Regenerate.
4726 * config/i386/i386.c (ix86_split_reduction): Implement
4727 TARGET_VECTORIZE_SPLIT_REDUCTION.
4729 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4732 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
4733 in PIC mode except for TARGET_VXWORKS_RTP.
4734 * config/sparc/sparc.c: Include cfgrtl.h.
4735 (TARGET_INIT_PIC_REG): Define.
4736 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
4737 (sparc_pic_register_p): New predicate.
4738 (sparc_legitimate_address_p): Use it.
4739 (sparc_legitimize_pic_address): Likewise.
4740 (sparc_delegitimize_address): Likewise.
4741 (sparc_mode_dependent_address_p): Likewise.
4742 (gen_load_pcrel_sym): Remove 4th parameter.
4743 (load_got_register): Adjust call to above. Remove obsolete stuff.
4744 (sparc_expand_prologue): Do not call load_got_register here.
4745 (sparc_flat_expand_prologue): Likewise.
4746 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
4747 (sparc_use_pseudo_pic_reg): New function.
4748 (sparc_init_pic_reg): Likewise.
4749 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4750 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
4752 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
4754 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
4755 Add item for branch_cost.
4757 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4759 PR rtl-optimization/83565
4760 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
4761 not extend the result to a larger mode for rotate operations.
4762 (num_sign_bit_copies1): Likewise.
4764 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4767 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
4769 Use values-Xc.o for -pedantic.
4770 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
4772 2018-01-12 Martin Liska <mliska@suse.cz>
4775 * ipa-devirt.c (final_warning_record::grow_type_warnings):
4777 (possible_polymorphic_call_targets): Use it.
4778 (ipa_devirt): Likewise.
4780 2018-01-12 Martin Liska <mliska@suse.cz>
4782 * profile-count.h (enum profile_quality): Use 0 as invalid
4783 enum value of profile_quality.
4785 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
4787 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
4788 -mext-string options.
4790 2018-01-12 Richard Biener <rguenther@suse.de>
4792 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
4793 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
4794 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
4796 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
4798 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4800 * configure.ac (--with-long-double-format): Add support for the
4801 configuration option to change the default long double format on
4803 * config.gcc (powerpc*-linux*-*): Likewise.
4804 * configure: Regenerate.
4805 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
4806 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
4807 used without modification.
4809 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4811 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
4812 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
4813 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
4814 MISC_BUILTIN_SPEC_BARRIER.
4815 (rs6000_init_builtins): Likewise.
4816 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
4818 (speculation_barrier): New define_insn.
4819 * doc/extend.texi: Document __builtin_speculation_barrier.
4821 2018-01-11 Jakub Jelinek <jakub@redhat.com>
4824 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
4825 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
4826 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
4828 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
4829 integral modes instead of "ss" and "sd".
4830 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
4831 vectors with 32-bit and 64-bit elements.
4832 (vecdupssescalarmodesuffix): New mode attribute.
4833 (vec_dup<mode>): Use it.
4835 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
4838 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
4839 frame if argument is passed on stack.
4841 2018-01-11 Jakub Jelinek <jakub@redhat.com>
4844 * ree.c (combine_reaching_defs): Optimize also
4845 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
4846 reg2=any_extend(exp); reg1=reg2;, formatting fix.
4848 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
4851 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
4853 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
4856 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
4857 after they are computed.
4859 2018-01-11 Bin Cheng <bin.cheng@arm.com>
4861 PR tree-optimization/83695
4862 * gimple-loop-linterchange.cc
4863 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
4864 reset cached scev information after interchange.
4865 (pass_linterchange::execute): Remove call to scev_reset_htab.
4867 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4869 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
4870 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
4871 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
4872 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
4873 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
4874 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
4875 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
4876 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
4877 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
4878 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
4879 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
4880 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
4881 (V_lane_reg): Likewise.
4882 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
4884 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
4885 (vfmal_lane_low<mode>_intrinsic,
4886 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
4887 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
4888 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
4889 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
4890 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
4891 vfmsl_lane_high<mode>_intrinsic): New define_insns.
4893 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4895 * config/arm/arm-cpus.in (fp16fml): New feature.
4896 (ALL_SIMD): Add fp16fml.
4897 (armv8.2-a): Add fp16fml as an option.
4898 (armv8.3-a): Likewise.
4899 (armv8.4-a): Add fp16fml as part of fp16.
4900 * config/arm/arm.h (TARGET_FP16FML): Define.
4901 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
4903 * config/arm/arm-modes.def (V2HF): Define.
4904 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
4905 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
4906 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
4907 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
4908 vfmsl_low, vfmsl_high): New set of builtins.
4909 * config/arm/iterators.md (PLUSMINUS): New code iterator.
4910 (vfml_op): New code attribute.
4911 (VFMLHALVES): New int iterator.
4912 (VFML, VFMLSEL): New mode attributes.
4913 (V_reg): Define mapping for V2HF.
4914 (V_hi, V_lo): New mode attributes.
4915 (VF_constraint): Likewise.
4916 (vfml_half, vfml_half_selector): New int attributes.
4917 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
4919 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
4920 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
4922 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
4923 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
4924 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
4925 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
4927 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
4928 Document new effective target and option set.
4930 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4932 * config/arm/arm-cpus.in (armv8_4): New feature.
4933 (ARMv8_4a): New fgroup.
4934 (armv8.4-a): New arch.
4935 * config/arm/arm-tables.opt: Regenerate.
4936 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
4937 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
4938 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
4939 Add matching rules for -march=armv8.4-a and extensions.
4940 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
4942 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
4945 * config/rx/rx.md (BW): New mode attribute.
4946 (sync_lock_test_and_setsi): Add mode suffix to insn output.
4948 2018-01-11 Richard Biener <rguenther@suse.de>
4950 PR tree-optimization/83435
4951 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
4952 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
4953 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
4955 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4956 Alan Hayward <alan.hayward@arm.com>
4957 David Sherwood <david.sherwood@arm.com>
4959 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
4961 (aarch64_classify_address): Initialize it. Track polynomial offsets.
4962 (aarch64_print_address_internal): Use it to check for a zero offset.
4964 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4965 Alan Hayward <alan.hayward@arm.com>
4966 David Sherwood <david.sherwood@arm.com>
4968 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
4969 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
4970 Return a poly_int64 rather than a HOST_WIDE_INT.
4971 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
4972 rather than a HOST_WIDE_INT.
4973 * config/aarch64/aarch64.h (aarch64_frame): Protect with
4974 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
4975 hard_fp_offset, frame_size, initial_adjust, callee_offset and
4976 final_offset from HOST_WIDE_INT to poly_int64.
4977 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
4978 to_constant when getting the number of units in an Advanced SIMD
4980 (aarch64_builtin_vectorized_function): Check for a constant number
4982 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
4984 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
4985 attribute instead of GET_MODE_NUNITS.
4986 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
4987 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
4988 GET_MODE_SIZE for fixed-size registers.
4989 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
4990 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
4991 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
4992 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
4993 (aarch64_print_operand, aarch64_print_address_internal)
4994 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
4995 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
4996 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
4997 Handle polynomial GET_MODE_SIZE.
4998 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
4999 wider than SImode without modification.
5000 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5001 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5002 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5003 passing and returning SVE modes.
5004 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5005 rather than GEN_INT.
5006 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5007 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5008 (aarch64_allocate_and_probe_stack_space): Likewise.
5009 (aarch64_layout_frame): Cope with polynomial offsets.
5010 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5011 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5013 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5014 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5015 poly_int64 rather than a HOST_WIDE_INT.
5016 (aarch64_get_separate_components, aarch64_process_components)
5017 (aarch64_expand_prologue, aarch64_expand_epilogue)
5018 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5019 (aarch64_anchor_offset): New function, split out from...
5020 (aarch64_legitimize_address): ...here.
5021 (aarch64_builtin_vectorization_cost): Handle polynomial
5022 TYPE_VECTOR_SUBPARTS.
5023 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5025 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5026 number of elements from the PARALLEL rather than the mode.
5027 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5028 rather than GET_MODE_BITSIZE.
5029 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5030 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5031 (aarch64_expand_vec_perm_const_1): Handle polynomial
5032 d->perm.length () and d->perm elements.
5033 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5034 Apply to_constant to d->perm elements.
5035 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5036 polynomial CONST_VECTOR_NUNITS.
5037 (aarch64_move_pointer): Take amount as a poly_int64 rather
5039 (aarch64_progress_pointer): Avoid temporary variable.
5040 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5041 the mode attribute instead of GET_MODE.
5043 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5044 Alan Hayward <alan.hayward@arm.com>
5045 David Sherwood <david.sherwood@arm.com>
5047 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5048 x exists before using it.
5049 (aarch64_add_constant_internal): Rename to...
5050 (aarch64_add_offset_1): ...this. Replace regnum with separate
5051 src and dest rtxes. Handle the case in which they're different,
5052 including when the offset is zero. Replace scratchreg with an rtx.
5053 Use 2 additions if there is no spare register into which we can
5054 move a 16-bit constant.
5055 (aarch64_add_constant): Delete.
5056 (aarch64_add_offset): Replace reg with separate src and dest
5057 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5058 Use aarch64_add_offset_1.
5059 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5060 an rtx rather than an int. Take the delta as a poly_int64
5061 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5062 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5063 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5064 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5065 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5067 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5068 aarch64_add_constant.
5070 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5072 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5073 Use scalar_float_mode.
5075 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5077 * config/aarch64/aarch64-simd.md
5078 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5079 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5080 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5081 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5082 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5083 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5084 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5085 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5086 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5087 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5089 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5092 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5093 targ_options->x_arm_arch_string is non NULL.
5095 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5097 * config/aarch64/aarch64.h
5098 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5100 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5103 * expmed.c (emit_store_flag_force): Swap if const op0
5104 and change VOIDmode to mode of op0.
5106 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5108 PR rtl-optimization/83761
5109 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5110 than bytes to mode_for_size.
5112 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5115 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5116 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5119 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5122 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5123 when in layout mode.
5124 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5125 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5128 2018-01-10 Michael Collison <michael.collison@arm.com>
5130 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5131 * config/aarch64/aarch64-option-extension.def: Add
5132 AARCH64_OPT_EXTENSION of 'fp16fml'.
5133 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5134 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5135 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5136 * config/aarch64/constraints.md (Ui7): New constraint.
5137 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5138 (VFMLA_SEL_W): Ditto.
5141 (VFMLA16_LOW): New int iterator.
5142 (VFMLA16_HIGH): Ditto.
5143 (UNSPEC_FMLAL): New unspec.
5144 (UNSPEC_FMLSL): Ditto.
5145 (UNSPEC_FMLAL2): Ditto.
5146 (UNSPEC_FMLSL2): Ditto.
5147 (f16mac): New code attribute.
5148 * config/aarch64/aarch64-simd-builtins.def
5149 (aarch64_fmlal_lowv2sf): Ditto.
5150 (aarch64_fmlsl_lowv2sf): Ditto.
5151 (aarch64_fmlalq_lowv4sf): Ditto.
5152 (aarch64_fmlslq_lowv4sf): Ditto.
5153 (aarch64_fmlal_highv2sf): Ditto.
5154 (aarch64_fmlsl_highv2sf): Ditto.
5155 (aarch64_fmlalq_highv4sf): Ditto.
5156 (aarch64_fmlslq_highv4sf): Ditto.
5157 (aarch64_fmlal_lane_lowv2sf): Ditto.
5158 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5159 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5160 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5161 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5162 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5163 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5164 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5165 (aarch64_fmlal_lane_highv2sf): Ditto.
5166 (aarch64_fmlsl_lane_highv2sf): Ditto.
5167 (aarch64_fmlal_laneq_highv2sf): Ditto.
5168 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5169 (aarch64_fmlalq_lane_highv4sf): Ditto.
5170 (aarch64_fmlsl_lane_highv4sf): Ditto.
5171 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5172 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5173 * config/aarch64/aarch64-simd.md:
5174 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5175 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5176 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5177 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5178 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5179 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5180 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5181 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5182 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5183 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5184 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5185 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5186 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5187 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5188 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5189 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5190 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5191 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5192 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5193 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5194 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5195 (vfmlsl_low_u32): Ditto.
5196 (vfmlalq_low_u32): Ditto.
5197 (vfmlslq_low_u32): Ditto.
5198 (vfmlal_high_u32): Ditto.
5199 (vfmlsl_high_u32): Ditto.
5200 (vfmlalq_high_u32): Ditto.
5201 (vfmlslq_high_u32): Ditto.
5202 (vfmlal_lane_low_u32): Ditto.
5203 (vfmlsl_lane_low_u32): Ditto.
5204 (vfmlal_laneq_low_u32): Ditto.
5205 (vfmlsl_laneq_low_u32): Ditto.
5206 (vfmlalq_lane_low_u32): Ditto.
5207 (vfmlslq_lane_low_u32): Ditto.
5208 (vfmlalq_laneq_low_u32): Ditto.
5209 (vfmlslq_laneq_low_u32): Ditto.
5210 (vfmlal_lane_high_u32): Ditto.
5211 (vfmlsl_lane_high_u32): Ditto.
5212 (vfmlal_laneq_high_u32): Ditto.
5213 (vfmlsl_laneq_high_u32): Ditto.
5214 (vfmlalq_lane_high_u32): Ditto.
5215 (vfmlslq_lane_high_u32): Ditto.
5216 (vfmlalq_laneq_high_u32): Ditto.
5217 (vfmlslq_laneq_high_u32): Ditto.
5218 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5219 (AARCH64_FL_FOR_ARCH8_4): New.
5220 (AARCH64_ISA_F16FML): New ISA flag.
5221 (TARGET_F16FML): New feature flag for fp16fml.
5222 (doc/invoke.texi): Document new fp16fml option.
5224 2018-01-10 Michael Collison <michael.collison@arm.com>
5226 * config/aarch64/aarch64-builtins.c:
5227 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5228 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5229 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5230 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5231 (AARCH64_ISA_SHA3): New ISA flag.
5232 (TARGET_SHA3): New feature flag for sha3.
5233 * config/aarch64/iterators.md (sha512_op): New int attribute.
5234 (CRYPTO_SHA512): New int iterator.
5235 (UNSPEC_SHA512H): New unspec.
5236 (UNSPEC_SHA512H2): Ditto.
5237 (UNSPEC_SHA512SU0): Ditto.
5238 (UNSPEC_SHA512SU1): Ditto.
5239 * config/aarch64/aarch64-simd-builtins.def
5240 (aarch64_crypto_sha512hqv2di): New builtin.
5241 (aarch64_crypto_sha512h2qv2di): Ditto.
5242 (aarch64_crypto_sha512su0qv2di): Ditto.
5243 (aarch64_crypto_sha512su1qv2di): Ditto.
5244 (aarch64_eor3qv8hi): Ditto.
5245 (aarch64_rax1qv2di): Ditto.
5246 (aarch64_xarqv2di): Ditto.
5247 (aarch64_bcaxqv8hi): Ditto.
5248 * config/aarch64/aarch64-simd.md:
5249 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5250 (aarch64_crypto_sha512su0qv2di): Ditto.
5251 (aarch64_crypto_sha512su1qv2di): Ditto.
5252 (aarch64_eor3qv8hi): Ditto.
5253 (aarch64_rax1qv2di): Ditto.
5254 (aarch64_xarqv2di): Ditto.
5255 (aarch64_bcaxqv8hi): Ditto.
5256 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5257 (vsha512h2q_u64): Ditto.
5258 (vsha512su0q_u64): Ditto.
5259 (vsha512su1q_u64): Ditto.
5260 (veor3q_u16): Ditto.
5261 (vrax1q_u64): Ditto.
5263 (vbcaxq_u16): Ditto.
5264 * config/arm/types.md (crypto_sha512): New type attribute.
5265 (crypto_sha3): Ditto.
5266 (doc/invoke.texi): Document new sha3 option.
5268 2018-01-10 Michael Collison <michael.collison@arm.com>
5270 * config/aarch64/aarch64-builtins.c:
5271 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5272 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5273 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5274 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5275 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5276 (AARCH64_ISA_SM4): New ISA flag.
5277 (TARGET_SM4): New feature flag for sm4.
5278 * config/aarch64/aarch64-simd-builtins.def
5279 (aarch64_sm3ss1qv4si): Ditto.
5280 (aarch64_sm3tt1aq4si): Ditto.
5281 (aarch64_sm3tt1bq4si): Ditto.
5282 (aarch64_sm3tt2aq4si): Ditto.
5283 (aarch64_sm3tt2bq4si): Ditto.
5284 (aarch64_sm3partw1qv4si): Ditto.
5285 (aarch64_sm3partw2qv4si): Ditto.
5286 (aarch64_sm4eqv4si): Ditto.
5287 (aarch64_sm4ekeyqv4si): Ditto.
5288 * config/aarch64/aarch64-simd.md:
5289 (aarch64_sm3ss1qv4si): Ditto.
5290 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5291 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5292 (aarch64_sm4eqv4si): Ditto.
5293 (aarch64_sm4ekeyqv4si): Ditto.
5294 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5295 (sm3part_op): Ditto.
5296 (CRYPTO_SM3TT): Ditto.
5297 (CRYPTO_SM3PART): Ditto.
5298 (UNSPEC_SM3SS1): New unspec.
5299 (UNSPEC_SM3TT1A): Ditto.
5300 (UNSPEC_SM3TT1B): Ditto.
5301 (UNSPEC_SM3TT2A): Ditto.
5302 (UNSPEC_SM3TT2B): Ditto.
5303 (UNSPEC_SM3PARTW1): Ditto.
5304 (UNSPEC_SM3PARTW2): Ditto.
5305 (UNSPEC_SM4E): Ditto.
5306 (UNSPEC_SM4EKEY): Ditto.
5307 * config/aarch64/constraints.md (Ui2): New constraint.
5308 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5309 * config/arm/types.md (crypto_sm3): New type attribute.
5310 (crypto_sm4): Ditto.
5311 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5312 (vsm3tt1aq_u32): Ditto.
5313 (vsm3tt1bq_u32): Ditto.
5314 (vsm3tt2aq_u32): Ditto.
5315 (vsm3tt2bq_u32): Ditto.
5316 (vsm3partw1q_u32): Ditto.
5317 (vsm3partw2q_u32): Ditto.
5318 (vsm4eq_u32): Ditto.
5319 (vsm4ekeyq_u32): Ditto.
5320 (doc/invoke.texi): Document new sm4 option.
5322 2018-01-10 Michael Collison <michael.collison@arm.com>
5324 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5325 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5326 (AARCH64_FL_FOR_ARCH8_4): New.
5327 (AARCH64_FL_V8_4): New flag.
5328 (doc/invoke.texi): Document new armv8.4-a option.
5330 2018-01-10 Michael Collison <michael.collison@arm.com>
5332 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5333 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5334 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5335 * config/aarch64/aarch64-option-extension.def: Add
5336 AARCH64_OPT_EXTENSION of 'sha2'.
5337 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5338 (crypto): Disable sha2 and aes if crypto disabled.
5339 (crypto): Enable aes and sha2 if enabled.
5340 (simd): Disable sha2 and aes if simd disabled.
5341 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5343 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5344 (TARGET_SHA2): New feature flag for sha2.
5345 (TARGET_AES): New feature flag for aes.
5346 * config/aarch64/aarch64-simd.md:
5347 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5348 conditional on TARGET_AES.
5349 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5350 (aarch64_crypto_sha1hsi): Make pattern conditional
5352 (aarch64_crypto_sha1hv4si): Ditto.
5353 (aarch64_be_crypto_sha1hv4si): Ditto.
5354 (aarch64_crypto_sha1su1v4si): Ditto.
5355 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5356 (aarch64_crypto_sha1su0v4si): Ditto.
5357 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5358 (aarch64_crypto_sha256su0v4si): Ditto.
5359 (aarch64_crypto_sha256su1v4si): Ditto.
5360 (doc/invoke.texi): Document new aes and sha2 options.
5362 2018-01-10 Martin Sebor <msebor@redhat.com>
5364 PR tree-optimization/83781
5365 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5368 2018-01-11 Martin Sebor <msebor@gmail.com>
5369 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5371 PR tree-optimization/83501
5372 PR tree-optimization/81703
5374 * tree-ssa-strlen.c (get_string_cst): Rename...
5375 (get_string_len): ...to this. Handle global constants.
5376 (handle_char_store): Adjust.
5378 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5379 Jim Wilson <jimw@sifive.com>
5381 * config/riscv/riscv-protos.h (riscv_output_return): New.
5382 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5383 (riscv_attribute_table, riscv_output_return),
5384 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5385 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5386 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5388 (riscv_expand_prologue): Add early return for naked function.
5389 (riscv_expand_epilogue): Likewise.
5390 (riscv_function_ok_for_sibcall): Return false for naked function.
5391 (riscv_set_current_function): New.
5392 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5393 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5394 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5395 * doc/extend.texi (RISC-V Function Attributes): New.
5397 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5399 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5400 check for 128-bit long double before checking TCmode.
5401 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5402 128-bit long doubles before checking TFmode or TCmode.
5403 (FLOAT128_IBM_P): Likewise.
5405 2018-01-10 Martin Sebor <msebor@redhat.com>
5407 PR tree-optimization/83671
5408 * builtins.c (c_strlen): Unconditionally return zero for the empty
5410 Use -Warray-bounds for warnings.
5411 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5412 for non-constant array indices with COMPONENT_REF, arrays of
5413 arrays, and pointers to arrays.
5414 (gimple_fold_builtin_strlen): Determine and set length range for
5415 non-constant character arrays.
5417 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5420 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5423 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5425 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5427 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5430 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5431 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5432 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5433 indexed_or_indirect_operand predicate.
5434 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5435 (*vsx_le_perm_load_v8hi): Likewise.
5436 (*vsx_le_perm_load_v16qi): Likewise.
5437 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5438 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5439 (*vsx_le_perm_store_v8hi): Likewise.
5440 (*vsx_le_perm_store_v16qi): Likewise.
5441 (eight unnamed splitters): Likewise.
5443 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5445 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5446 * config/rs6000/emmintrin.h: Likewise.
5447 * config/rs6000/mmintrin.h: Likewise.
5448 * config/rs6000/xmmintrin.h: Likewise.
5450 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5453 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5455 * tree.c (tree_nop_conversion): Return true for location wrapper
5457 (maybe_wrap_with_location): New function.
5458 (selftest::check_strip_nops): New function.
5459 (selftest::test_location_wrappers): New function.
5460 (selftest::tree_c_tests): Call it.
5461 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5462 (maybe_wrap_with_location): New decl.
5463 (EXPR_LOCATION_WRAPPER_P): New macro.
5464 (location_wrapper_p): New inline function.
5465 (tree_strip_any_location_wrapper): New inline function.
5467 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5470 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5471 stack_realign_offset for the largest alignment of stack slot
5473 (ix86_find_max_used_stack_alignment): New function.
5474 (ix86_finalize_stack_frame_flags): Use it. Set
5475 max_used_stack_alignment if we don't realign stack.
5476 * config/i386/i386.h (machine_function): Add
5477 max_used_stack_alignment.
5479 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5481 * config/arm/arm.opt (-mbranch-cost): New option.
5482 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5485 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5488 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5489 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5491 2018-01-10 Richard Biener <rguenther@suse.de>
5494 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
5495 early out so it also covers the case where we have a non-NULL
5498 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5500 PR tree-optimization/83753
5501 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
5502 for non-strided grouped accesses if the number of elements is 1.
5504 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5507 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
5508 * i386.h (TARGET_USE_GATHER): Define.
5509 * x86-tune.def (X86_TUNE_USE_GATHER): New.
5511 2018-01-10 Martin Liska <mliska@suse.cz>
5514 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
5515 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
5517 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
5518 CLEANUP_NO_PARTITIONING is not set.
5520 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5522 * doc/rtl.texi: Remove documentation of (const ...) wrappers
5523 for vectors, as a partial revert of r254296.
5524 * rtl.h (const_vec_p): Delete.
5525 (const_vec_duplicate_p): Don't test for vector CONSTs.
5526 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
5527 * expmed.c (make_tree): Likewise.
5530 * common.md (E, F): Use CONSTANT_P instead of checking for
5532 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
5533 checking for CONST_VECTOR.
5535 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5538 * predict.c (force_edge_cold): Handle in more sane way edges
5541 2018-01-09 Carl Love <cel@us.ibm.com>
5543 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
5545 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
5546 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
5547 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
5548 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
5549 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
5550 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
5551 * config/rs6000/rs6000-protos.h: Add extern defition for
5552 rs6000_generate_float2_double_code.
5553 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
5555 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
5556 (float2_v2df): Add define_expand.
5558 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
5561 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
5562 op_mode in the force_to_mode call.
5564 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5566 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
5567 instead of checking each element individually.
5568 (aarch64_evpc_uzp): Likewise.
5569 (aarch64_evpc_zip): Likewise.
5570 (aarch64_evpc_ext): Likewise.
5571 (aarch64_evpc_rev): Likewise.
5572 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
5573 instead of checking each element individually. Return true without
5575 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
5576 whether all selected elements come from the same input, instead of
5577 checking each element individually. Remove calls to gen_rtx_REG,
5578 start_sequence and end_sequence and instead assert that no rtl is
5581 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5583 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
5584 order of HIGH and CONST checks.
5586 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5588 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
5589 if the destination isn't an SSA_NAME.
5591 2018-01-09 Richard Biener <rguenther@suse.de>
5593 PR tree-optimization/83668
5594 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
5596 (canonicalize_loop_form): ... here, renamed from ...
5597 (canonicalize_loop_closed_ssa_form): ... this and amended to
5598 swap successor edges for loop exit blocks to make us use
5599 the RPO order we need for initial schedule generation.
5601 2018-01-09 Joseph Myers <joseph@codesourcery.com>
5603 PR tree-optimization/64811
5604 * match.pd: When optimizing comparisons with Inf, avoid
5605 introducing or losing exceptions from comparisons with NaN.
5607 2018-01-09 Martin Liska <mliska@suse.cz>
5610 * asan.c (shadow_mem_size): Add gcc_assert.
5612 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
5614 Don't save registers in main().
5617 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
5618 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
5619 * config/avr/avr.c (avr_set_current_function): Don't error if
5620 naked, OS_task or OS_main are specified at the same time.
5621 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
5623 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
5625 * common/config/avr/avr-common.c (avr_option_optimization_table):
5626 Switch on -mmain-is-OS_task for optimizing compilations.
5628 2018-01-09 Richard Biener <rguenther@suse.de>
5630 PR tree-optimization/83572
5631 * graphite.c: Include cfganal.h.
5632 (graphite_transform_loops): Connect infinite loops to exit
5633 and remove fake edges at the end.
5635 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5637 * ipa-inline.c (edge_badness): Revert accidental checkin.
5639 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5642 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
5643 symbols; not inline clones.
5645 2018-01-09 Jakub Jelinek <jakub@redhat.com>
5648 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
5649 hard registers. Formatting fixes.
5651 PR preprocessor/83722
5652 * gcc.c (try_generate_repro): Pass
5653 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
5654 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
5657 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
5658 Kito Cheng <kito.cheng@gmail.com>
5660 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
5661 (riscv_leaf_function_p): Delete.
5662 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
5664 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5666 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
5668 (do_ifelse): New function.
5669 (do_isel): New function.
5670 (do_sub3): New function.
5671 (do_add3): New function.
5672 (do_load_mask_compare): New function.
5673 (do_overlap_load_compare): New function.
5674 (expand_compare_loop): New function.
5675 (expand_block_compare): Call expand_compare_loop() when appropriate.
5676 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
5678 (-mblock-compare-inline-loop-limit): New option.
5680 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5683 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
5684 Reverse order of second and third operands in first alternative.
5685 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
5686 of first and second elements in UNSPEC_VPERMR vector.
5687 (altivec_expand_vec_perm_le): Likewise.
5689 2018-01-08 Jeff Law <law@redhat.com>
5691 PR rtl-optimizatin/81308
5692 * tree-switch-conversion.c (cfg_altered): New file scoped static.
5693 (process_switch): If group_case_labels makes a change, then set
5695 (pass_convert_switch::execute): If a switch is converted, then
5696 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
5698 PR rtl-optimization/81308
5699 * recog.c (split_all_insns): Conditionally cleanup the CFG after
5702 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
5704 PR target/83663 - Revert r255946
5705 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
5706 generation for cases where splatting a value is not useful.
5707 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
5708 across a vec_duplicate and a paradoxical subreg forming a vector
5709 mode to a vec_concat.
5711 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5713 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
5714 -march=armv8.3-a variants.
5715 * config/arm/t-multilib: Likewise.
5716 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
5718 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5720 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
5722 (cceq_ior_compare_complement): Give it a name so I can use it, and
5723 change boolean_or_operator predicate to boolean_operator so it can
5724 be used to generate a crand.
5725 (eqne): New code iterator.
5726 (bd/bd_neg): New code_attrs.
5727 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
5728 a single define_insn.
5729 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
5730 decrement (bdnzt/bdnzf/bdzt/bdzf).
5731 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
5732 with the new names of the branch decrement patterns, and added the
5733 names of the branch decrement conditional patterns.
5735 2018-01-08 Richard Biener <rguenther@suse.de>
5737 PR tree-optimization/83563
5738 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
5741 2018-01-08 Richard Biener <rguenther@suse.de>
5744 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
5746 2018-01-08 Richard Biener <rguenther@suse.de>
5748 PR tree-optimization/83685
5749 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
5750 references to abnormals.
5752 2018-01-08 Richard Biener <rguenther@suse.de>
5755 * dwarf2out.c (output_indirect_strings): Handle empty
5756 skeleton_debug_str_hash.
5757 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
5759 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5761 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
5762 (emit_store_direct): Likewise.
5763 (arc_trampoline_adjust_address): Likewise.
5764 (arc_asm_trampoline_template): New function.
5765 (arc_initialize_trampoline): Use asm_trampoline_template.
5766 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
5767 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
5768 * config/arc/arc.md (flush_icache): Delete pattern.
5770 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5772 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
5773 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
5776 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5779 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
5780 by not USED_FOR_TARGET.
5781 (make_pass_resolve_sw_modes): Likewise.
5783 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5785 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
5788 2018-01-08 Richard Biener <rguenther@suse.de>
5791 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
5793 2018-01-08 Richard Biener <rguenther@suse.de>
5796 * match.pd ((t * 2) / 2) -> t): Add missing :c.
5798 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
5801 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
5802 basic blocks with a small number of successors.
5803 (convert_control_dep_chain_into_preds): Improve handling of
5805 (dump_predicates): Split apart into...
5806 (dump_pred_chain): ...here...
5807 (dump_pred_info): ...and here.
5808 (can_one_predicate_be_invalidated_p): Add debugging printfs.
5809 (can_chain_union_be_invalidated_p): Improve check for invalidation
5811 (uninit_uses_cannot_happen): Avoid unnecessary if
5812 convert_control_dep_chain_into_preds yielded nothing.
5814 2018-01-06 Martin Sebor <msebor@redhat.com>
5816 PR tree-optimization/83640
5817 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
5818 subtracting negative offset from size.
5819 (builtin_access::overlap): Adjust offset bounds of the access to fall
5820 within the size of the object if possible.
5822 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
5824 PR rtl-optimization/83699
5825 * expmed.c (extract_bit_field_1): Restrict the vector usage of
5826 extract_bit_field_as_subreg to cases in which the extracted
5827 value is also a vector.
5829 * lra-constraints.c (process_alt_operands): Test for the equivalence
5830 substitutions when detecting a possible reload cycle.
5832 2018-01-06 Jakub Jelinek <jakub@redhat.com>
5835 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
5836 by default if flag_selective_schedling{,2}. Formatting fixes.
5838 PR rtl-optimization/83682
5839 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
5840 if it has non-VECTOR_MODE element mode.
5841 (vec_duplicate_p): Likewise.
5844 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
5845 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
5847 2018-01-05 Jakub Jelinek <jakub@redhat.com>
5850 * config/i386/i386-builtin.def
5851 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
5852 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
5853 Require also OPTION_MASK_ISA_AVX512F in addition to
5854 OPTION_MASK_ISA_GFNI.
5855 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
5856 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
5857 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
5858 to OPTION_MASK_ISA_GFNI.
5859 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
5860 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
5861 OPTION_MASK_ISA_AVX512BW.
5862 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
5863 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
5864 addition to OPTION_MASK_ISA_GFNI.
5865 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
5866 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
5867 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
5868 to OPTION_MASK_ISA_GFNI.
5869 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
5870 a requirement for all ISAs rather than any of them with a few
5872 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
5874 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
5875 bitmasks to be enabled with 3 exceptions, instead of requiring any
5876 enabled ISA with lots of exceptions.
5877 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
5878 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
5879 Change avx512bw in isa attribute to avx512f.
5880 * config/i386/sgxintrin.h: Add license boilerplate.
5881 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
5882 to __AVX512F__ and __AVX512VL to __AVX512VL__.
5883 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
5884 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
5886 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
5887 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
5888 temporarily sse2 rather than sse if not enabled already.
5891 * config/i386/sse.md (VI248_VLBW): Rename to ...
5892 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
5893 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
5894 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
5895 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
5896 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
5897 mode iterator instead of VI248_VLBW.
5899 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
5901 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
5902 (record_modified): Skip clobbers; add debug output.
5903 (param_change_prob): Use sreal frequencies.
5905 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
5907 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
5908 punt for user-aligned variables.
5910 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
5912 * tree-chrec.c (chrec_contains_symbols): Return true for
5915 2018-01-05 Sudakshina Das <sudi.das@arm.com>
5918 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
5919 of (x|y) == x for BICS pattern.
5921 2018-01-05 Jakub Jelinek <jakub@redhat.com>
5923 PR tree-optimization/83605
5924 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
5925 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
5928 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
5930 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
5931 * config/epiphany/rtems.h: New file.
5933 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5934 Uros Bizjak <ubizjak@gmail.com>
5937 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
5938 QIreg_operand instead of register_operand predicate.
5939 * config/i386/i386.c (ix86_rop_should_change_byte_p,
5940 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
5941 comments instead of -fmitigate[-_]rop.
5943 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5946 * cgraphunit.c (symbol_table::compile): Switch to text_section
5947 before calling assembly_start debug hook.
5948 * run-rtl-passes.c (run_rtl_passes): Likewise.
5951 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5953 * tree-vrp.c (extract_range_from_binary_expr_1): Check
5954 range_int_cst_p rather than !symbolic_range_p before calling
5955 extract_range_from_multiplicative_op_1.
5957 2018-01-04 Jeff Law <law@redhat.com>
5959 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
5960 redundant test in assertion.
5962 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5964 * doc/rtl.texi: Document machine_mode wrapper classes.
5966 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5968 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
5971 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5973 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
5974 the VEC_PERM_EXPR fold to fail.
5976 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5979 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
5980 to switched_sections.
5982 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5985 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
5988 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
5991 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
5992 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
5994 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5997 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
5998 is BLKmode and bitpos not zero or mode change is needed.
6000 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6003 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6006 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6009 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6010 instead of MULT rtx. Update all corresponding splitters.
6012 (*ssub<modesuffix>): Ditto.
6014 (*cmp_sadd_di): Update split patterns.
6015 (*cmp_sadd_si): Ditto.
6016 (*cmp_sadd_sidi): Ditto.
6017 (*cmp_ssub_di): Ditto.
6018 (*cmp_ssub_si): Ditto.
6019 (*cmp_ssub_sidi): Ditto.
6020 * config/alpha/predicates.md (const23_operand): New predicate.
6021 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6022 Look for ASHIFT, not MULT inner operand.
6023 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6025 2018-01-04 Martin Liska <mliska@suse.cz>
6027 PR gcov-profile/83669
6028 * gcov.c (output_intermediate_file): Add version to intermediate
6030 * doc/gcov.texi: Document new field 'version' in intermediate
6031 file format. Fix location of '-k' option of gcov command.
6033 2018-01-04 Martin Liska <mliska@suse.cz>
6036 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6038 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6040 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6042 2018-01-03 Martin Sebor <msebor@redhat.com>
6044 PR tree-optimization/83655
6045 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6046 checking calls with invalid arguments.
6048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6050 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6051 (vectorizable_mask_load_store): Delete.
6052 (vectorizable_call): Return false for masked loads and stores.
6053 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6054 instead of gimple_assign_rhs1.
6055 (vectorizable_load): Handle IFN_MASK_LOAD.
6056 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6060 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6062 (vectorizable_mask_load_store): ...here.
6063 (vectorizable_load): ...and here.
6065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6067 * tree-vect-stmts.c (vect_build_all_ones_mask)
6068 (vect_build_zero_merge_argument): New functions, split out from...
6069 (vectorizable_load): ...here.
6071 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6073 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6075 (vectorizable_mask_load_store): ...here.
6076 (vectorizable_store): ...and here.
6078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6080 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6082 (vectorizable_mask_load_store): ...here.
6084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6086 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6087 (vect_model_store_cost): Take a vec_load_store_type instead of a
6089 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6090 (vect_model_store_cost): Take a vec_load_store_type instead of a
6092 (vectorizable_mask_load_store): Update accordingly.
6093 (vectorizable_store): Likewise.
6094 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6096 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6098 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6099 IFN_MASK_LOAD calls here rather than...
6100 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6102 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6103 Alan Hayward <alan.hayward@arm.com>
6104 David Sherwood <david.sherwood@arm.com>
6106 * expmed.c (extract_bit_field_1): For vector extracts,
6107 fall back to extract_bit_field_as_subreg if vec_extract
6110 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6111 Alan Hayward <alan.hayward@arm.com>
6112 David Sherwood <david.sherwood@arm.com>
6114 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6115 they are variable or constant sized.
6116 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6117 slots for constant-sized data.
6119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6120 Alan Hayward <alan.hayward@arm.com>
6121 David Sherwood <david.sherwood@arm.com>
6123 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6124 handling COND_EXPRs with boolean comparisons, try to find a better
6125 basis for the mask type than the boolean itself.
6127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6129 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6130 is calculated and how it can be overridden.
6131 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6132 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6134 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6138 Alan Hayward <alan.hayward@arm.com>
6139 David Sherwood <david.sherwood@arm.com>
6141 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6142 Remove the mode argument.
6143 (aarch64_simd_valid_immediate): Remove the mode and inverse
6145 * config/aarch64/iterators.md (bitsize): New iterator.
6146 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6147 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6148 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6149 aarch64_simd_valid_immediate.
6150 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6151 (aarch64_reg_or_bic_imm): Likewise.
6152 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6153 with an insn_type enum and msl with a modifier_type enum.
6154 Replace element_width with a scalar_mode. Change the shift
6155 to unsigned int. Add constructors for scalar_float_mode and
6156 scalar_int_mode elements.
6157 (aarch64_vect_float_const_representable_p): Delete.
6158 (aarch64_can_const_movi_rtx_p)
6159 (aarch64_simd_scalar_immediate_valid_for_move)
6160 (aarch64_simd_make_constant): Update call to
6161 aarch64_simd_valid_immediate.
6162 (aarch64_advsimd_valid_immediate_hs): New function.
6163 (aarch64_advsimd_valid_immediate): Likewise.
6164 (aarch64_simd_valid_immediate): Remove mode and inverse
6165 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6166 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6167 and aarch64_float_const_representable_p on the result.
6168 (aarch64_output_simd_mov_immediate): Remove mode argument.
6169 Update call to aarch64_simd_valid_immediate and use of
6170 simd_immediate_info.
6171 (aarch64_output_scalar_simd_mov_immediate): Update call
6174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6175 Alan Hayward <alan.hayward@arm.com>
6176 David Sherwood <david.sherwood@arm.com>
6178 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6179 (mode_nunits): Likewise CONST_MODE_NUNITS.
6180 * machmode.def (ADJUST_NUNITS): Document.
6181 * genmodes.c (mode_data::need_nunits_adj): New field.
6182 (blank_mode): Update accordingly.
6183 (adj_nunits): New variable.
6184 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6186 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6187 listed in adj_nunits.
6188 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6189 listed in adj_nunits. Don't emit case statements for such modes.
6190 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6191 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6192 nothing if adj_nunits is nonnull.
6193 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6194 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6195 (emit_mode_fbit): Update use of print_maybe_const_decl.
6196 (emit_move_size): Likewise. Treat the array as non-const
6198 (emit_mode_adjustments): Handle adj_nunits.
6200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6202 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6203 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6204 (VECTOR_MODES): Use it.
6205 (make_vector_modes): Take the prefix as an argument.
6207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6208 Alan Hayward <alan.hayward@arm.com>
6209 David Sherwood <david.sherwood@arm.com>
6211 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6212 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6213 for MODE_VECTOR_BOOL.
6214 * machmode.def (VECTOR_BOOL_MODE): Document.
6215 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6216 (make_vector_bool_mode): New function.
6217 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6219 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6220 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6222 * stor-layout.c (int_mode_for_mode): Likewise.
6223 * tree.c (build_vector_type_for_mode): Likewise.
6224 * varasm.c (output_constant_pool_2): Likewise.
6225 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6226 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6227 for MODE_VECTOR_BOOL.
6228 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6229 of mode class checks.
6230 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6231 instead of a list of mode class checks.
6232 (expand_vector_scalar_condition): Likewise.
6233 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6235 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6236 Alan Hayward <alan.hayward@arm.com>
6237 David Sherwood <david.sherwood@arm.com>
6239 * machmode.h (mode_size): Change from unsigned short to
6241 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6242 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6243 or if measurement_type is not polynomial.
6244 (fixed_size_mode::includes_p): Check for constant-sized modes.
6245 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6246 return a poly_uint16 rather than an unsigned short.
6247 (emit_mode_size): Change the type of mode_size from unsigned short
6248 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6249 (emit_mode_adjustments): Cope with polynomial vector sizes.
6250 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6252 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6254 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6255 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6256 * caller-save.c (setup_save_areas): Likewise.
6257 (replace_reg_with_saved_mem): Likewise.
6258 * calls.c (emit_library_call_value_1): Likewise.
6259 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6260 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6261 (gen_lowpart_for_combine): Likewise.
6262 * convert.c (convert_to_integer_1): Likewise.
6263 * cse.c (equiv_constant, cse_insn): Likewise.
6264 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6265 (cselib_subst_to_values): Likewise.
6266 * dce.c (word_dce_process_block): Likewise.
6267 * df-problems.c (df_word_lr_mark_ref): Likewise.
6268 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6269 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6270 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6271 (rtl_for_decl_location): Likewise.
6272 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6273 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6274 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6275 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6276 (expand_expr_real_1): Likewise.
6277 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6278 (pad_below): Likewise.
6279 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6280 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6281 * ira.c (get_subreg_tracking_sizes): Likewise.
6282 * ira-build.c (ira_create_allocno_objects): Likewise.
6283 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6284 (ira_sort_regnos_for_alter_reg): Likewise.
6285 * ira-costs.c (record_operand_costs): Likewise.
6286 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6287 (resolve_simple_move): Likewise.
6288 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6289 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6290 (lra_constraints): Likewise.
6291 (CONST_POOL_OK_P): Reject variable-sized modes.
6292 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6293 (add_pseudo_to_slot, lra_spill): Likewise.
6294 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6295 * optabs-query.c (get_best_extraction_insn): Likewise.
6296 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6297 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6298 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6299 * recog.c (offsettable_address_addr_space_p): Likewise.
6300 * regcprop.c (maybe_mode_change): Likewise.
6301 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6302 * regrename.c (build_def_use): Likewise.
6303 * regstat.c (dump_reg_info): Likewise.
6304 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6305 (find_reloads, find_reloads_subreg_address): Likewise.
6306 * reload1.c (eliminate_regs_1): Likewise.
6307 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6308 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6309 (simplify_binary_operation_1, simplify_subreg): Likewise.
6310 * targhooks.c (default_function_arg_padding): Likewise.
6311 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6312 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6313 (verify_gimple_assign_ternary): Likewise.
6314 * tree-inline.c (estimate_move_cost): Likewise.
6315 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6316 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6317 (get_address_cost_ainc): Likewise.
6318 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6319 (vect_supportable_dr_alignment): Likewise.
6320 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6321 (vectorizable_reduction): Likewise.
6322 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6323 (vectorizable_operation, vectorizable_load): Likewise.
6324 * tree.c (build_same_sized_truth_vector_type): Likewise.
6325 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6326 * var-tracking.c (emit_note_insn_var_location): Likewise.
6327 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6328 (ADDR_VEC_ALIGN): Likewise.
6330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6331 Alan Hayward <alan.hayward@arm.com>
6332 David Sherwood <david.sherwood@arm.com>
6334 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6336 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6337 or if measurement_type is polynomial.
6338 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6339 * combine.c (make_extraction): Likewise.
6340 * dse.c (find_shift_sequence): Likewise.
6341 * dwarf2out.c (mem_loc_descriptor): Likewise.
6342 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6343 (extract_bit_field, extract_low_bits): Likewise.
6344 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6345 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6346 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6347 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6348 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6349 * reload.c (find_reloads): Likewise.
6350 * reload1.c (alter_reg): Likewise.
6351 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6352 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6353 * tree-if-conv.c (predicate_mem_writes): Likewise.
6354 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6355 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6356 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6357 * valtrack.c (dead_debug_insert_temp): Likewise.
6358 * varasm.c (mergeable_constant_section): Likewise.
6359 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6361 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6362 Alan Hayward <alan.hayward@arm.com>
6363 David Sherwood <david.sherwood@arm.com>
6365 * expr.c (expand_assignment): Cope with polynomial mode sizes
6366 when assigning to a CONCAT.
6368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6369 Alan Hayward <alan.hayward@arm.com>
6370 David Sherwood <david.sherwood@arm.com>
6372 * machmode.h (mode_precision): Change from unsigned short to
6374 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6376 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6377 or if measurement_type is not polynomial.
6378 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6379 in which the mode is already known to be a scalar_int_mode.
6380 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6381 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6383 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6384 for GET_MODE_PRECISION.
6385 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6386 for GET_MODE_PRECISION.
6387 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6389 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6390 (expand_field_assignment, make_extraction): Likewise.
6391 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6392 (get_last_value): Likewise.
6393 * convert.c (convert_to_integer_1): Likewise.
6394 * cse.c (cse_insn): Likewise.
6395 * expr.c (expand_expr_real_1): Likewise.
6396 * lra-constraints.c (simplify_operand_subreg): Likewise.
6397 * optabs-query.c (can_atomic_load_p): Likewise.
6398 * optabs.c (expand_atomic_load): Likewise.
6399 (expand_atomic_store): Likewise.
6400 * ree.c (combine_reaching_defs): Likewise.
6401 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6402 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6403 * tree.h (type_has_mode_precision_p): Likewise.
6404 * ubsan.c (instrument_si_overflow): Likewise.
6406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6407 Alan Hayward <alan.hayward@arm.com>
6408 David Sherwood <david.sherwood@arm.com>
6410 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6411 polynomial numbers of units.
6412 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6413 (valid_vector_subparts_p): New function.
6414 (build_vector_type): Remove temporary shim and take the number
6415 of units as a poly_uint64 rather than an int.
6416 (build_opaque_vector_type): Take the number of units as a
6417 poly_uint64 rather than an int.
6418 * tree.c (build_vector_from_ctor): Handle polynomial
6419 TYPE_VECTOR_SUBPARTS.
6420 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6421 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6422 (build_vector_from_val): If the number of units is variable,
6423 use build_vec_duplicate_cst for constant operands and
6424 VEC_DUPLICATE_EXPR otherwise.
6425 (make_vector_type): Remove temporary is_constant ().
6426 (build_vector_type, build_opaque_vector_type): Take the number of
6427 units as a poly_uint64 rather than an int.
6428 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6430 * cfgexpand.c (expand_debug_expr): Likewise.
6431 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6432 (store_constructor, expand_expr_real_1): Likewise.
6433 (const_scalar_mask_from_tree): Likewise.
6434 * fold-const-call.c (fold_const_reduction): Likewise.
6435 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6436 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6437 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6438 (fold_relational_const): Likewise.
6439 (native_interpret_vector): Likewise. Change the size from an
6440 int to an unsigned int.
6441 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6442 TYPE_VECTOR_SUBPARTS.
6443 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6444 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6445 duplicating a non-constant operand into a variable-length vector.
6446 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6447 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6448 * ipa-icf.c (sem_variable::equals): Likewise.
6449 * match.pd: Likewise.
6450 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6451 * print-tree.c (print_node): Likewise.
6452 * stor-layout.c (layout_type): Likewise.
6453 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6454 * tree-cfg.c (verify_gimple_comparison): Likewise.
6455 (verify_gimple_assign_binary): Likewise.
6456 (verify_gimple_assign_ternary): Likewise.
6457 (verify_gimple_assign_single): Likewise.
6458 * tree-pretty-print.c (dump_generic_node): Likewise.
6459 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6460 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6461 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6462 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6463 (vect_shift_permute_load_chain): Likewise.
6464 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6465 (expand_vector_condition, optimize_vector_constructor): Likewise.
6466 (lower_vec_perm, get_compute_type): Likewise.
6467 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6468 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6469 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6470 (vect_recog_mask_conversion_pattern): Likewise.
6471 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6472 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6473 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6474 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6475 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6476 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6477 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6478 (supportable_widening_operation): Likewise.
6479 (supportable_narrowing_operation): Likewise.
6480 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6482 * varasm.c (output_constant): Likewise.
6484 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6485 Alan Hayward <alan.hayward@arm.com>
6486 David Sherwood <david.sherwood@arm.com>
6488 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6489 so that both the length == 3 and length != 3 cases set up their
6490 own permute vectors. Add comments explaining why we know the
6491 number of elements is constant.
6492 (vect_permute_load_chain): Likewise.
6494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6495 Alan Hayward <alan.hayward@arm.com>
6496 David Sherwood <david.sherwood@arm.com>
6498 * machmode.h (mode_nunits): Change from unsigned char to
6500 (ONLY_FIXED_SIZE_MODES): New macro.
6501 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
6502 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
6503 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
6505 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
6506 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
6507 or if measurement_type is not polynomial.
6508 * genmodes.c (ZERO_COEFFS): New macro.
6509 (emit_mode_nunits_inline): Make mode_nunits_inline return a
6511 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
6512 Use ZERO_COEFFS when emitting initializers.
6513 * data-streamer.h (bp_pack_poly_value): New function.
6514 (bp_unpack_poly_value): Likewise.
6515 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6516 for GET_MODE_NUNITS.
6517 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6518 for GET_MODE_NUNITS.
6519 * tree.c (make_vector_type): Remove temporary shim and make
6520 the real function take the number of units as a poly_uint64
6522 (build_vector_type_for_mode): Handle polynomial nunits.
6523 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
6524 * emit-rtl.c (const_vec_series_p_1): Likewise.
6525 (gen_rtx_CONST_VECTOR): Likewise.
6526 * fold-const.c (test_vec_duplicate_folding): Likewise.
6527 * genrecog.c (validate_pattern): Likewise.
6528 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
6529 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6530 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
6531 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
6532 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
6533 * rtlanal.c (subreg_get_info): Likewise.
6534 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
6535 (vect_grouped_load_supported): Likewise.
6536 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
6537 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
6538 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
6539 (simplify_const_unary_operation, simplify_binary_operation_1)
6540 (simplify_const_binary_operation, simplify_ternary_operation)
6541 (test_vector_ops_duplicate, test_vector_ops): Likewise.
6542 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
6543 instead of CONST_VECTOR_NUNITS.
6544 * varasm.c (output_constant_pool_2): Likewise.
6545 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
6546 explicit-encoded elements in the XVEC for variable-length vectors.
6548 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6550 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
6552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6553 Alan Hayward <alan.hayward@arm.com>
6554 David Sherwood <david.sherwood@arm.com>
6556 * coretypes.h (fixed_size_mode): Declare.
6557 (fixed_size_mode_pod): New typedef.
6558 * builtins.h (target_builtins::x_apply_args_mode)
6559 (target_builtins::x_apply_result_mode): Change type to
6560 fixed_size_mode_pod.
6561 * builtins.c (apply_args_size, apply_result_size, result_vector)
6562 (expand_builtin_apply_args_1, expand_builtin_apply)
6563 (expand_builtin_return): Update accordingly.
6565 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6567 * cse.c (hash_rtx_cb): Hash only the encoded elements.
6568 * cselib.c (cselib_hash_rtx): Likewise.
6569 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
6570 CONST_VECTOR encoding.
6572 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6573 Jeff Law <law@redhat.com>
6576 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
6577 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
6578 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
6579 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
6582 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
6583 explicitly probe *sp in a noreturn function if there were any callee
6584 register saves or frame pointer is needed.
6586 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6589 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
6590 BLKmode for ternary, binary or unary expressions.
6593 * var-tracking.c (delete_vta_debug_insn): New inline function.
6594 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
6595 insns from get_insns () to NULL instead of each bb separately.
6596 Use delete_vta_debug_insn. No longer static.
6597 (vt_debug_insns_local, variable_tracking_main_1): Adjust
6598 delete_vta_debug_insns callers.
6599 * rtl.h (delete_vta_debug_insns): Declare.
6600 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
6601 instead of variable_tracking_main.
6603 2018-01-03 Martin Sebor <msebor@redhat.com>
6605 PR tree-optimization/83603
6606 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
6607 arguments past the endof the argument list in functions declared
6608 without a prototype.
6609 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6610 Avoid checking when arguments are null.
6612 2018-01-03 Martin Sebor <msebor@redhat.com>
6615 * doc/extend.texi (attribute const): Fix a typo.
6616 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
6617 issuing -Wsuggest-attribute for void functions.
6619 2018-01-03 Martin Sebor <msebor@redhat.com>
6621 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
6622 offset_int::from instead of wide_int::to_shwi.
6623 (maybe_diag_overlap): Remove assertion.
6624 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
6625 * gimple-ssa-sprintf.c (format_directive): Same.
6626 (parse_directive): Same.
6627 (sprintf_dom_walker::compute_format_length): Same.
6628 (try_substitute_return_value): Same.
6630 2018-01-03 Jeff Law <law@redhat.com>
6633 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
6634 non-constant residual for zero at runtime and avoid probing in
6635 that case. Reorganize code for trailing problem to mirror handling
6638 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6640 PR tree-optimization/83501
6641 * tree-ssa-strlen.c (get_string_cst): New.
6642 (handle_char_store): Call get_string_cst.
6644 2018-01-03 Martin Liska <mliska@suse.cz>
6646 PR tree-optimization/83593
6647 * tree-ssa-strlen.c: Include tree-cfg.h.
6648 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
6649 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
6650 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
6652 (strlen_dom_walker::before_dom_children): Call
6653 gimple_purge_dead_eh_edges. Dump tranformation with details
6655 (strlen_dom_walker::before_dom_children): Update call by adding
6656 new argument cleanup_eh.
6657 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
6659 2018-01-03 Martin Liska <mliska@suse.cz>
6662 * cif-code.def (VARIADIC_THUNK): New enum value.
6663 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
6666 2018-01-03 Jan Beulich <jbeulich@suse.com>
6668 * sse.md (mov<mode>_internal): Tighten condition for when to use
6669 vmovdqu<ssescalarsize> for TI and OI modes.
6671 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6673 Update copyright years.
6675 2018-01-03 Martin Liska <mliska@suse.cz>
6678 * ipa-visibility.c (function_and_variable_visibility): Skip
6679 functions with noipa attribure.
6681 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6683 * gcc.c (process_command): Update copyright notice dates.
6684 * gcov-dump.c (print_version): Ditto.
6685 * gcov.c (print_version): Ditto.
6686 * gcov-tool.c (print_version): Ditto.
6687 * gengtype.c (create_file): Ditto.
6688 * doc/cpp.texi: Bump @copying's copyright year.
6689 * doc/cppinternals.texi: Ditto.
6690 * doc/gcc.texi: Ditto.
6691 * doc/gccint.texi: Ditto.
6692 * doc/gcov.texi: Ditto.
6693 * doc/install.texi: Ditto.
6694 * doc/invoke.texi: Ditto.
6696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6698 * vector-builder.h (vector_builder::m_full_nelts): Change from
6699 unsigned int to poly_uint64.
6700 (vector_builder::full_nelts): Update prototype accordingly.
6701 (vector_builder::new_vector): Likewise.
6702 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
6703 (vector_builder::operator ==): Likewise.
6704 (vector_builder::finalize): Likewise.
6705 * int-vector-builder.h (int_vector_builder::int_vector_builder):
6706 Take the number of elements as a poly_uint64 rather than an
6708 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
6709 from unsigned int to poly_uint64.
6710 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
6711 (vec_perm_indices::new_vector): Likewise.
6712 (vec_perm_indices::length): Likewise.
6713 (vec_perm_indices::nelts_per_input): Likewise.
6714 (vec_perm_indices::input_nelts): Likewise.
6715 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
6716 number of elements per input as a poly_uint64 rather than an
6717 unsigned int. Use the original encoding for variable-length
6718 vectors, rather than clamping each individual element.
6719 For the second and subsequent elements in each pattern,
6720 clamp the step and base before clamping their sum.
6721 (vec_perm_indices::series_p): Handle polynomial element counts.
6722 (vec_perm_indices::all_in_range_p): Likewise.
6723 (vec_perm_indices_to_tree): Likewise.
6724 (vec_perm_indices_to_rtx): Likewise.
6725 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
6726 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
6727 (tree_vector_builder::new_binary_operation): Handle polynomial
6728 element counts. Return false if we need to know the number
6729 of elements at compile time.
6730 * fold-const.c (fold_vec_perm): Punt if the number of elements
6731 isn't known at compile time.
6733 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6735 * vec-perm-indices.h (vec_perm_builder): Change element type
6736 from HOST_WIDE_INT to poly_int64.
6737 (vec_perm_indices::element_type): Update accordingly.
6738 (vec_perm_indices::clamp): Handle polynomial element_types.
6739 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
6740 (vec_perm_indices::all_in_range_p): Likewise.
6741 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
6743 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
6744 polynomial vec_perm_indices element types.
6745 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
6746 * fold-const.c (fold_vec_perm): Likewise.
6747 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
6748 * tree-vect-generic.c (lower_vec_perm): Likewise.
6749 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
6750 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
6751 element type to HOST_WIDE_INT.
6753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6754 Alan Hayward <alan.hayward@arm.com>
6755 David Sherwood <david.sherwood@arm.com>
6757 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
6758 rather than an int. Use plus_constant.
6759 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
6760 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
6762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6763 Alan Hayward <alan.hayward@arm.com>
6764 David Sherwood <david.sherwood@arm.com>
6766 * calls.c (emit_call_1, expand_call): Change struct_value_size from
6767 a HOST_WIDE_INT to a poly_int64.
6769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6770 Alan Hayward <alan.hayward@arm.com>
6771 David Sherwood <david.sherwood@arm.com>
6773 * calls.c (load_register_parameters): Cope with polynomial
6774 mode sizes. Require a constant size for BLKmode parameters
6775 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
6776 forces a parameter to be padded at the lsb end in order to
6777 fill a complete number of words, require the parameter size
6778 to be ordered wrt UNITS_PER_WORD.
6780 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6781 Alan Hayward <alan.hayward@arm.com>
6782 David Sherwood <david.sherwood@arm.com>
6784 * reload1.c (spill_stack_slot_width): Change element type
6785 from unsigned int to poly_uint64_pod.
6786 (alter_reg): Treat mode sizes as polynomial.
6788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6789 Alan Hayward <alan.hayward@arm.com>
6790 David Sherwood <david.sherwood@arm.com>
6792 * reload.c (complex_word_subreg_p): New function.
6793 (reload_inner_reg_of_subreg, push_reload): Use it.
6795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6796 Alan Hayward <alan.hayward@arm.com>
6797 David Sherwood <david.sherwood@arm.com>
6799 * lra-constraints.c (process_alt_operands): Reject matched
6800 operands whose sizes aren't ordered.
6801 (match_reload): Refer to this check here.
6803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6804 Alan Hayward <alan.hayward@arm.com>
6805 David Sherwood <david.sherwood@arm.com>
6807 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
6808 that the mode size is in the set {1, 2, 4, 8, 16}.
6810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6811 Alan Hayward <alan.hayward@arm.com>
6812 David Sherwood <david.sherwood@arm.com>
6814 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
6815 Use plus_constant instead of gen_rtx_PLUS.
6817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6818 Alan Hayward <alan.hayward@arm.com>
6819 David Sherwood <david.sherwood@arm.com>
6821 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
6822 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
6823 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
6824 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
6825 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
6826 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
6827 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
6828 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
6829 * config/i386/i386.c (ix86_push_rounding): ...this new function.
6830 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
6832 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
6833 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
6834 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
6835 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
6836 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
6837 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
6838 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
6839 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
6840 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
6841 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
6843 * expr.c (emit_move_resolve_push): Treat the input and result
6844 of PUSH_ROUNDING as a poly_int64.
6845 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
6846 (emit_push_insn): Likewise.
6847 * lra-eliminations.c (mark_not_eliminable): Likewise.
6848 * recog.c (push_operand): Likewise.
6849 * reload1.c (elimination_effects): Likewise.
6850 * rtlanal.c (nonzero_bits1): Likewise.
6851 * calls.c (store_one_arg): Likewise. Require the padding to be
6852 known at compile time.
6854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6855 Alan Hayward <alan.hayward@arm.com>
6856 David Sherwood <david.sherwood@arm.com>
6858 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
6859 Use plus_constant instead of gen_rtx_PLUS.
6861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6862 Alan Hayward <alan.hayward@arm.com>
6863 David Sherwood <david.sherwood@arm.com>
6865 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
6868 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6869 Alan Hayward <alan.hayward@arm.com>
6870 David Sherwood <david.sherwood@arm.com>
6872 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
6873 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
6874 via stack temporaries. Treat the mode size as polynomial too.
6876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6877 Alan Hayward <alan.hayward@arm.com>
6878 David Sherwood <david.sherwood@arm.com>
6880 * expr.c (expand_expr_real_2): When handling conversions involving
6881 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
6882 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
6883 as a poly_uint64 too.
6885 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6886 Alan Hayward <alan.hayward@arm.com>
6887 David Sherwood <david.sherwood@arm.com>
6889 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
6891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6892 Alan Hayward <alan.hayward@arm.com>
6893 David Sherwood <david.sherwood@arm.com>
6895 * combine.c (can_change_dest_mode): Handle polynomial
6896 REGMODE_NATURAL_SIZE.
6897 * expmed.c (store_bit_field_1): Likewise.
6898 * expr.c (store_constructor): Likewise.
6899 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
6900 and polynomial REGMODE_NATURAL_SIZE.
6901 (gen_lowpart_common): Likewise.
6902 * reginfo.c (record_subregs_of_mode): Likewise.
6903 * rtlanal.c (read_modify_subreg_p): Likewise.
6905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6906 Alan Hayward <alan.hayward@arm.com>
6907 David Sherwood <david.sherwood@arm.com>
6909 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
6910 numbers of elements.
6912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6913 Alan Hayward <alan.hayward@arm.com>
6914 David Sherwood <david.sherwood@arm.com>
6916 * match.pd: Cope with polynomial numbers of vector elements.
6918 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6919 Alan Hayward <alan.hayward@arm.com>
6920 David Sherwood <david.sherwood@arm.com>
6922 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
6923 in a POINTER_PLUS_EXPR.
6925 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6926 Alan Hayward <alan.hayward@arm.com>
6927 David Sherwood <david.sherwood@arm.com>
6929 * omp-simd-clone.c (simd_clone_subparts): New function.
6930 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
6931 (ipa_simd_modify_function_body): Likewise.
6933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6934 Alan Hayward <alan.hayward@arm.com>
6935 David Sherwood <david.sherwood@arm.com>
6937 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
6938 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
6939 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
6940 (expand_vector_condition, vector_element): Likewise.
6941 (subparts_gt): New function.
6942 (get_compute_type): Use subparts_gt.
6943 (count_type_subparts): Delete.
6944 (expand_vector_operations_1): Use subparts_gt instead of
6945 count_type_subparts.
6947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6948 Alan Hayward <alan.hayward@arm.com>
6949 David Sherwood <david.sherwood@arm.com>
6951 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
6952 (vect_compile_time_alias): ...this new function. Do the calculation
6953 on poly_ints rather than trees.
6954 (vect_prune_runtime_alias_test_list): Update call accordingly.
6956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6957 Alan Hayward <alan.hayward@arm.com>
6958 David Sherwood <david.sherwood@arm.com>
6960 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
6962 (vect_schedule_slp_instance): Likewise.
6964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6965 Alan Hayward <alan.hayward@arm.com>
6966 David Sherwood <david.sherwood@arm.com>
6968 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
6969 constant and extern definitions for variable-length vectors.
6970 (vect_get_constant_vectors): Note that the number of units
6971 is known to be constant.
6973 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6974 Alan Hayward <alan.hayward@arm.com>
6975 David Sherwood <david.sherwood@arm.com>
6977 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
6978 of units as polynomial. Choose between WIDE and NARROW based
6981 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6982 Alan Hayward <alan.hayward@arm.com>
6983 David Sherwood <david.sherwood@arm.com>
6985 * tree-vect-stmts.c (simd_clone_subparts): New function.
6986 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
6988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6989 Alan Hayward <alan.hayward@arm.com>
6990 David Sherwood <david.sherwood@arm.com>
6992 * tree-vect-stmts.c (vectorizable_call): Treat the number of
6993 vectors as polynomial. Use build_index_vector for
6996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6997 Alan Hayward <alan.hayward@arm.com>
6998 David Sherwood <david.sherwood@arm.com>
7000 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7001 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7002 for variable-length vectors.
7003 (vectorizable_mask_load_store): Treat the number of units as
7004 polynomial, asserting that it is constant if the condition has
7005 already been enforced.
7006 (vectorizable_store, vectorizable_load): Likewise.
7008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7009 Alan Hayward <alan.hayward@arm.com>
7010 David Sherwood <david.sherwood@arm.com>
7012 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7013 of units as polynomial. Punt if we can't tell at compile time
7014 which vector contains the final result.
7016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7017 Alan Hayward <alan.hayward@arm.com>
7018 David Sherwood <david.sherwood@arm.com>
7020 * tree-vect-loop.c (vectorizable_induction): Treat the number
7021 of units as polynomial. Punt on SLP inductions. Use an integer
7022 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7023 cast of such a series for variable-length floating-point
7026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7027 Alan Hayward <alan.hayward@arm.com>
7028 David Sherwood <david.sherwood@arm.com>
7030 * tree.h (build_index_vector): Declare.
7031 * tree.c (build_index_vector): New function.
7032 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7033 of units as polynomial, forcibly converting it to a constant if
7034 vectorizable_reduction has already enforced the condition.
7035 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7036 to create a {1,2,3,...} vector.
7037 (vectorizable_reduction): Treat the number of units as polynomial.
7038 Choose vectype_in based on the largest scalar element size rather
7039 than the smallest number of units. Enforce the restrictions
7042 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7043 Alan Hayward <alan.hayward@arm.com>
7044 David Sherwood <david.sherwood@arm.com>
7046 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7047 number of units as polynomial.
7049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7050 Alan Hayward <alan.hayward@arm.com>
7051 David Sherwood <david.sherwood@arm.com>
7053 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7054 * target.def (autovectorize_vector_sizes): Return the vector sizes
7055 by pointer, using vector_sizes rather than a bitmask.
7056 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7057 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7058 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7060 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7061 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7062 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7063 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7064 * omp-general.c (omp_max_vf): Likewise.
7065 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7066 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7067 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7068 * tree-vect-slp.c (vect_slp_bb): Likewise.
7069 * doc/tm.texi: Regenerate.
7070 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7072 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7073 the vector size as a poly_uint64 rather than an unsigned int.
7074 (current_vector_size): Change from an unsigned int to a poly_uint64.
7075 (get_vectype_for_scalar_type): Update accordingly.
7076 * tree.h (build_truth_vector_type): Take the size and number of
7077 units as a poly_uint64 rather than an unsigned int.
7078 (build_vector_type): Add a temporary overload that takes
7079 the number of units as a poly_uint64 rather than an unsigned int.
7080 * tree.c (make_vector_type): Likewise.
7081 (build_truth_vector_type): Take the number of units as a poly_uint64
7082 rather than an unsigned int.
7084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7085 Alan Hayward <alan.hayward@arm.com>
7086 David Sherwood <david.sherwood@arm.com>
7088 * target.def (get_mask_mode): Take the number of units and length
7089 as poly_uint64s rather than unsigned ints.
7090 * targhooks.h (default_get_mask_mode): Update accordingly.
7091 * targhooks.c (default_get_mask_mode): Likewise.
7092 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7093 * doc/tm.texi: Regenerate.
7095 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7096 Alan Hayward <alan.hayward@arm.com>
7097 David Sherwood <david.sherwood@arm.com>
7099 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7100 * omp-general.c (omp_max_vf): Likewise.
7101 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7102 (expand_omp_simd): Handle polynomial safelen.
7103 * omp-low.c (omplow_simd_context): Add a default constructor.
7104 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7105 (lower_rec_simd_input_clauses): Update accordingly.
7106 (lower_rec_input_clauses): Likewise.
7108 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7109 Alan Hayward <alan.hayward@arm.com>
7110 David Sherwood <david.sherwood@arm.com>
7112 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7113 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7114 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7115 (vect_analyze_slp_cost): Likewise.
7116 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7117 (vect_model_load_cost): Likewise.
7119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7120 Alan Hayward <alan.hayward@arm.com>
7121 David Sherwood <david.sherwood@arm.com>
7123 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7124 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7125 from an unsigned int * to a poly_uint64_pod *.
7126 (calculate_unrolling_factor): New function.
7127 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7129 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7130 Alan Hayward <alan.hayward@arm.com>
7131 David Sherwood <david.sherwood@arm.com>
7133 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7134 from an unsigned int to a poly_uint64.
7135 (_loop_vec_info::slp_unrolling_factor): Likewise.
7136 (_loop_vec_info::vectorization_factor): Change from an int
7138 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7139 (vect_get_num_vectors): New function.
7140 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7141 (vect_get_num_copies): Use vect_get_num_vectors.
7142 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7143 to an unsigned int *.
7144 (vect_analyze_data_refs): Change min_vf from an int * to a
7146 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7147 than an unsigned HOST_WIDE_INT.
7148 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7149 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7150 to an unsigned int *.
7151 (vect_analyze_data_ref_dependences): Likewise.
7152 (vect_compute_data_ref_alignment): Handle polynomial vf.
7153 (vect_enhance_data_refs_alignment): Likewise.
7154 (vect_prune_runtime_alias_test_list): Likewise.
7155 (vect_shift_permute_load_chain): Likewise.
7156 (vect_supportable_dr_alignment): Likewise.
7157 (dependence_distance_ge_vf): Take the vectorization factor as a
7158 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7159 (vect_analyze_data_refs): Change min_vf from an int * to a
7161 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7162 vfm1 as a poly_uint64 rather than an int. Make the same change
7163 for the returned bound_scalar.
7164 (vect_gen_vector_loop_niters): Handle polynomial vf.
7165 (vect_do_peeling): Likewise. Update call to
7166 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7167 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7169 * tree-vect-loop.c (vect_determine_vectorization_factor)
7170 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7171 (vect_get_known_peeling_cost): Likewise.
7172 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7173 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7174 (vect_transform_loop): Likewise. Use the lowest possible VF when
7175 updating the upper bounds of the loop.
7176 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7178 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7179 polynomial unroll factors.
7180 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7181 (vect_make_slp_decision): Likewise.
7182 (vect_supported_load_permutation_p): Likewise, and polynomial
7184 (vect_analyze_slp_cost): Handle polynomial vf.
7185 (vect_slp_analyze_node_operations): Likewise.
7186 (vect_slp_analyze_bb_1): Likewise.
7187 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7188 than an unsigned HOST_WIDE_INT.
7189 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7190 (vectorizable_load): Handle polynomial vf.
7191 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7193 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7196 Alan Hayward <alan.hayward@arm.com>
7197 David Sherwood <david.sherwood@arm.com>
7199 * match.pd: Handle bit operations involving three constants
7200 and try to fold one pair.
7202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7204 * tree-vect-loop-manip.c: Include gimple-fold.h.
7205 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7206 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7207 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7208 Add a path that uses a step of VF instead of 1, but disable it
7210 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7211 and niters_no_overflow parameters. Update calls to
7212 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7213 Create a new SSA name if the latter choses to use a ste other
7214 than zero, and return it via niters_vector_mult_vf_var.
7215 * tree-vect-loop.c (vect_transform_loop): Update calls to
7216 vect_do_peeling, vect_gen_vector_loop_niters and
7217 slpeel_make_loop_iterate_ntimes.
7218 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7219 (vect_gen_vector_loop_niters): Update declarations after above changes.
7221 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7223 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7224 128-bit round to integer instructions.
7225 (ceil<mode>2): Likewise.
7226 (btrunc<mode>2): Likewise.
7227 (round<mode>2): Likewise.
7229 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7231 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7232 unaligned VSX load/store on P8/P9.
7233 (expand_block_clear): Allow the use of unaligned VSX
7234 load/store on P8/P9.
7236 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7238 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7240 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7241 swap associated with both a load and a store.
7243 2018-01-02 Andrew Waterman <andrew@sifive.com>
7245 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7246 * config/riscv/riscv.md (clear_cache): Use it.
7248 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7250 * web.c: Remove out-of-date comment.
7252 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7254 * expr.c (fixup_args_size_notes): Check that any existing
7255 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7256 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7257 (emit_single_push_insn): ...here.
7259 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7261 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7262 (const_vector_encoded_nelts): New function.
7263 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7264 (const_vector_int_elt, const_vector_elt): Declare.
7265 * emit-rtl.c (const_vector_int_elt_1): New function.
7266 (const_vector_elt): Likewise.
7267 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7268 of CONST_VECTOR_ELT.
7270 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7272 * expr.c: Include rtx-vector-builder.h.
7273 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7274 directly on the tree encoding.
7275 (const_vector_from_tree): Likewise.
7276 * optabs.c: Include rtx-vector-builder.h.
7277 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7278 sequence of "u" values.
7279 * vec-perm-indices.c: Include rtx-vector-builder.h.
7280 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7281 directly on the vec_perm_indices encoding.
7283 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7285 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7286 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7287 * rtx-vector-builder.h: New file.
7288 * rtx-vector-builder.c: Likewise.
7289 * rtl.h (rtx_def::u2): Add a const_vector field.
7290 (CONST_VECTOR_NPATTERNS): New macro.
7291 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7292 (CONST_VECTOR_DUPLICATE_P): Likewise.
7293 (CONST_VECTOR_STEPPED_P): Likewise.
7294 (CONST_VECTOR_ENCODED_ELT): Likewise.
7295 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7296 (unwrap_const_vec_duplicate): Likewise.
7297 (const_vec_series_p): Check for a non-duplicated vector encoding.
7298 Say that the function only returns true for integer vectors.
7299 * emit-rtl.c: Include rtx-vector-builder.h.
7300 (gen_const_vec_duplicate_1): Delete.
7301 (gen_const_vector): Call gen_const_vec_duplicate instead of
7302 gen_const_vec_duplicate_1.
7303 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7304 (gen_const_vec_duplicate): Use rtx_vector_builder.
7305 (gen_const_vec_series): Likewise.
7306 (gen_rtx_CONST_VECTOR): Likewise.
7307 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7308 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7309 Build a new vector rather than modifying a CONST_VECTOR in-place.
7310 (handle_special_swappables): Update call accordingly.
7311 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7312 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7313 Build a new vector rather than modifying a CONST_VECTOR in-place.
7314 (handle_special_swappables): Update call accordingly.
7316 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7318 * simplify-rtx.c (simplify_const_binary_operation): Use
7319 CONST_VECTOR_ELT instead of XVECEXP.
7321 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7323 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7324 the selector elements to be different from the data elements
7325 if the selector is a VECTOR_CST.
7326 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7327 ssizetype for the selector.
7329 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7331 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7332 before testing each element individually.
7333 * tree-vect-generic.c (lower_vec_perm): Likewise.
7335 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7337 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7338 * selftest-run-tests.c (selftest::run_tests): Call it.
7339 * vector-builder.h (vector_builder::operator ==): New function.
7340 (vector_builder::operator !=): Likewise.
7341 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7342 (vec_perm_indices::all_from_input_p): New function.
7343 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7344 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7345 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7346 instead of reading the VECTOR_CST directly. Detect whether both
7347 vector inputs are the same before constructing the vec_perm_indices,
7348 and update the number of inputs argument accordingly. Use the
7349 utility functions added above. Only construct sel2 if we need to.
7351 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7353 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7354 the broadcast of the low byte.
7355 (expand_mult_highpart): Use an explicit encoding for the permutes.
7356 * optabs-query.c (can_mult_highpart_p): Likewise.
7357 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7358 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7359 (vectorizable_bswap): Likewise.
7360 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7361 explicit encoding for the power-of-2 permutes.
7362 (vect_permute_store_chain): Likewise.
7363 (vect_grouped_load_supported): Likewise.
7364 (vect_permute_load_chain): Likewise.
7366 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7368 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7369 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7370 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7371 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7372 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7373 (vect_gen_perm_mask_any): Likewise.
7375 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7377 * int-vector-builder.h: New file.
7378 * vec-perm-indices.h: Include int-vector-builder.h.
7379 (vec_perm_indices): Redefine as an int_vector_builder.
7380 (auto_vec_perm_indices): Delete.
7381 (vec_perm_builder): Redefine as a stand-alone class.
7382 (vec_perm_indices::vec_perm_indices): New function.
7383 (vec_perm_indices::clamp): Likewise.
7384 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7385 (vec_perm_indices::new_vector): New function.
7386 (vec_perm_indices::new_expanded_vector): Update for new
7387 vec_perm_indices class.
7388 (vec_perm_indices::rotate_inputs): New function.
7389 (vec_perm_indices::all_in_range_p): Operate directly on the
7390 encoded form, without computing elided elements.
7391 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7392 encoding. Update for new vec_perm_indices class.
7393 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7394 the given vec_perm_builder.
7395 (expand_vec_perm_var): Update vec_perm_builder constructor.
7396 (expand_mult_highpart): Use vec_perm_builder instead of
7397 auto_vec_perm_indices.
7398 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7399 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7400 or double series encoding as appropriate.
7401 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7402 vec_perm_indices instead of auto_vec_perm_indices.
7403 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7404 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7405 (vect_permute_store_chain): Likewise.
7406 (vect_grouped_load_supported): Likewise.
7407 (vect_permute_load_chain): Likewise.
7408 (vect_shift_permute_load_chain): Likewise.
7409 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7410 (vect_transform_slp_perm_load): Likewise.
7411 (vect_schedule_slp_instance): Likewise.
7412 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7413 (vectorizable_mask_load_store): Likewise.
7414 (vectorizable_bswap): Likewise.
7415 (vectorizable_store): Likewise.
7416 (vectorizable_load): Likewise.
7417 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7418 vec_perm_indices instead of auto_vec_perm_indices. Use
7419 tree_to_vec_perm_builder to read the vector from a tree.
7420 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7421 vec_perm_builder instead of a vec_perm_indices.
7422 (have_whole_vector_shift): Use vec_perm_builder and
7423 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7424 truncation to calc_vec_perm_mask_for_shift.
7425 (vect_create_epilog_for_reduction): Likewise.
7426 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7427 from auto_vec_perm_indices to vec_perm_indices.
7428 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7429 instead of changing individual elements.
7430 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7431 the vector in d.perm.
7432 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7433 from auto_vec_perm_indices to vec_perm_indices.
7434 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7435 instead of changing individual elements.
7436 (arm_vectorize_vec_perm_const): Use new_vector to install
7437 the vector in d.perm.
7438 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7439 Update vec_perm_builder constructor.
7440 (rs6000_expand_interleave): Likewise.
7441 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7442 (rs6000_expand_interleave): Likewise.
7444 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7446 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7447 to qimode could truncate the indices.
7448 * optabs.c (expand_vec_perm_var): Likewise.
7450 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7452 * Makefile.in (OBJS): Add vec-perm-indices.o.
7453 * vec-perm-indices.h: New file.
7454 * vec-perm-indices.c: Likewise.
7455 * target.h (vec_perm_indices): Replace with a forward class
7457 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7458 * optabs.h: Include vec-perm-indices.h.
7459 (expand_vec_perm): Delete.
7460 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7461 (expand_vec_perm_const): Declare.
7462 * target.def (vec_perm_const_ok): Replace with...
7463 (vec_perm_const): ...this new hook.
7464 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7465 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7466 * doc/tm.texi: Regenerate.
7467 * optabs.def (vec_perm_const): Delete.
7468 * doc/md.texi (vec_perm_const): Likewise.
7469 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7470 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7471 expand_vec_perm for constant permutation vectors. Assert that
7472 the mode of variable permutation vectors is the integer equivalent
7473 of the mode that is being permuted.
7474 * optabs-query.h (selector_fits_mode_p): Declare.
7475 * optabs-query.c: Include vec-perm-indices.h.
7476 (selector_fits_mode_p): New function.
7477 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7478 is defined, instead of checking whether the vec_perm_const_optab
7479 exists. Use targetm.vectorize.vec_perm_const instead of
7480 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7481 fit in the vector mode before using a variable permute.
7482 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7483 vec_perm_indices instead of an rtx.
7484 (expand_vec_perm): Replace with...
7485 (expand_vec_perm_const): ...this new function. Take the selector
7486 as a vec_perm_indices rather than an rtx. Also take the mode of
7487 the selector. Update call to shift_amt_for_vec_perm_mask.
7488 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7489 Use vec_perm_indices::new_expanded_vector to expand the original
7490 selector into bytes. Check whether the indices fit in the vector
7491 mode before using a variable permute.
7492 (expand_vec_perm_var): Make global.
7493 (expand_mult_highpart): Use expand_vec_perm_const.
7494 * fold-const.c: Includes vec-perm-indices.h.
7495 * tree-ssa-forwprop.c: Likewise.
7496 * tree-vect-data-refs.c: Likewise.
7497 * tree-vect-generic.c: Likewise.
7498 * tree-vect-loop.c: Likewise.
7499 * tree-vect-slp.c: Likewise.
7500 * tree-vect-stmts.c: Likewise.
7501 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
7503 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
7504 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
7505 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
7506 (aarch64_vectorize_vec_perm_const): ...this new function.
7507 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7508 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7509 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
7510 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
7511 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7512 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7513 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
7515 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
7516 check for NEON modes.
7517 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
7518 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
7519 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
7520 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
7522 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
7523 the old VEC_PERM_CONST conditions.
7524 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
7525 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
7526 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
7527 (ia64_vectorize_vec_perm_const_ok): Merge into...
7528 (ia64_vectorize_vec_perm_const): ...this new function.
7529 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
7530 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
7531 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
7532 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
7533 * config/mips/mips.c (mips_expand_vec_perm_const)
7534 (mips_vectorize_vec_perm_const_ok): Merge into...
7535 (mips_vectorize_vec_perm_const): ...this new function.
7536 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
7537 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
7538 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
7539 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
7540 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
7541 (rs6000_expand_vec_perm_const): Delete.
7542 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
7544 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7545 (altivec_expand_vec_perm_const_le): Take each operand individually.
7546 Operate on constant selectors rather than rtxes.
7547 (altivec_expand_vec_perm_const): Likewise. Update call to
7548 altivec_expand_vec_perm_const_le.
7549 (rs6000_expand_vec_perm_const): Delete.
7550 (rs6000_vectorize_vec_perm_const_ok): Delete.
7551 (rs6000_vectorize_vec_perm_const): New function.
7552 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7553 an element count and rtx array.
7554 (rs6000_expand_extract_even): Update call accordingly.
7555 (rs6000_expand_interleave): Likewise.
7556 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
7557 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
7558 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
7559 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
7560 (rs6000_expand_vec_perm_const): Delete.
7561 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7562 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7563 (altivec_expand_vec_perm_const_le): Take each operand individually.
7564 Operate on constant selectors rather than rtxes.
7565 (altivec_expand_vec_perm_const): Likewise. Update call to
7566 altivec_expand_vec_perm_const_le.
7567 (rs6000_expand_vec_perm_const): Delete.
7568 (rs6000_vectorize_vec_perm_const_ok): Delete.
7569 (rs6000_vectorize_vec_perm_const): New function. Remove stray
7570 reference to the SPE evmerge intructions.
7571 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7572 an element count and rtx array.
7573 (rs6000_expand_extract_even): Update call accordingly.
7574 (rs6000_expand_interleave): Likewise.
7575 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
7576 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
7578 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7580 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7582 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
7583 vector mode and that that mode matches the mode of the data
7585 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
7586 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
7587 directly using expand_vec_perm_1 when forcing selectors into
7589 (expand_vec_perm_var): New function, split out from expand_vec_perm.
7591 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7593 * optabs-query.h (can_vec_perm_p): Delete.
7594 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
7595 * optabs-query.c (can_vec_perm_p): Split into...
7596 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
7597 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
7598 particular selector is valid.
7599 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7600 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7601 (vect_grouped_load_supported): Likewise.
7602 (vect_shift_permute_load_chain): Likewise.
7603 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7604 (vect_transform_slp_perm_load): Likewise.
7605 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7606 (vectorizable_bswap): Likewise.
7607 (vect_gen_perm_mask_checked): Likewise.
7608 * fold-const.c (fold_ternary_loc): Likewise. Don't take
7609 implementations of variable permutation vectors into account
7610 when deciding which selector to use.
7611 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
7612 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
7613 with a false third argument.
7614 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
7615 to test whether the constant selector is valid and can_vec_perm_var_p
7616 to test whether a variable selector is valid.
7618 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7620 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
7621 * optabs-query.c (can_vec_perm_p): Likewise.
7622 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
7623 instead of vec_perm_indices.
7624 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
7625 (vect_gen_perm_mask_checked): Likewise,
7626 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
7627 (vect_gen_perm_mask_checked): Likewise,
7629 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7631 * optabs-query.h (qimode_for_vec_perm): Declare.
7632 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
7633 (qimode_for_vec_perm): ...this new function.
7634 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
7636 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7638 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
7639 does not have a conditional at the top.
7641 2018-01-02 Richard Biener <rguenther@suse.de>
7643 * ipa-inline.c (big_speedup_p): Fix expression.
7645 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7648 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
7651 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7655 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
7656 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
7657 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
7658 cond_taken_branch_cost 3->4.
7660 2018-01-01 Jakub Jelinek <jakub@redhat.com>
7662 PR tree-optimization/83581
7663 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
7664 TODO_cleanup_cfg if any changes have been made.
7667 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
7668 convert_modes if target mode has the right side, but different mode
7672 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
7673 last argument when extracting from CONCAT. If either from_real or
7674 from_imag is NULL, use expansion through memory. If result is not
7675 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
7676 the parts directly to inner mode, if even that fails, use expansion
7680 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
7681 check for bswap in mode rather than HImode and use that in expand_unop
7684 Copyright (C) 2018 Free Software Foundation, Inc.
7686 Copying and distribution of this file, with or without modification,
7687 are permitted in any medium without royalty provided the copyright
7688 notice and this notice are preserved.