1 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
4 p->max as pointers rather than using iterative_hash_expr.
6 2018-02-23 Carl Love <cel@us.ibm.com>
8 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
11 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
12 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
13 P8V_BUILTIN_VEC_VUNSIGNED2.
15 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
18 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
19 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
21 * lra-constraints.c (curr_insn_transform): Set up
22 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
24 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
25 * lra-eliminations.c (spill_pseudos): Ditto.
26 (process_insn_for_elimination): Ditto.
27 * lra-lives.c (reg_early_clobber_p): Use the new macros.
28 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
29 LRA_NON_CLOBBERED_ALT.
31 2018-02-22 Martin Sebor <msebor@redhat.com>
33 PR tree-optimization/84480
34 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
35 to maybe_diag_stxncpy_trunc. Call it.
36 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
37 from gimple_fold_builtin_strcpy. Print inlining stack.
38 (handle_builtin_stxncpy): Print inlining stack.
39 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
41 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
44 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
45 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
46 and -fcheck-pointer-bounds are used together.
47 (indirect_thunk_prefix): New enum.
48 (indirect_thunk_need_prefix): New function.
49 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
50 "_nt" instead of "_bnd" for NOTRACK prefix.
51 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
52 (output_indirect_thunk_function): Likewise.
54 (ix86_code_end): Update output_indirect_thunk_function calls.
55 (ix86_output_indirect_branch_via_reg): Replace
56 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
57 (ix86_output_indirect_branch_via_push): Likewise.
58 (ix86_output_function_return): Likewise.
59 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
60 incompatible with -fcf-protection=branch and
61 -fcheck-pointer-bounds.
63 2018-02-22 Steve Ellcey <sellcey@cavium.com>
66 * config/aarch64/aarch64.c (aarch64_print_address_internal):
67 Change gcc_assert call to output_operand_lossage.
69 2018-02-22 Steve Ellcey <sellcey@cavium.com>
71 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
73 2018-02-22 DJ Delorie <dj@redhat.com>
74 Sebastian Perta <sebastian.perta@renesas.com>
75 Oleg Endo <olegendo@gcc.gnu.org>
77 * config/rx/rx.c (rx_rtx_costs): New function.
78 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
80 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
82 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
84 2018-02-22 Martin Liska <mliska@suse.cz>
87 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
88 Add "native" as a possible value.
90 2018-02-22 Martin Liska <mliska@suse.cz>
93 * config/i386/i386.c (ix86_option_override_internal):
94 Add "native" as a possible value for -march and -mtune.
96 2018-02-22 Jakub Jelinek <jakub@redhat.com>
99 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
100 to all type variants.
102 PR tree-optimization/84503
103 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
104 width as info->bitpos + info->bitsize - start.
105 (merged_store_group::merge_overlapping): Simplify width computation.
106 (check_no_overlap): New function.
107 (imm_store_chain_info::try_coalesce_bswap): Compute expected
108 start + width and last_order of the group, fail if check_no_overlap
110 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
111 to group if check_no_overlap fails.
113 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
115 * config/rs6000/altivec.md: Delete contraint arguments to
116 define_expand, define_split, and define_peephole2, and in
117 define_insn_and_split if always unused.
118 * config/rs6000/darwin.md: Ditto.
119 * config/rs6000/dfp.md: Ditto.
120 * config/rs6000/rs6000.md: Ditto.
121 * config/rs6000/sync.md: Ditto.
122 * config/rs6000/vector.md: Ditto.
123 * config/rs6000/vsx.md: Ditto.
125 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
127 * config/rs6000/altivec.md: Write output control strings as braced
128 blocks instead of double-quoted strings.
129 * config/rs6000/darwin.md: Ditto.
130 * config/rs6000/rs6000.md: Ditto.
131 * config/rs6000/vector.md: Ditto.
132 * config/rs6000/vsx.md: Ditto.
134 2018-02-21 Jason Merrill <jason@redhat.com>
136 PR c++/84314 - ICE with templates and fastcall attribute.
137 * attribs.c (build_type_attribute_qual_variant): Remove assert.
139 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
141 * ipa-cp.c (determine_versionability): Fix comment typos.
143 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
146 * ipa-cp.c (determine_versionability): Do not version functions caling
149 2018-02-21 Martin Liska <mliska@suse.cz>
152 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
153 Add "native" as a possible value.
154 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
155 the macro when native cpu detection is available.
157 2018-02-21 Martin Liska <mliska@suse.cz>
160 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
161 Add "native" as a possible value.
162 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
163 when native cpu detection is available.
165 2018-02-21 Jakub Jelinek <jakub@redhat.com>
166 Martin Sebor <msebor@redhat.com>
168 PR tree-optimization/84478
169 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
171 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
172 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
173 support which is conservatively correct, for 2 only stay conservative
174 for maxlen. Formatting and comment capitalization fixes. Add STRICT
175 argument to the 2 argument get_range_strlen, adjust 6 arg
176 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
178 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
179 (gimple_fold_builtin_strlen): Pass true as last argument to
182 2018-02-20 Martin Sebor <msebor@redhat.com>
185 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
186 (builtin_memref::set_base_and_offset): Same. Handle inner references.
187 (builtin_memref::builtin_memref): Factor out parts into
188 set_base_and_offset and call it.
190 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
193 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
194 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
195 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
196 search at the associated MODE_INT.
198 2018-02-20 Jeff Law <law@redhat.com>
201 PR tree-optimization/81592
203 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
204 for range data rather than using global data.
205 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
206 range data rather than using global data.
207 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
208 pass it to children as needed.
209 (struct directive::fmtresult): Similarly.
210 (struct directive::set_width): Similarly.
211 (struct directive::set_precision): Similarly.
212 (format_integer, format_directive, parse_directive): Similarly.
213 (format_none): Accept unnamed vr_values parameter.
214 (format_percent, format_floating, format_character): Similarly.
215 (format_string, format_plain): Similarly.
216 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
217 the EVRP range analyzer for range data rather than using global data.
218 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
219 gimple-ssa-evrp-analyze.h
220 (class sprintf_dom_walker): Add after_dom_children member function.
221 Add evrp_range_analyzer member.
222 (sprintf_dom_walker::before_dom_children): Call into the EVRP
223 range analyzer as needed.
224 (sprintf_dom_walker::after_dom_children): New member function.
225 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
227 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
228 (evrp_range_analyzer::pop_to_marker): Likewise.
230 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
232 PR tree-optimization/84419
233 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
234 with the required type if its current type is compatible but
237 2018-02-20 Jakub Jelinek <jakub@redhat.com>
240 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
243 2018-02-20 Martin Liska <mliska@suse.cz>
246 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
247 possible values if we don't have a hint.
249 2018-02-20 Martin Liska <mliska@suse.cz>
253 * final.c (shorten_branches): Build align_tab array with one
255 * opts.c (finish_options): Add alignment option limit check.
256 (MAX_CODE_ALIGN): Likewise.
257 (MAX_CODE_ALIGN_VALUE): Likewise.
258 * doc/invoke.texi: Document maximum allowed option value for
259 all -falign-* options.
261 2018-02-19 Jakub Jelinek <jakub@redhat.com>
264 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
265 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
266 * var-tracking.c (emit_note_insn_var_location): Remove all references
267 to NOTE_INSN_CALL_ARG_LOCATION.
268 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
269 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
270 Use copy_rtx_if_shared.
271 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
272 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
273 (dwarf2out_var_location): Remove handling of
274 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
276 * final.c (final_scan_insn): Remove all references to
277 NOTE_INSN_CALL_ARG_LOCATION.
278 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
279 before dumping final insns.
280 * except.c (emit_note_eh_region_end): Remove all references to
281 NOTE_INSN_CALL_ARG_LOCATION.
282 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
283 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
284 * config/arc/arc.c (hwloop_optimize): Likewise.
285 * config/arm/arm.c (create_fix_barrier): Likewise.
286 * config/s390/s390.c (s390_chunkify_start): Likewise.
287 * config/sh/sh.c (find_barrier): Likewise.
288 * config/i386/i386.c (rest_of_insert_endbranch,
289 ix86_seh_fixup_eh_fallthru): Likewise.
290 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
291 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
292 * config/frv/frv.c (frv_function_prologue): Likewise.
293 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
295 (note_outside_basic_block_p): Remove all references to
296 NOTE_INSN_CALL_ARG_LOCATION.
297 * gengtype.c (adjust_field_rtx_def): Likewise.
298 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
300 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
301 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
304 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
307 PR tree-optimization/84452
308 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
309 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
312 2018-02-19 Martin Liska <mliska@suse.cz>
315 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
317 2018-02-19 Martin Liska <mliska@suse.cz>
318 Richard Sandiford <richard.sandiford@linaro.org>
320 PR tree-optimization/82491
321 * gimple-fold.c (get_base_constructor): Make earlier bail out
324 2018-02-19 Carl Love <cel@us.ibm.com>
326 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
327 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
329 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
332 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
334 * config/rl78/rl78.md (movdf): New define expand.
336 2018-02-19 Martin Liska <mliska@suse.cz>
339 * doc/invoke.texi: Fix typo.
340 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
342 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
344 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
345 handle rs6000_single_float and rs6000_double_float specially for
348 2018-02-16 Jeff Law <law@redhat.com>
350 * config/rx/rx.c (add_pop_cfi_notes): New function.;
353 2018-02-16 Jakub Jelinek <jakub@redhat.com>
356 * ipa-inline.c (inline_small_functions): Fix a typo.
358 2018-02-16 Nathan Sidwell <nathan@acm.org>
360 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
362 2018-02-16 Carl Love <cel@us.ibm.com>
364 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
365 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
366 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
367 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
368 expansion to P8V_BUILTIN_VEC_FLOAT2.
370 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
372 PR rtl-optimization/70023
373 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
374 src_regno into account.
376 2018-02-16 Carl Love <cel@us.ibm.com>
378 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
379 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
380 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
381 * config/rs6000/rs6000.c: Remove case statements for
382 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
383 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
384 and P9V_BUILTIN_VEC_VINSERT4B.
385 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
386 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
387 * config/rs6000/vsx.md:
388 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
391 2018-02-16 Carl Love <cel@us.ibm.com>
393 * config/rs6000/altivec.h: Add builtin names vec_extract4b
395 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
397 * config/rs6000/rs6000-c.c: Add the definitions for
398 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
399 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
400 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
401 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
402 definition for insert4b and define insn *insert3b_internal.
403 * doc/extend.texi: Add documentation for vec_extract4b.
405 2018-02-16 Nathan Sidwell <nathan@acm.org>
407 * doc/extend.texi (Backwards Compatibility): Mention friend
408 injection. Note for-scope is deprecated.
409 * doc/invoke.texi (-ffriend-injection): Deprecate.
411 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
413 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
414 that moved to I2, also allow destinations that are a paradoxical
415 subreg (instead of a normal reg).
417 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
420 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
423 2018-02-16 Richard Biener <rguenther@suse.de>
425 PR tree-optimization/84037
426 PR tree-optimization/84016
428 * config/i386/i386.c (ix86_builtin_vectorization_cost):
429 Adjust vec_construct for the fact we need additional higher latency
430 128bit inserts for AVX256 and AVX512 vector builds.
431 (ix86_add_stmt_cost): Scale vector construction cost for
434 2018-02-16 Richard Biener <rguenther@suse.de>
436 PR tree-optimization/84417
437 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
438 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
439 (non_rewritable_lvalue_p): Likewise, use poly-ints.
441 2018-02-16 Martin Liska <mliska@suse.cz>
444 * internal-fn.def (ASAN_CHECK): Set proper flags.
445 (ASAN_MARK): Likewise.
447 2018-02-16 Julia Koval <julia.koval@intel.com>
449 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
452 2018-02-16 Jakub Jelinek <jakub@redhat.com>
455 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
456 Use ++iter rather than iter++ for std::list iterators.
457 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
458 defer deleting them until all nodes in the forest are processed. Do
459 free even leaf nodes. Change to_process into auto_vec.
462 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
463 * vec.h (vec_default_construct): Use memset instead of placement new
464 if BROKEN_VALUE_INITIALIZATION is defined.
465 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
466 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
469 PR rtl-optimization/83723
470 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
471 * lra.c (lra_substitute_pseudo): Likewise. If true, use
472 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
474 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
476 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
478 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
480 PR rtl-optimization/81443
481 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
482 from inner REGs to paradoxical SUBREGs.
484 2018-02-16 Richard Biener <rguenther@suse.de>
486 PR tree-optimization/84399
487 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
488 For operands we can analyze at their definition make sure we can
489 analyze them at each use as well.
491 2018-02-16 Richard Biener <rguenther@suse.de>
493 PR tree-optimization/84190
494 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
495 volatile accesses if the decl isn't volatile.
497 2018-02-15 Jason Merrill <jason@redhat.com>
499 PR c++/84314 - ICE with templates and fastcall attribute.
500 * attribs.c (build_type_attribute_qual_variant): Don't clobber
501 TYPE_CANONICAL on an existing type.
503 2018-02-15 Jakub Jelinek <jakub@redhat.com>
505 PR tree-optimization/84383
506 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
507 dstoff nor call operand_equal_p if dstbase is NULL.
509 PR tree-optimization/84334
510 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
511 also a CONSTANT_CLASS_P, punt.
513 2018-02-14 Jim Wilson <jimw@sifive.com>
515 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
516 first SMALL_OPERAND check. New local min_second_step. Move assert
517 to where locals are set. Add TARGET_RVC support.
518 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
520 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
522 * doc/invoke.texi: Correct -Wformat-overflow code sample.
524 2018-02-14 Martin Sebor <msebor@redhat.com>
526 PR tree-optimization/83698
527 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
528 arrays constrain the offset range to their bounds.
529 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
530 (builtin_access::overlap): Avoid setting the size of overlap if it's
532 (maybe_diag_overlap): Also consider arrays when deciding what values
533 of offsets to include in diagnostics.
535 2018-02-14 Martin Sebor <msebor@redhat.com>
538 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
539 that correspond to the kind of a declaration.
541 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
544 * config/pa/pa.md: Load address of PIC label using the linkage table
545 if the label is nonlocal.
547 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
549 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
550 warning message if user requests -maltivec=be.
551 * doc/invoke.texi: Document deprecation of -maltivec=be.
553 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
556 * config/rs6000/rs6000-c.c: Update definitions for
557 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
558 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
560 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
563 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
564 add _get_ssp intrinsics. Remove argument from
565 __builtin_ia32_rdssp[d|q].
566 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
567 * config/i386/i386-builtin.def: Remove argument from
568 __builtin_ia32_rdssp[d|q].
569 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
570 ix86_expand_special_args_builtin for _rdssp[d|q].
571 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
572 Clear register before usage.
573 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
574 Add documentation for new _get_ssp and _inc_ssp intrinsics.
576 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
578 PR tree-optimization/84357
579 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
580 operand 1 of an ARRAY_REF too.
582 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
585 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
586 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
588 (set_of_reg): New struct.
589 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
590 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
591 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
593 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
594 Split into bitclr, bitset, bitinvert patterns if appropriate.
595 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
596 use rx_fuse_in_memory_bitop.
597 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
598 to named insn, correct maximum insn length.
600 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
603 * machmode.def: Define a complex mode for PARTIAL_INT.
604 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
606 * doc/rtl.texi: Document CSPImode.
607 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
609 (msp430_hard_regno_nregs_with_padding): Likewise.
611 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
614 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
616 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
618 PR rtl-optimization/84169
619 * combine.c (try_combine): New variable split_i2i3. Set it to true if
620 we generated a parallel as new i3 and we split that to new i2 and i3
621 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
622 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
623 those to i2, not i1. Partially rewrite this scan code.
625 2018-02-13 Jakub Jelinek <jakub@redhat.com>
628 * stor-layout.c (place_field): For variable length fields, adjust
629 offset_align afterwards not just based on the field's alignment,
630 but also on the size.
633 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
634 of exps and logs in the use_exp2 case.
636 2018-02-13 Jeff Law <law@redhat.com>
638 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
641 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
644 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
648 * common.opt (gas-loc-support, gas-locview-support): New.
649 (ginline-points, ginternal-reset-location-views): New.
650 * doc/invoke.texi: Document them. Use @itemx where intended.
651 (gvariable-location-views): Adjust.
652 * target.def (reset_location_view): New.
653 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
654 (TARGET_RESET_LOCATION_VIEW): New.
655 * doc/tm.texi: Rebuilt.
656 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
657 (dwarf2out_default_as_locview_support): New.
658 (output_asm_line_debug_info): Use option variables.
659 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
660 (output_loc_list): Likewise.
661 (add_high_low_attributes): Check option variables.
662 Don't output entry view attribute in strict mode.
663 (gen_inlined_subroutine_die): Check option variables.
664 (dwarf2out_inline_entry): Likewise.
665 (init_sections_and_labels): Likewise.
666 (dwarf2out_early_finish): Likewise.
667 (maybe_reset_location_view): New, from...
668 (dwarf2out_var_location): ... here. Call it.
669 * debug.h (dwarf2out_default_as_loc_support): Declare.
670 (dwarf2out_default_as_locview_support): Declare.
671 * hooks.c (hook_int_rtx_insn_0): New.
672 * hooks.h (hook_int_rtx_insn_0): Declare.
673 * toplev.c (process_options): Take -gas-loc-support and
674 -gas-locview-support from dwarf2out. Enable
675 -gvariable-location-views by default only with locview
676 assembler support. Enable -ginternal-reset-location-views by
677 default only if the target defines the corresponding hook.
678 Enable -ginline-points by default if location views are
679 enabled; force it disabled if statement frontiers are
681 * tree-inline.c (expand_call_inline): Check option variables.
682 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
684 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
686 PR tree-optimization/84321
687 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
688 handling. Also check whether the anti-range contains any values
689 that satisfy the mask; switch to a VR_RANGE if not.
691 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
694 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
696 2018-02-13 Martin Jambor <mjambor@suse.cz>
699 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
700 of call statements, also set location of a load to a temporary.
702 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
704 * config/rl78/rl78.c (add_vector_labels): New function.
705 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
706 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
707 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
708 which checks that no arguments are passed.
709 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
710 * doc/extend.texi: Documentation for the new attribute.
712 2018-02-13 Andreas Schwab <schwab@suse.de>
714 * config/riscv/linux.h (CPP_SPEC): Define.
716 2018-02-13 Jakub Jelinek <jakub@redhat.com>
719 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
720 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
721 OPTION_MASK_ISA_AES as first argument to def_builtin_const
722 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
723 instead of OPTION_MASK_ISA_PCLMUL as first argument to
724 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
725 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
726 temporarily for AES and PCLMUL builtins.
728 PR tree-optimization/84339
729 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
730 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
734 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
735 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
736 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
738 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
740 * omp-simd-clone.h: New file.
741 * omp-simd-clone.c: Include omp-simd-clone.h.
742 (expand_simd_clones): No longer static.
743 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
744 cgraph.h and omp-simd-clone.h.
745 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
746 (vect_recog_widen_shift_pattern): Formatting fix.
747 (vect_pattern_recog_1): Don't check optab for calls.
750 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
751 operands[2] into a REG before using gen_lowpart on it.
753 2018-02-12 Jeff Law <law@redhat.com>
756 * config/sh/sh.c (find_barrier): Consider a sibling call
759 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
760 successfully back substituting a reg.
762 2018-02-12 Richard Biener <rguenther@suse.de>
764 PR tree-optimization/84037
765 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
766 parameter, move visited init to caller.
767 (vect_slp_analyze_operations): Separate cost from validity
768 check, initialize visited once for all instances.
769 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
771 * tree-vect-stmts.c (vect_model_simple_cost): Make early
773 (vect_model_promotion_demotion_cost): Likewise.
774 (vectorizable_bswap): Guard cost modeling with !slp_node
775 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
777 (vectorizable_call): Likewise.
778 (vectorizable_conversion): Likewise.
779 (vectorizable_assignment): Likewise.
780 (vectorizable_shift): Likewise.
781 (vectorizable_operation): Likewise.
782 (vectorizable_store): Likewise.
783 (vectorizable_load): Likewise.
784 (vectorizable_condition): Likewise.
785 (vectorizable_comparison): Likewise.
787 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
790 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
791 (ASAN_MARK): Fix fnspec to account for return value, change pointer
792 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
794 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
797 * params.def (inline-min-speedup): Increase from 8 to 15.
798 (max-inline-insns-auto): Decrease from 40 to 30.
799 * ipa-split.c (consider_split): Add some buffer for function to
800 be considered inlining candidate.
801 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
804 2018-02-12 Richard Biener <rguenther@suse.de>
806 PR tree-optimization/84037
807 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
808 matched stmts if we cannot swap the non-matched ones.
810 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
812 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
813 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
814 _mm_maskz_scalef_round_ss): New intrinsics.
815 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
816 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
817 __builtin_ia32_scalefss_round): Remove.
818 (__builtin_ia32_scalefsd_mask_round,
819 __builtin_ia32_scalefss_mask_round): New intrinsics.
820 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
821 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
822 ((match_operand:VF_128 2 "<round_nimm_predicate>"
823 "<round_constraint>")): Changed to ...
824 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
825 "<round_scalar_constraint>")): ... this.
826 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
827 %0, %1, %2<round_op3>}"): Changed to ...
828 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
829 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
830 %2<round_scalar_mask_op3>}"): ... this.
831 * config/i386/subst.md (round_scalar_nimm_predicate): New.
833 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
835 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
836 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
837 (_mm_maskz_sqrt_round_ss): New intrinsics.
838 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
839 (__builtin_ia32_sqrtsd_mask_round)
840 (__builtin_ia32_sqrtss_mask_round): New builtins.
841 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
842 (__builtin_ia32_sqrtss_round): Remove.
843 (__builtin_ia32_sqrtsd_mask_round)
844 (__builtin_ia32_sqrtss_mask_round): New builtins.
845 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
846 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
847 ((match_operand:VF_128 1 "vector_operand"
848 "xBm,<round_constraint>")): Changed to ...
849 ((match_operand:VF_128 1 "vector_operand"
850 "xBm,<round_scalar_constraint>")): ... this.
851 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
852 %0, %2, %<iptr>1<round_op3>}): Changed to ...
853 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
854 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
855 %<iptr>1<round_scalar_mask_op3>}): ... this.
856 ((set_attr "prefix" "<round_prefix>")): Changed to ...
857 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
859 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
862 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
863 Cast vec_cmpeq result to correct type.
864 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
865 Cast vec_cmpgt result to correct type.
867 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
869 * final.c (final_scan_insn_1): Renamed from...
870 (final_scan_insn): ... this. New wrapper, to recover
871 seen from the outermost call in recursive ones.
872 * config/sparc/sparc.c (output_return): Drop seen from call.
873 (output_sibcall): Likewise.
874 * config/visium/visium.c (output_branch): Likewise.
876 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
878 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
881 2018-02-10 Alan Modra <amodra@gmail.com>
884 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
885 Specify LR as an input.
887 2018-02-10 Jakub Jelinek <jakub@redhat.com>
890 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
891 remove_member_access_dummy_vars): New functions.
892 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
893 lower_omp_1, execute_lower_omp): Use them.
895 PR rtl-optimization/84308
896 * shrink-wrap.c (spread_components): Release todo vector.
898 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
900 PR rtl-optimization/57193
901 * ira-color.c (struct allocno_color_data): Add member
902 conflict_allocno_hard_prefs.
903 (update_conflict_allocno_hard_prefs): New.
904 (bucket_allocno_compare_func): Add a preference based on
905 conflict_allocno_hard_prefs.
906 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
907 (color_allocnos): Remove a dead code. Initiate
908 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
910 2018-02-09 Jakub Jelinek <jakub@redhat.com>
913 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
914 constraint from =wa to wa. Avoid a subreg on the output operand,
915 instead use a pseudo and subreg it in a move.
916 (p9_xxbrd_<mode>): Changed to ...
917 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
918 (p9_xxbrd_v2df): New expander.
919 (p9_xxbrw_<mode>): Changed to ...
920 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
921 (p9_xxbrw_v4sf): New expander.
923 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
925 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
927 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
930 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
931 multiply in 32-bit mode.
932 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
933 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
936 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
938 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
939 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
940 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
941 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
943 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
946 * dwarf2out.c (is_trivial_indirect_ref): New function.
947 (dwarf2out_late_global_decl): Do not generate a location
948 attribute for variables that have a non-trivial DECL_VALUE_EXPR
949 and that are not defined in the current unit.
951 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
953 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
954 instead of a libcall for UNORDERED.
956 2018-02-09 Tamar Christina <tamar.christina@arm.com>
959 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
960 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
962 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
965 * config/s390/s390.c (s390_set_current_function): Invoke
966 s390_indirect_branch_settings also if fndecl didn't change.
968 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
970 * config/rs6000/rs6000.md (blockage): Set length to zero.
972 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
974 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
976 2018-02-09 Jakub Jelinek <jakub@redhat.com>
979 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
980 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
984 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
985 PARALLEL incoming that failed vt_get_decl_and_offset check.
988 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
989 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
991 (get_variable_section): For decls in named .bss* sections pass true as
992 second argument to bss_initializer_p.
994 2018-02-09 Marek Polacek <polacek@redhat.com>
995 Jakub Jelinek <jakub@redhat.com>
998 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
999 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1000 Sync some changes from cxx_fold_indirect_ref.
1002 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1004 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1006 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1007 (BLOCK_INLINE_ENTRY_LABEL): New.
1008 (dwarf2out_var_location): Disregard inline entry markers.
1009 (inline_entry_data): New struct.
1010 (inline_entry_data_hasher): New hashtable type.
1011 (inline_entry_data_hasher::hash): New.
1012 (inline_entry_data_hasher::equal): New.
1013 (inline_entry_data_table): New variable.
1014 (add_high_low_attributes): Add DW_AT_entry_pc and
1015 DW_AT_GNU_entry_view attributes if a pending entry is found
1016 in inline_entry_data_table. Add old entry_pc attribute only
1017 if debug nonbinding markers are disabled.
1018 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1019 markers are enabled.
1020 (block_within_block_p, dwarf2out_inline_entry): New.
1021 (dwarf2out_finish): Check that no entries remained in
1022 inline_entry_data_table.
1023 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1024 (final_scan_insn, notice_source_line): Likewise.
1025 (rest_of_clean_state): Skip inline entry markers.
1026 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1028 * gimple.c (gimple_build_debug_inline_entry): New.
1029 * gimple.h (enum gimple_debug_subcode): Add
1030 GIMPLE_DEBUG_INLINE_ENTRY.
1031 (gimple_build_debug_inline_entry): Declare.
1032 (gimple_debug_inline_entry_p): New.
1033 (gimple_debug_nonbind_marker_p): Adjust.
1034 * insn-notes.def (INLINE_ENTRY): New.
1035 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1036 inline entry marker notes.
1037 (print_insn): Likewise.
1038 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1039 (INSN_DEBUG_MARKER_KIND): Likewise.
1040 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1041 * tree-inline.c (expand_call_inline): Build and insert
1042 debug_inline_entry stmt.
1043 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1044 inline entry blocks early, if nonbind markers are enabled.
1045 (dump_scope_block): Dump fragment info.
1046 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1047 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1048 (gimple_build_debug_inline_entry): New.
1049 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1050 Enable/disable inline entry points too.
1051 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1052 (DEBUG_INSN): Describe inline entry markers.
1054 * common.opt (gvariable-location-views): New.
1055 (gvariable-location-views=incompat5): New.
1056 * config.in: Rebuilt.
1057 * configure: Rebuilt.
1058 * configure.ac: Test assembler for view support.
1059 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1060 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1061 * dwarf2out.c (var_loc_view): New typedef.
1062 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1063 (dwarf2out_locviews_in_attribute): New.
1064 (dwarf2out_locviews_in_loclist): New.
1065 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1066 (enum dw_line_info_opcode): Add LI_adv_address.
1067 (struct dw_line_info_table): Add view.
1068 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1069 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1070 (zero_view_p): New variable.
1071 (ZERO_VIEW_P): New macro.
1072 (output_asm_line_debug_info): New.
1073 (struct var_loc_node): Add view.
1074 (add_AT_view_list, AT_loc_list): New.
1075 (add_var_loc_to_decl): Add view param. Test it against last.
1076 (new_loc_list): Add view params. Record them.
1077 (AT_loc_list_ptr): Handle loc and view lists.
1078 (view_list_to_loc_list_val_node): New.
1079 (print_dw_val): Handle dw_val_class_view_list.
1080 (size_of_die): Likewise.
1081 (value_format): Likewise.
1082 (loc_list_has_views): New.
1083 (gen_llsym): Set vl_symbol too.
1084 (maybe_gen_llsym, skip_loc_list_entry): New.
1085 (dwarf2out_maybe_output_loclist_view_pair): New.
1086 (output_loc_list): Output view list or entries too.
1087 (output_view_list_offset): New.
1088 (output_die): Handle dw_val_class_view_list.
1089 (output_dwarf_version): New.
1090 (output_compilation_unit_header): Use it.
1091 (output_skeleton_debug_sections): Likewise.
1092 (output_rnglists, output_line_info): Likewise.
1093 (output_pubnames, output_aranges): Update version comments.
1094 (output_one_line_info_table): Output view numbers in asm comments.
1095 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1096 Call maybe_gen_llsym.
1097 (loc_list_from_tree_1): Adjust.
1098 (add_AT_location_description): Create view list attribute if
1099 needed, check it's absent otherwise.
1100 (convert_cfa_to_fb_loc_list): Adjust.
1101 (maybe_emit_file): Call output_asm_line_debug_info for test.
1102 (dwarf2out_var_location): Reset views as needed. Precompute
1103 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1104 attribute. Set view.
1105 (new_line_info_table): Reset next view.
1106 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1107 (dwarf2out_source_line): Likewise. Output view resets and labels to
1108 the assembler, or select appropriate line info opcodes.
1109 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1110 (optimize_string_length): Catch it. Adjust.
1111 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1112 dw_val_class_view_list, and remove it if no longer needed.
1113 (hash_loc_list): Hash view numbers.
1114 (loc_list_hasher::equal): Compare them.
1115 (optimize_location_lists): Check whether a view list symbol is
1116 needed, and whether the locview attribute is present, and
1117 whether they match. Remove the locview attribute if no longer
1119 (index_location_lists): Call skip_loc_list_entry for test.
1120 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1121 Use output_dwarf_version.
1122 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1123 (struct dw_val_node): Add val_view_list.
1124 * final.c (SEEN_NEXT_VIEW): New.
1125 (set_next_view_needed): New.
1126 (clear_next_view_needed): New.
1127 (maybe_output_next_view): New.
1128 (final_start_function): Rename to...
1129 (final_start_function_1): ... this. Take pointer to FIRST,
1130 add SEEN parameter. Emit param bindings in the initial view.
1131 (final_start_function): Reintroduce SEEN-less interface.
1132 (final): Rename to...
1133 (final_1): ... this. Take SEEN parameter. Output final pending
1134 next view at the end.
1135 (final): Reintroduce seen-less interface.
1136 (final_scan_insn): Output pending next view before switching
1137 sections or ending a block. Mark the next view as needed when
1138 outputting variable locations. Notify debug backend of section
1139 changes, and of location view changes.
1140 (rest_of_handle_final): Adjust.
1141 * toplev.c (process_options): Autodetect value for debug variable
1142 location views option. Warn on incompat5 without -gdwarf-5.
1143 * doc/invoke.texi (gvariable-location-views): New.
1144 (gvariable-location-views=incompat5): New.
1145 (gno-variable-location-views): New.
1147 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1149 PR tree-optimization/84136
1150 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1151 that the result of find_edge is non-NULL.
1153 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1156 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1157 storing integer register in SImode. Fix cost of 256 and 512
1158 byte aligned SSE register store.
1160 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1162 * config/i386/i386.c (ix86_multiplication_cost): Fix
1163 multiplication cost for TARGET_AVX512DQ.
1165 2018-02-08 Marek Polacek <polacek@redhat.com>
1167 PR tree-optimization/84238
1168 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1171 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1173 PR tree-optimization/84265
1174 * tree-vect-stmts.c (vectorizable_store): Don't treat
1175 VMAT_CONTIGUOUS accesses as grouped.
1176 (vectorizable_load): Likewise.
1178 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1180 PR tree-optimization/81635
1181 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1182 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1183 (test_round_for_mask): New functions.
1184 (wide_int_cc_tests): Call test_round_for_mask.
1185 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1186 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1187 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1188 range returned by get_range_info.
1190 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1193 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1194 * symtab.c: Include builtins.h
1195 (symtab_node::output_to_lto_symbol_table_p): Move here
1196 from lto-streamer-out.c:output_symbol_p.
1197 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1198 (output_symbol_p): Move all logic to symtab.c
1199 (produce_symtab): Update.
1201 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1203 * config/s390/s390-opts.h (enum indirect_branch): Define.
1204 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1205 (s390_indirect_branch_via_thunk)
1206 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1207 (enum s390_indirect_branch_type): Define.
1208 * config/s390/s390.c (struct s390_frame_layout, struct
1209 machine_function): Remove.
1210 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1211 (indirect_branch_table_label_no, indirect_branch_table_name):
1213 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1214 (enum s390_indirect_branch_option): Define.
1215 (s390_return_addr_from_memory): New function.
1216 (s390_handle_string_attribute): New function.
1217 (s390_attribute_table): Add new attribute handler.
1218 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1219 (s390_indirect_branch_via_thunk): New function.
1220 (s390_indirect_branch_via_inline_thunk): New function.
1221 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1222 sibling call optimization for non z10 compiles.
1223 (s390_emit_call): Force indirect branch target to be a single
1224 register. Add r1 clobber for non-z10 compiles.
1225 (s390_emit_epilogue): Emit return jump via return_use expander.
1226 (s390_reorg): Handle JUMP_INSNs as execute targets.
1227 (s390_option_override_internal): Perform validity checks for the
1228 new command line options.
1229 (s390_indirect_branch_attrvalue): New function.
1230 (s390_indirect_branch_settings): New function.
1231 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1232 (s390_output_indirect_thunk_function): New function.
1233 (s390_code_end): Implement target hook.
1234 (s390_case_values_threshold): Implement target hook.
1235 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1237 * config/s390/s390.h (struct s390_frame_layout)
1238 (struct machine_function): Move here from s390.c.
1239 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1240 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1241 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1242 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1243 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1244 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1245 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1246 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1247 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1248 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1249 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1250 (mnemonic attribute): Add values which aren't recognized
1252 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1253 pattern for branch conversion. Fix mnemonic attribute.
1254 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1255 indirect branch via thunk if requested.
1256 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1257 ("*indirect_jump"): Disable for branch conversion using out of
1259 ("indirect_jump_via_thunk<mode>_z10")
1260 ("indirect_jump_via_thunk<mode>")
1261 ("indirect_jump_via_inlinethunk<mode>_z10")
1262 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1263 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1264 ("casesi_jump_via_inlinethunk<mode>_z10")
1265 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1266 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1267 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1268 ("*indirect2_jump"): Disable for branch conversion.
1269 ("casesi_jump"): Turn into expander and expand patterns for branch
1271 ("return_use"): New expander.
1272 ("*return"): Emit return via thunk and rename it to ...
1273 ("*return<mode>"): ... this one.
1274 * config/s390/s390.opt: Add new options and and enum for the
1277 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1279 * lra-constraints.c (match_reload): Unconditionally use
1280 gen_lowpart_SUBREG, rather than selecting between that
1281 and equivalent gen_rtx_SUBREG code.
1283 2018-02-08 Richard Biener <rguenther@suse.de>
1285 PR tree-optimization/84233
1286 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1287 changed flag instead of boguously re-using phi_inserted.
1289 2018-02-08 Martin Jambor <mjambor@suse.cz>
1291 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1292 static local variables.
1294 2018-02-08 Richard Biener <rguenther@suse.de>
1296 PR tree-optimization/84278
1297 * tree-vect-stmts.c (vectorizable_store): When looking for
1298 smaller vector types to perform grouped strided loads/stores
1299 make sure the mode is supported by the target.
1300 (vectorizable_load): Likewise.
1302 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1304 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1305 Increase LDP/STP opportunities by adding adjacent callee-saves.
1307 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1309 PR rtl-optimization/84068
1310 PR rtl-optimization/83459
1311 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1313 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1315 PR tree-optimization/84224
1316 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1317 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1320 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1323 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1324 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1325 count, remove one USE.
1327 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1329 * doc/install.texi (Configuration): Document the
1330 --with-long-double-format={ibm,ieee} PowerPC configuration
1334 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1335 Convert from define_expand to be define_insn_and_split. Rework
1336 float/double/_Float128 conversions to QI/HI/SImode to work with
1337 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1338 conversions to QI/HImode types did a store and then a load to
1339 truncate the value. For conversions to VSX registers, don't split
1340 the insn, instead emit the code directly. Use the code iterator
1341 any_fix to combine signed and unsigned conversions.
1342 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1343 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1344 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1345 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1346 (fix_<mode>di2_hw): Likewise.
1347 (fixuns_<mode>di2_hw): Likewise.
1348 (fix_<mode>si2_hw): Likewise.
1349 (fixuns_<mode>si2_hw): Likewise.
1350 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1351 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1352 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1353 fix<uns>_trunc<SFDF:mode>si2_p8.
1354 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1356 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1357 (fix<uns>_<mode>_mem): Likewise.
1358 (fctiw<u>z_<mode>_mem): Likewise.
1359 (fix<uns>_<mode>_mem): Likewise.
1360 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1361 the register allocator from doing a direct move to the GPRs to do
1362 a store, and instead use the ISA 3.0 store byte/half-word from
1363 vector register instruction. For IEEE 128-bit floating point,
1364 also optimize stores of 32-bit ints.
1365 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1367 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1369 * genextract.c (push_pathstr_operand): New function to support
1371 (walk_rtx): Call push_pathstr_operand.
1372 (print_path): Support [a-zA-Z].
1374 2018-02-07 Richard Biener <rguenther@suse.de>
1376 PR tree-optimization/84037
1377 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
1378 (cse_and_gimplify_to_preheader): Declare.
1379 (vect_get_place_in_interleaving_chain): Likewise.
1380 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1382 (_loop_vec_info::~_loop_vec_info): Delete it.
1383 (cse_and_gimplify_to_preheader): New function.
1384 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
1385 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
1386 (vectorizable_load): Likewise. For grouped stores always base
1387 the IV on the first element.
1388 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
1389 condition before gimplifying.
1391 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1393 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
1394 *DIV_EXPR and *MOD_EXPR.
1396 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
1399 * config/i386/i386.c (ix86_option_override_internal): Mask out
1400 the CF_SET bit when checking -fcf-protection.
1402 2018-02-07 Tom de Vries <tom@codesourcery.com>
1405 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
1408 2018-02-07 Richard Biener <rguenther@suse.de>
1410 PR tree-optimization/84204
1411 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
1414 PR tree-optimization/84205
1415 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
1416 special-case isl_ast_op_zdiv_r.
1418 PR tree-optimization/84223
1419 * graphite-scop-detection.c (gather_bbs::before_dom_children):
1420 Only add conditions from within the region.
1421 (gather_bbs::after_dom_children): Adjust.
1423 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
1426 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
1427 * config/avr/avr.md: Only post-reload split REG-REG moves if
1428 either register is REGERAL_REG_P.
1430 2018-02-07 Jakub Jelinek <jakub@redhat.com>
1432 PR tree-optimization/84235
1433 * tree-ssa-scopedtables.c
1434 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
1435 if the subtraction is performed in floating point type where NaNs are
1436 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
1437 build 1. Formatting fix.
1439 2018-02-06 Jakub Jelinek <jakub@redhat.com>
1442 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
1443 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
1444 and skip it regardless of bb boundaries. Use CALL_P macro,
1445 don't test INSN_P (insn) together with CALL_P or JUMP_P check
1446 unnecessarily, formatting fix.
1448 2018-02-06 Michael Collison <michael.collison@arm.com>
1450 * config/arm/thumb2.md:
1451 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
1452 (*thumb_mov_notscc): Ditto.
1454 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
1457 * config/rs6000/rs6000.md (su code attribute): Use "u" for
1458 unsigned_fix, not "s".
1460 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1462 * configure.ac (gcc_fn_eh_frame_ro): New function.
1463 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
1464 correct .eh_frame permissions.
1465 * configure: Regenerate.
1467 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
1469 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
1472 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1474 * config/rs6000/rs6000.c (rs6000_option_override_internal):
1475 Display warning message for -mno-speculate-indirect-jumps.
1477 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
1479 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
1481 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
1483 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
1485 PR tree-optimization/84225
1486 * tree-eh.c (find_trapping_overflow): Only call
1487 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
1489 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1492 * config/i386/i386.c: Reimplement the check of possible options
1493 -mibt/-mshstk conbination. Change error messages.
1494 * doc/invoke.texi: Fix a typo: remove extra '='.
1496 2018-02-06 Marek Polacek <polacek@redhat.com>
1498 PR tree-optimization/84228
1499 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
1501 2018-02-06 Tamar Christina <tamar.christina@arm.com>
1504 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
1505 emitted arch directives.
1506 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
1507 __ARM_FEATURE_COPROC before changing architectures.
1509 2018-02-06 Richard Biener <rguenther@suse.de>
1511 * config/i386/i386.c (print_reg): Fix typo.
1512 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
1514 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
1516 * configure: Regenerate.
1518 2018-02-05 Martin Sebor <msebor@redhat.com>
1520 PR tree-optimization/83369
1521 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
1524 2018-02-05 Martin Liska <mliska@suse.cz>
1526 * doc/invoke.texi: Cherry-pick upstream r323995.
1528 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
1530 * ira.c (ira_init_register_move_cost): Adjust comment.
1532 2018-02-05 Martin Liska <mliska@suse.cz>
1534 PR gcov-profile/84137
1535 * doc/gcov.texi: Fix typo in documentation.
1537 2018-02-05 Martin Liska <mliska@suse.cz>
1539 PR gcov-profile/83879
1540 * doc/gcov.texi: Document necessity of --dynamic-list-data when
1541 using dlopen functionality.
1543 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
1545 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
1546 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
1547 _mm_maskz_range_ss, _mm_mask_range_round_ss,
1548 _mm_maskz_range_round_ss): New intrinsics.
1549 (__builtin_ia32_rangesd128_round)
1550 (__builtin_ia32_rangess128_round): Remove.
1551 (__builtin_ia32_rangesd128_mask_round,
1552 __builtin_ia32_rangess128_mask_round): New builtins.
1553 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
1554 __builtin_ia32_rangess128_round): Remove.
1555 (__builtin_ia32_rangesd128_mask_round,
1556 __builtin_ia32_rangess128_mask_round): New builtins.
1557 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
1558 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
1559 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
1560 "<round_saeonly_constraint>")): Changed to ...
1561 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
1562 "<round_saeonly_scalar_constraint>")): ... this.
1563 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
1564 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
1565 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
1566 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
1567 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
1569 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
1571 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
1573 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
1574 Remove all values except native, 8540 and 8548.
1576 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
1578 * config/i386/i386.c (ix86_output_function_return): Pass
1579 INVALID_REGNUM, instead of -1, as invalid register number to
1580 indirect_thunk_name and output_indirect_thunk.
1582 2018-02-02 Julia Koval <julia.koval@intel.com>
1584 * config.gcc: Add -march=icelake.
1585 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
1586 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
1587 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
1588 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
1589 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
1590 (processor_target_table): Add icelake.
1591 (ix86_option_override_internal): Handle new PTAs.
1592 (get_builtin_code_for_version): Handle icelake.
1593 (M_INTEL_COREI7_ICELAKE): New.
1594 (fold_builtin_cpu): Handle icelake.
1595 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
1596 * doc/invoke.texi: Add -march=icelake.
1598 2018-02-02 Julia Koval <julia.koval@intel.com>
1600 * config/i386/i386.c (ix86_option_override_internal): Change flags type
1601 to wide_int_bitmask.
1602 * wide-int-bitmask.h: New.
1604 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1607 * config/i386/i386.md: Replace Pmode with word_mode in
1608 builtin_setjmp_setup and builtin_longjmp to support x32.
1610 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
1614 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
1616 (rs6000_supported_cpu_names): New static variable.
1617 (linux_cpu_translation_table): Likewise.
1618 (elf_platform) <cpu>: Define new static variable and use it.
1619 Translate kernel AT_PLATFORM name to canonical name if needed.
1620 Error if platform name is unknown.
1622 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
1625 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
1627 2018-02-01 Jeff Law <law@redhat.com>
1630 * config/i386/i386.c (release_scratch_register_on_entry): Add new
1631 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
1632 the scratch if RELEASE_VIA_POP is false.
1633 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
1634 If we have to save a temporary register, decrement SIZE appropriately.
1635 Pass new arguments to release_scratch_register_on_entry.
1636 (ix86_adjust_stack_and_probe): Likewise.
1637 (ix86_emit_probe_stack_range): Pass new arguments to
1638 release_scratch_register_on_entry.
1640 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
1642 PR rtl-optimization/84157
1643 * combine.c (change_zero_ext): Use REG_P predicate in
1644 front of HARD_REGISTER_P predicate.
1646 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
1648 * config/avr/avr.c (avr_option_override): Move disabling of
1649 -fdelete-null-pointer-checks to...
1650 * common/config/avr/avr-common.c (avr_option_optimization_table):
1653 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1655 PR tree-optimization/81635
1656 * tree-data-ref.c (split_constant_offset_1): For types that
1657 wrap on overflow, try to use range info to prove that wrapping
1660 2018-02-01 Renlin Li <renlin.li@arm.com>
1663 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
1665 (aarch64_register_move_cost): Likewise.
1666 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
1668 (REG_CLASS_NAMES): Likewise.
1669 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
1670 TAILCALL_ADDR_REGS. Remove IP registers.
1671 * config/aarch64/aarch64.md (Ucs): Update register constraint.
1673 2018-02-01 Richard Biener <rguenther@suse.de>
1675 * domwalk.h (dom_walker::dom_walker): Add additional constructor
1676 for specifying RPO order and allow NULL for that.
1677 * domwalk.c (dom_walker::dom_walker): Likewise.
1678 (dom_walker::walk): Handle NULL RPO order.
1679 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
1681 (rewrite_update_dom_walker): Likewise.
1682 (mark_def_dom_walker): Likewise.
1684 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1686 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
1687 (aarch64_maybe_expand_sve_subreg_move): Declare.
1688 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
1689 * config/aarch64/predicates.md (aarch64_any_register_operand): New
1691 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
1692 that are semantically a reverse operation.
1693 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
1694 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
1695 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
1697 (aarch64_can_change_mode_class): For big-endian, forbid changes
1698 between two SVE modes if they have different element sizes.
1700 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1702 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
1703 the TImode handling for big-endian targets.
1705 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1707 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
1708 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
1710 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
1711 Remove BSWAP handing for big-endian targets and use the form of
1712 LD1RQ appropariate for the mode.
1714 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1716 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
1717 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
1720 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
1723 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
1724 check for operands that need to go through aarch64_sve_reload_be.
1726 2018-02-01 Jakub Jelinek <jakub@redhat.com>
1728 PR tree-optimization/81661
1729 PR tree-optimization/84117
1730 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
1731 * tree-eh.c: Include gimplify.h.
1732 (find_trapping_overflow, replace_trapping_overflow,
1733 rewrite_to_non_trapping_overflow): New functions.
1734 * tree-vect-loop.c: Include tree-eh.h.
1735 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
1736 * tree-data-ref.c: Include tree-eh.h.
1737 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
1739 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
1741 PR rtl-optimization/84123
1742 * combine.c (change_zero_ext): Check if hard register satisfies
1743 can_change_dest_mode before calling gen_lowpart_SUBREG.
1745 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
1748 * ira.c (ira_init_register_move_cost): Remove assert.
1750 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
1752 PR rtl-optimization/84071
1753 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
1754 * doc/tm.texi: Regenerate.
1756 2018-01-31 Richard Biener <rguenther@suse.de>
1758 PR tree-optimization/84132
1759 * tree-data-ref.c (analyze_miv_subscript): Properly
1760 check whether evolution_function_is_affine_multivariate_p
1761 before calling gcd_of_steps_may_divide_p.
1763 2018-01-31 Julia Koval <julia.koval@intel.com>
1766 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
1767 * config/i386/i386.md (rdpid_rex64) New.
1768 (rdpid): Make 32bit only.
1770 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
1773 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
1774 an IDENTIFIER_NODE for FUNCTION_TYPE's.
1776 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
1779 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
1781 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
1783 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
1785 PR rtl-optimization/84071
1786 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
1787 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
1789 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
1791 * config/arc/arc.c (arc_handle_aux_attribute): New function.
1792 (arc_attribute_table): Add 'aux' attribute.
1793 (arc_in_small_data_p): Consider aux like variables.
1794 (arc_is_aux_reg_p): New function.
1795 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
1796 (arc_get_aux_arg): New function.
1797 (prepare_move_operands): Handle aux-register access.
1798 (arc_handle_aux_attribute): New function.
1799 * doc/extend.texi (ARC Variable attributes): Add subsection.
1801 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
1803 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
1804 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
1805 (arc_attribute_table): Add 'uncached' attribute.
1806 (arc_print_operand): Print '.di' flag for uncached memory
1808 (arc_in_small_data_p): Do not consider for small data the uncached
1810 (arc_is_uncached_mem_p): New function.
1811 * config/arc/predicates.md (compact_store_memory_operand): Check
1812 for uncached memory accesses.
1813 (nonvol_nonimm_operand): Likewise.
1814 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
1816 2018-01-31 Jakub Jelinek <jakub@redhat.com>
1819 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
1820 falign-loops=): Add Optimization flag.
1822 2018-01-30 Jeff Law <law@redhat.com>
1825 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
1826 INT_REGISTERS_SAVED. Check it prior to calling
1827 get_scratch_register_on_entry.
1828 (ix86_adjust_stack_and_probe): Similarly.
1829 (ix86_emit_probe_stack_range): Similarly.
1830 (ix86_expand_prologue): Corresponding changes.
1832 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1835 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
1836 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
1838 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
1841 * lra-constraints.c (curr_insn_transform): Process AND in the
1844 2018-01-30 Jakub Jelinek <jakub@redhat.com>
1846 PR rtl-optimization/83986
1847 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
1848 dependence against last_pending_memory_flush in addition to
1851 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
1853 PR tree-optimization/81611
1854 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
1857 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1860 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
1863 2018-01-30 Richard Biener <rguenther@suse.de>
1864 Jakub Jelinek <jakub@redhat.com>
1866 PR tree-optimization/84111
1867 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
1868 inner loops added during recursion, as they don't have up-to-date
1871 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
1874 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
1875 (can_inline_edge_by_limits_p): ... here.
1876 (can_early_inline_edge_p, check_callers,
1877 update_caller_keys, update_callee_keys, recursive_inlining,
1878 add_new_edges_to_heap, speculation_useful_p,
1879 inline_small_functions,
1880 inline_small_functions, flatten_function,
1881 inline_to_all_callers_1): Update.
1883 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
1885 * profile-count.c (profile_count::combine_with_ipa_count): Handle
1888 2018-01-30 Richard Biener <rguenther@suse.de>
1890 PR tree-optimization/83008
1891 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
1892 invariant and constant vector uses in stmts when they need
1895 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1898 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
1899 * configure: Regenerate.
1901 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
1903 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
1905 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
1906 Use gen_rtx_REG rather than gen_lowpart.
1908 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
1910 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
1911 rather than 0 when creating partial subregs.
1913 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
1915 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
1918 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
1921 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
1922 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
1923 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
1924 flags. This restores the settings used before the 2017-07-24.
1925 Turning off pre increment/decrement/modify allows IVOPTS to
1926 optimize DF/SF loops where the index is an int.
1928 2018-01-29 Richard Biener <rguenther@suse.de>
1929 Kelvin Nilsen <kelvin@gcc.gnu.org>
1932 * tree-vect-stmts.c (vectorizable_call): Don't call
1933 targetm.vectorize_builtin_md_vectorized_function if callee is
1936 2018-01-22 Carl Love <cel@us.ibm.com>
1938 * doc/extend.tex: Fix typo in second arg in
1939 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
1941 2018-01-29 Richard Biener <rguenther@suse.de>
1943 PR tree-optimization/84086
1944 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
1945 (flush_ssaname_freelist): When SSA names were released reset
1946 the SCEV hash table.
1948 2018-01-29 Richard Biener <rguenther@suse.de>
1950 PR tree-optimization/84057
1951 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
1952 removed paths when removing edges.
1954 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
1956 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
1957 -mfunction-return=@var{choice}.
1959 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
1962 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
1964 (layout::print_source_line): Likewise.
1965 (test_get_line_width_without_trailing_whitespace): Add test cases.
1967 2018-01-27 Jakub Jelinek <jakub@redhat.com>
1970 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
1973 2018-01-26 Jim Wilson <jimw@sifive.com>
1975 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
1977 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
1980 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1982 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
1983 and CMP + SUB-immediate -> SUBS.
1985 2018-01-26 Martin Sebor <msebor@redhat.com>
1987 PR tree-optimization/83896
1988 * tree-ssa-strlen.c (get_string_len): Rename...
1989 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
1990 Avoid assuming length is constant.
1991 (handle_char_store): Use HOST_WIDE_INT for string length.
1993 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
1996 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
1997 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
1999 2018-01-26 Richard Biener <rguenther@suse.de>
2001 PR rtl-optimization/84003
2002 * dse.c (record_store): Only record redundant stores when
2003 the earlier store aliases at least all accesses the later one does.
2005 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2007 PR rtl-optimization/83985
2008 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2009 REG_CFA_RESTORE insns.
2010 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2011 REG_CFA_RESTORE insns here.
2014 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2015 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2017 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2019 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2020 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2021 (arc_init): Likewise.
2022 (arc_override_options): Likewise.
2023 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2025 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2027 * config/arc/arc.h (TARGET_DBNZ): Define.
2028 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2029 properly set the tune attribute.
2030 (dbnz): Use TARGET_DBNZ guard.
2031 * config/arc/arc.opt (mtune): Add core3 option.
2033 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2035 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2036 recognize new pic like addresses.
2037 (arc_delegitimize_address): Clean up.
2039 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2041 * config/arc/arc-arches.def: Option mrf16 valid for all
2043 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2044 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2045 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2046 * config/arc/arc-tables.opt: Regenerate.
2047 * config/arc/arc.c (arc_conditional_register_usage): Handle
2048 reduced register file case.
2049 (arc_file_start): Set must have build attributes.
2050 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2052 * config/arc/arc.opt (mrf16): Add new option.
2053 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2054 * config/arc/genmultilib.awk: Handle new mrf16 option.
2055 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2056 * config/arc/t-multilib: Regenerate.
2057 * doc/invoke.texi (ARC Options): Document mrf16 option.
2059 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2061 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2062 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2063 (arc_attribute_table): Add 'secure_call' attribute.
2064 (arc_print_operand): Print secure call operand.
2065 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2067 (arc_is_secure_call_p): New function. * config/arc/arc.md
2068 (call_i): Add support for sjli instruction.
2069 (call_value_i): Likewise.
2070 * config/arc/constraints.md (Csc): New constraint.
2072 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2073 John Eric Martin <John.Martin@emmicro-us.com>
2075 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2076 * config/arc/arc.c (_arc_jli_section): New struct.
2077 (arc_jli_section): New type.
2078 (rc_jli_sections): New static variable.
2079 (arc_handle_jli_attribute): New function.
2080 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2081 (arc_file_end): New function.
2082 (TARGET_ASM_FILE_END): Define.
2083 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2084 (arc_add_jli_section): New function.
2085 (jli_call_scan): Likewise.
2086 (arc_reorg): Call jli_call_scan.
2087 (arc_output_addsi): Remove 'S' from printing asm operand.
2088 (arc_is_jli_call_p): New function.
2089 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2091 (movhi_insn): Likewise.
2092 (movsi_insn): Likewise.
2093 (movsi_set_cc_insn): Likewise.
2094 (loadqi_update): Likewise.
2095 (load_zeroextendqisi_update): Likewise.
2096 (load_signextendqisi_update): Likewise.
2097 (loadhi_update): Likewise.
2098 (load_zeroextendhisi_update): Likewise.
2099 (load_signextendhisi_update): Likewise.
2100 (loadsi_update): Likewise.
2101 (loadsf_update): Likewise.
2102 (movsicc_insn): Likewise.
2103 (bset_insn): Likewise.
2104 (bxor_insn): Likewise.
2105 (bclr_insn): Likewise.
2106 (bmsk_insn): Likewise.
2107 (bicsi3_insn): Likewise.
2108 (cmpsi_cc_c_insn): Likewise.
2109 (movsi_ne): Likewise.
2110 (movsi_cond_exec): Likewise.
2111 (clrsbsi2): Likewise.
2122 (call_i): Remove 'S' asm letter, add jli instruction.
2123 (call_value_i): Likewise.
2124 * config/arc/arc.op (mjli-always): New option.
2125 * config/arc/constraints.md (Cji): New constraint.
2126 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2128 (subsf3_fpx): Likewise.
2129 (mulsf3_fpx): Likewise.
2130 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2132 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2134 * doc/invoke.texi (ARC): Document mjli-always option.
2136 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2138 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2139 avoid addition with 0 and use incw and decw where possible.
2141 2018-01-26 Richard Biener <rguenther@suse.de>
2143 PR tree-optimization/81082
2144 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2145 association if it requires casting to unsigned.
2146 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2147 from fold_plusminus_mult_expr to catch important cases late when
2148 range info is available.
2150 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2152 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2153 * configure.ac (hidden_linkonce): New test.
2154 * configure: Regenerate.
2155 * config.in: Regenerate.
2157 2018-01-26 Julia Koval <julia.koval@intel.com>
2159 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2160 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2161 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2162 _mm_mask_bitshuffle_epi64_mask): Fix type.
2163 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2164 USI_FTYPE_V4DI_V4DI_USI): Remove.
2165 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2166 __builtin_ia32_vpshufbitqmb256_mask,
2167 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2168 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2169 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2171 2018-01-26 Alan Modra <amodra@gmail.com>
2174 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2175 UNSPEC_VBPERMQ. Sort other unspecs.
2177 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2179 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2181 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2184 * predict.c (drop_profile): Do not push/pop cfun; update also
2186 (handle_missing_profiles): Fix logic looking for zero profiles.
2188 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2191 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2192 on functions with #pragma omp declare simd or functions with simd
2194 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2195 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2196 Remove trailing \n from warning_at calls.
2198 2018-01-25 Tom de Vries <tom@codesourcery.com>
2201 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2202 for neutered workers.
2204 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2207 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2208 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2210 2018-01-24 Jeff Law <law@redhat.com>
2213 * i386.c (get_probe_interval): Move to earlier point.
2214 (ix86_compute_frame_layout): If -fstack-clash-protection and
2215 the frame is larger than the probe interval, then use pushes
2216 to save registers rather than reg->mem moves.
2217 (ix86_expand_prologue): Remove conditional for int_registers_saved
2220 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2223 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2224 min/max for never referenced object.
2226 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2229 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2231 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2232 attributes from DECL_ATTRIBUTES (decl) without affecting
2233 DECL_ATTRIBUTES (current_function_decl).
2234 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2235 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2237 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2239 PR tree-optimization/83979
2240 * fold-const.c (fold_comparison): Use constant_boolean_node
2241 instead of boolean_{true,false}_node.
2243 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2245 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2248 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2250 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2251 Simplify the clause that sets the length attribute.
2252 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2253 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2254 clause that sets the length attribute.
2255 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2257 2018-01-24 Tom de Vries <tom@codesourcery.com>
2260 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2261 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2262 Add strict parameter.
2263 (prevent_branch_around_nothing): Insert dummy insn between branch to
2264 label and label with no ptx insn inbetween.
2265 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2267 2018-01-24 Tom de Vries <tom@codesourcery.com>
2270 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2271 for neutered threads in warp.
2272 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2274 2018-01-24 Richard Biener <rguenther@suse.de>
2276 PR tree-optimization/83176
2277 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2280 2018-01-24 Richard Biener <rguenther@suse.de>
2282 PR tree-optimization/82819
2283 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2284 code generating pluses that are no-ops in the target precision.
2286 2018-01-24 Richard Biener <rguenther@suse.de>
2289 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2291 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2293 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2294 to merge probabilities.
2295 * predict.c (probably_never_executed): Also mark as cold functions
2296 with global 0 profile and guessed local profile.
2297 * profile-count.c (profile_probability::combine_with_count): New
2299 * profile-count.h (profile_probability::operator*,
2300 profile_probability::operator*=, profile_probability::operator/,
2301 profile_probability::operator/=): Reduce precision to adjusted
2302 and set value to guessed on contradictory divisions.
2303 (profile_probability::combine_with_freq): Remove.
2304 (profile_probability::combine_wiht_count): Declare.
2305 (profile_count::force_nonzero):: Set to adjusted.
2306 (profile_count::probability_in):: Set quality to adjusted.
2307 * tree-ssa-tail-merge.c (replace_block_by): Use
2310 2018-01-23 Andrew Waterman <andrew@sifive.com>
2311 Jim Wilson <jimw@sifive.com>
2313 * config/riscv/riscv.c (riscv_stack_boundary): New.
2314 (riscv_option_override): Set riscv_stack_boundary. Handle
2315 riscv_preferred_stack_boundary_arg.
2316 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2317 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2318 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2319 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2320 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2321 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2323 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2326 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2327 of struct ix86_frame.
2328 (ix86_expand_epilogue): Likewise. Add a local variable for
2329 the reg_save_offset field in struct ix86_frame.
2331 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2333 PR tree-optimization/82604
2334 * tree-loop-distribution.c (enum partition_kind): New enum item
2335 PKIND_PARTIAL_MEMSET.
2336 (partition_builtin_p): Support above new enum item.
2337 (generate_code_for_partition): Ditto.
2338 (compute_access_range): Differentiate cases that equality can be
2339 proven at all loops, the innermost loops or no loops.
2340 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2341 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2342 (finalize_partitions, distribute_loop): Don't fuse partition of
2343 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2344 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2347 2018-01-23 Martin Liska <mliska@suse.cz>
2349 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2350 order to ignore the predictor.
2351 (PRED_POLYMORPHIC_CALL): Likewise.
2352 (PRED_RECURSIVE_CALL): Likewise.
2354 2018-01-23 Martin Liska <mliska@suse.cz>
2356 * tree-profile.c (tree_profiling): Print function header to
2357 aware reader which function we are working on.
2358 * value-prof.c (gimple_find_values_to_profile): Do not print
2359 not interesting value histograms.
2361 2018-01-23 Martin Liska <mliska@suse.cz>
2363 * profile-count.h (enum profile_quality): Add
2364 profile_uninitialized as the first value. Do not number values
2365 as they are zero based.
2366 (profile_count::verify): Update sanity check.
2367 (profile_probability::verify): Likewise.
2369 2018-01-23 Nathan Sidwell <nathan@acm.org>
2371 * doc/invoke.texi (ffor-scope): Deprecate.
2373 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2375 PR tree-optimization/83510
2376 * domwalk.c (set_all_edges_as_executable): New function.
2377 (dom_walker::dom_walker): Convert bool param
2378 "skip_unreachable_blocks" to enum reachability. Move setup of
2379 edge flags to set_all_edges_as_executable and only do it when
2380 reachability is REACHABLE_BLOCKS.
2381 * domwalk.h (enum dom_walker::reachability): New enum.
2382 (dom_walker::dom_walker): Convert bool param
2383 "skip_unreachable_blocks" to enum reachability.
2384 (set_all_edges_as_executable): New decl.
2385 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
2386 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
2388 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
2389 but converting true to REACHABLE_BLOCKS.
2390 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
2392 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
2393 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
2394 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
2396 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
2397 if check_all_array_refs will be called.
2399 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2401 * tree.c (selftest::test_location_wrappers): Add more test
2404 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2406 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
2407 (selftest::test_bit_in_range): Likewise.
2409 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2412 * doc/sourcebuild.texi (vect_float): Say that the selector
2413 only describes the situation when -funsafe-math-optimizations is on.
2414 (vect_float_strict): Document.
2416 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
2418 PR tree-optimization/83965
2419 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
2420 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
2421 instead of checking only for a reduction.
2422 (vect_recog_widen_sum_pattern): Likewise.
2424 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2426 * predict.c (probably_never_executed): Only use precise profile info.
2427 (compute_function_frequency): Skip after inlining hack since we now
2428 have quality checking.
2430 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2432 * profile-count.h (profile_probability::very_unlikely,
2433 profile_probability::unlikely, profile_probability::even): Set
2434 precision to guessed.
2436 2018-01-23 Richard Biener <rguenther@suse.de>
2438 PR tree-optimization/83963
2439 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
2440 Properly terminate dominator walk when crossing the exit edge not
2441 when visiting its source block.
2443 2018-01-23 Jakub Jelinek <jakub@redhat.com>
2446 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
2447 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
2449 2018-01-22 Jakub Jelinek <jakub@redhat.com>
2451 PR tree-optimization/83957
2452 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
2453 semicolon after for body surrounded by braces.
2455 PR tree-optimization/83081
2456 * profile-count.h (profile_probability::split): New method.
2457 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
2458 Use profile_probability::split.
2459 (do_compare_rtx_and_jump): Fix adjustment of probabilities
2460 when splitting a single conditional jump into 2.
2462 2018-01-22 David Malcolm <dmalcolm@redhat.com>
2464 PR tree-optimization/69452
2465 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
2468 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2470 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
2471 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
2472 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
2474 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2476 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
2477 * config/rl78/rl78.md (movdi): New define_expand.
2478 * config/rl78/rl78.c (rl78_split_movdi): New function.
2480 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
2483 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
2485 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
2486 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
2487 128-bit to produce an UNSPEC move to get the double word with the
2488 signbit and then a shift directly to do signbit.
2489 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
2490 implementation with a new version that just does either a direct
2491 move or a regular move. Move memory interface to separate insns.
2492 Move insns so they are next to the expander.
2493 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
2494 with signbit move. Split big and little endian case.
2495 (signbit<mode>2_dm_mem_le): Likewise.
2496 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
2497 (signbit<mode>2_dm2): Likewise.
2499 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2501 * config/rl78/rl78.md (anddi3): New define_expand.
2503 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2505 * config/rl78/rl78.md (umindi3): New define_expand.
2507 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2509 * config/rl78/rl78.md (smindi3): New define_expand.
2511 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2513 * config/rl78/rl78.md (smaxdi3): New define_expand.
2515 2018-01-22 Carl Love <cel@us.ibm.com>
2517 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
2518 LVX_V1TI): Add macro expansion.
2519 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
2520 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
2521 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
2522 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
2523 Change check to determine if the instruction is a byte reversing
2524 entry. Fix typo in comment.
2525 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
2526 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
2527 Add def_builtin calls for new builtins.
2528 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
2529 Add define_insn expansion.
2531 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2533 * config/rl78/rl78.md (umaxdi3): New define_expand.
2535 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
2537 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
2538 for non-QImode registers.
2540 2018-01-22 Richard Biener <rguenther@suse.de>
2542 PR tree-optimization/83963
2543 * graphite-scop-detection.c (scop_detection::get_sese): Delay
2544 including the loop exit block.
2545 (scop_detection::merge_sese): Likewise.
2546 (scop_detection::add_scop): Do it here instead.
2548 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2550 * doc/sourcebuild.texi (arm_softfloat): Document.
2552 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
2555 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
2556 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
2557 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
2559 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2560 David Edelsohn <dje.gcc@gmail.com>
2563 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2564 Change "crset eq" to "crset 2".
2565 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2566 (*call_indirect_aix<mode>_nospec): Likewise.
2567 (*call_value_indirect_aix<mode>_nospec): Likewise.
2568 (*call_indirect_elfv2<mode>_nospec): Likewise.
2569 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
2570 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
2571 change assembly output from . to $.
2572 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2573 (indirect_jump<mode>_nospec): Change assembly output from . to $.
2574 (*tablejump<mode>_internal1_nospec): Likewise.
2576 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
2579 * config/sh/sh_optimize_sett_clrt.cc:
2580 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
2582 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
2584 PR tree-optimization/83940
2585 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
2586 offset_dt to vect_constant_def rather than vect_unknown_def_type.
2587 (vect_check_load_store_mask): Add a mask_dt_out parameter and
2588 use it to pass back the definition type.
2589 (vect_check_store_rhs): Likewise rhs_dt_out.
2590 (vect_build_gather_load_calls): Add a mask_dt argument and use
2591 it instead of a call to vect_is_simple_use.
2592 (vectorizable_store): Update calls to vect_check_load_store_mask
2593 and vect_check_store_rhs. Use the dt returned by the latter instead
2594 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
2595 instead of calls to vect_is_simple_use. Pass the scalar rather
2596 than the vector operand to vect_is_simple_use when handling
2597 second and subsequent copies of an rhs value.
2598 (vectorizable_load): Update calls to vect_check_load_store_mask
2599 and vect_build_gather_load_calls. Use the cached mask_dt and
2600 gs_info.offset_dt instead of calls to vect_is_simple_use.
2602 2018-01-20 Jakub Jelinek <jakub@redhat.com>
2605 * tree-emutls.c: Include gimplify.h.
2606 (lower_emutls_2): New function.
2607 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
2608 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
2609 it before further processing.
2612 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
2613 UINTVAL (trueop1) instead of INTVAL (op1).
2615 2018-01-19 Jakub Jelinek <jakub@redhat.com>
2619 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
2620 INCOMING_FRAME_SP_OFFSET if not defined.
2621 (scan_trace): Add ENTRY argument. If true and
2622 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
2623 emit a note to adjust the CFA offset.
2624 (create_cfi_notes): Adjust scan_trace callers.
2625 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
2626 INCOMING_FRAME_SP_OFFSET in the CIE.
2627 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
2628 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
2630 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
2631 * doc/tm.texi: Regenerated.
2633 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2635 PR rtl-optimization/83147
2636 * lra-constraints.c (remove_inheritance_pseudos): Use
2637 lra_substitute_pseudo_within_insn.
2639 2018-01-19 Tom de Vries <tom@codesourcery.com>
2640 Cesar Philippidis <cesar@codesourcery.com>
2643 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
2645 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
2648 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
2649 spaces for function labels.
2651 2018-01-19 Martin Liska <mliska@suse.cz>
2653 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
2654 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
2655 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
2656 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
2657 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
2658 (PRED_CONST_RETURN): Change from 69 to 65.
2659 (PRED_NULL_RETURN): Change from 91 to 71.
2660 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
2661 (PRED_LOOP_GUARD): Change from 66 to 73.
2663 2018-01-19 Martin Liska <mliska@suse.cz>
2665 * predict.c (predict_insn_def): Add new assert.
2666 (struct branch_predictor): Change type to signed integer.
2667 (test_prediction_value_range): Amend test to cover
2669 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
2670 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
2671 (PRED_LOOP_ITERATIONS_MAX): Likewise.
2672 (PRED_LOOP_IV_COMPARE): Likewise.
2673 * predict.h (PROB_UNINITIALIZED): Define new constant.
2675 2018-01-19 Martin Liska <mliska@suse.cz>
2677 * predict.c (dump_prediction): Add new format for
2678 analyze_brprob.py script which is enabled with -details
2680 * profile-count.h (precise_p): New function.
2682 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
2684 PR tree-optimization/83922
2685 * tree-vect-loop.c (vect_verify_full_masking): Return false if
2686 there are no statements that need masking.
2687 (vect_active_double_reduction_p): New function.
2688 (vect_analyze_loop_operations): Use it when handling phis that
2689 are not in the loop header.
2691 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
2693 PR tree-optimization/83914
2694 * tree-vect-loop.c (vectorizable_induction): Don't convert
2695 init_expr or apply the peeling adjustment for inductions
2696 that are nested within the vectorized loop.
2698 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2700 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
2703 2018-01-18 Jakub Jelinek <jakub@redhat.com>
2707 * function.h (gimplify_parameters): Add gimple_seq * argument.
2708 * function.c: Include gimple.h and options.h.
2709 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
2710 for the added local temporaries if needed.
2711 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
2712 if there are any parameter cleanups, wrap whole body into a
2713 try/finally with the cleanups.
2715 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
2718 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
2719 Use GET_MODE_CLASS for scalar floating point.
2721 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
2725 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
2726 Fix call of call_cgraph_insertion_hooks.
2728 2018-01-18 Martin Sebor <msebor@redhat.com>
2730 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
2732 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
2735 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
2738 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
2741 * common.opt: (-ffile-prefix-map): New option.
2742 * opts.c (common_handle_option): Defer it.
2743 * opts-global.c (handle_common_deferred_options): Handle it.
2744 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
2745 * file-prefix-map.h: New file.
2746 (remap_debug_filename, add_debug_prefix_map): ...here.
2747 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
2748 * final.c (debug_prefix_map, add_debug_prefix_map
2749 remap_debug_filename): Move to...
2750 * file-prefix-map.c: New file.
2751 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
2752 generalize, get rid of alloca(), use strrchr() instead of strchr().
2753 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
2754 Implement in terms of add_prefix_map().
2755 (remap_macro_filename, remap_debug_filename): Implement in term of
2757 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
2758 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
2759 * dbxout.c: Include file-prefix-map.h.
2760 * varasm.c: Likewise.
2761 * vmsdbgout.c: Likewise.
2762 * xcoffout.c: Likewise.
2763 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
2764 * doc/cppopts.texi (-fmacro-prefix-map): Document.
2765 * doc/invoke.texi (-ffile-prefix-map): Document.
2766 (-fdebug-prefix-map): Update description.
2768 2018-01-18 Martin Liska <mliska@suse.cz>
2770 * config/i386/i386.c (indirect_thunk_name): Document that also
2772 (output_indirect_thunk): Document why both instructions
2773 (pause and lfence) are generated.
2775 2018-01-18 Richard Biener <rguenther@suse.de>
2777 PR tree-optimization/83887
2778 * graphite-scop-detection.c
2779 (scop_detection::get_nearest_dom_with_single_entry): Remove.
2780 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
2781 (scop_detection::merge_sese): Re-implement with a flood-fill
2782 algorithm that properly finds a SESE region if it exists.
2784 2018-01-18 Jakub Jelinek <jakub@redhat.com>
2787 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
2788 pointer_diff optimizations use view_convert instead of convert.
2790 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2792 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2793 Generate different code for -mno-speculate-indirect-jumps.
2794 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2795 (*call_indirect_aix<mode>): Disable for
2796 -mno-speculate-indirect-jumps.
2797 (*call_indirect_aix<mode>_nospec): New define_insn.
2798 (*call_value_indirect_aix<mode>): Disable for
2799 -mno-speculate-indirect-jumps.
2800 (*call_value_indirect_aix<mode>_nospec): New define_insn.
2801 (*sibcall_nonlocal_sysv<mode>): Generate different code for
2802 -mno-speculate-indirect-jumps.
2803 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2805 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
2807 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
2808 long double type, set the flags for noting the default long double
2809 type, even if we don't pass or return a long double type.
2811 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
2814 * ipa-inline.c (flatten_function): Do not overwrite final inlining
2817 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
2819 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
2820 support for merge[hl].
2821 (fold_mergehl_helper): New helper function.
2822 (tree-vector-builder.h): New #include for tree_vector_builder usage.
2823 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
2824 (altivec_vmrglw_direct): Add xxmrglw insn.
2826 2018-01-17 Andrew Waterman <andrew@sifive.com>
2828 * config/riscv/riscv.c (riscv_conditional_register_usage): If
2829 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
2831 2018-01-17 David Malcolm <dmalcolm@redhat.com>
2834 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
2835 call the lto_location_cache before reading the
2836 DECL_SOURCE_LOCATION of the types.
2838 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
2839 Richard Sandiford <richard.sandiford@linaro.org>
2841 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
2842 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
2843 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
2844 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
2845 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
2847 * config/aarch64/constraints.md (aarch64_movti_operand):
2849 * config/aarch64/predicates.md (Uti): Add new constraint.
2851 2018-01-17 Carl Love <cel@us.ibm.com>
2853 * config/rs6000/vsx.md (define_expand xl_len_r,
2854 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
2855 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
2857 (define_expand, define_insn): Move the shift left from the
2858 define_insn to the define_expand for lxvl and stxvl instructions.
2859 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
2860 and XL_LEN_R definitions to PURE.
2862 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
2864 * config/i386/i386.c (indirect_thunk_name): Declare regno
2865 as unsigned int. Compare regno with INVALID_REGNUM.
2866 (output_indirect_thunk): Ditto.
2867 (output_indirect_thunk_function): Ditto.
2868 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
2869 in the call to output_indirect_thunk_function.
2871 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
2874 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
2875 rather than the size of inner_type to determine the stack slot size
2876 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
2878 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
2881 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
2884 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
2886 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
2887 endian Linux systems to optionally enable multilibs for selecting
2888 the long double type if the user configured an explicit type.
2889 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
2890 have no long double multilibs if not defined.
2891 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
2892 warn if the user used -mabi={ieee,ibm}longdouble and we built
2893 multilibs for long double.
2894 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
2895 appropriate multilib option.
2896 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
2898 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
2899 for building long double multilibs.
2900 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
2902 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
2904 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
2907 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
2909 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
2912 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
2915 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
2918 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
2920 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
2921 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
2923 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
2925 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
2926 different rtl trees depending on TARGET_64BIT.
2927 (rs6000_gen_lvx): Likewise.
2929 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
2931 * config/visium/visium.md (nop): Tweak comment.
2932 (hazard_nop): Likewise.
2934 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2936 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
2937 -mspeculate-indirect-jumps.
2938 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
2939 for -mno-speculate-indirect-jumps.
2940 (*call_indirect_elfv2<mode>_nospec): New define_insn.
2941 (*call_value_indirect_elfv2<mode>): Disable for
2942 -mno-speculate-indirect-jumps.
2943 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
2944 (indirect_jump): Emit different RTL for
2945 -mno-speculate-indirect-jumps.
2946 (*indirect_jump<mode>): Disable for
2947 -mno-speculate-indirect-jumps.
2948 (*indirect_jump<mode>_nospec): New define_insn.
2949 (tablejump): Emit different RTL for
2950 -mno-speculate-indirect-jumps.
2951 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
2952 (tablejumpsi_nospec): New define_expand.
2953 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
2954 (tablejumpdi_nospec): New define_expand.
2955 (*tablejump<mode>_internal1): Disable for
2956 -mno-speculate-indirect-jumps.
2957 (*tablejump<mode>_internal1_nospec): New define_insn.
2958 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
2961 2018-01-16 Artyom Skrobov tyomitch@gmail.com
2963 * caller-save.c (insert_save): Drop unnecessary parameter. All
2966 2018-01-16 Jakub Jelinek <jakub@redhat.com>
2967 Richard Biener <rguenth@suse.de>
2970 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
2971 return early, inline manually is_gimple_sizepos. Make sure if we
2972 call gimplify_expr we don't end up with a gimple constant.
2973 * tree.c (variably_modified_type_p): Don't return true for
2974 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
2975 * gimplify.h (is_gimple_sizepos): Remove.
2977 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
2979 PR tree-optimization/83857
2980 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
2981 vectorizable_live_operation for pure SLP statements.
2982 (vectorizable_live_operation): Handle PHIs.
2984 2018-01-16 Richard Biener <rguenther@suse.de>
2986 PR tree-optimization/83867
2987 * tree-vect-stmts.c (vect_transform_stmt): Precompute
2988 nested_in_vect_loop_p since the scalar stmt may get invalidated.
2990 2018-01-16 Jakub Jelinek <jakub@redhat.com>
2993 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
2994 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
2995 If off is not INTEGER_CST, issue a may not be aligned warning
2996 rather than isn't aligned. Use isn%'t rather than isn't.
2997 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
2999 <case MULT_EXPR>: Improve the case when bottom and one of the
3000 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3001 operand, in that case check if the other operand is multiple of
3002 bottom divided by the INTEGER_CST operand.
3004 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3007 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3008 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3009 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3010 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3011 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3012 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3013 (pa_function_arg_size): New function.
3015 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3017 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3018 in a separate statement.
3020 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3022 PR tree-optimization/83847
3023 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3024 group gathers and scatters.
3026 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3028 PR rtl-optimization/86620
3029 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3031 PR rtl-optimization/83213
3032 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3033 to last if both are JUMP_INSNs.
3035 PR tree-optimization/83843
3036 * gimple-ssa-store-merging.c
3037 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3038 store_immediate_info for bswap/nop orig_stores.
3040 2018-01-15 Andrew Waterman <andrew@sifive.com>
3042 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3044 <UDIV>: Increase cost if !TARGET_DIV.
3046 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3048 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3049 (define_attr "cr_logical_3op"): New.
3050 (cceq_ior_compare): Adjust.
3051 (cceq_ior_compare_complement): Adjust.
3052 (*cceq_rev_compare): Adjust.
3053 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3054 (is_cracked_insn): Adjust.
3055 (insn_must_be_first_in_group): Adjust.
3056 * config/rs6000/40x.md: Adjust.
3057 * config/rs6000/440.md: Adjust.
3058 * config/rs6000/476.md: Adjust.
3059 * config/rs6000/601.md: Adjust.
3060 * config/rs6000/603.md: Adjust.
3061 * config/rs6000/6xx.md: Adjust.
3062 * config/rs6000/7450.md: Adjust.
3063 * config/rs6000/7xx.md: Adjust.
3064 * config/rs6000/8540.md: Adjust.
3065 * config/rs6000/cell.md: Adjust.
3066 * config/rs6000/e300c2c3.md: Adjust.
3067 * config/rs6000/e500mc.md: Adjust.
3068 * config/rs6000/e500mc64.md: Adjust.
3069 * config/rs6000/e5500.md: Adjust.
3070 * config/rs6000/e6500.md: Adjust.
3071 * config/rs6000/mpc.md: Adjust.
3072 * config/rs6000/power4.md: Adjust.
3073 * config/rs6000/power5.md: Adjust.
3074 * config/rs6000/power6.md: Adjust.
3075 * config/rs6000/power7.md: Adjust.
3076 * config/rs6000/power8.md: Adjust.
3077 * config/rs6000/power9.md: Adjust.
3078 * config/rs6000/rs64.md: Adjust.
3079 * config/rs6000/titan.md: Adjust.
3081 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3083 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3084 ix86_indirect_branch_register logic.
3086 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3088 * config/i386/constraints.md (Bs): Update
3089 ix86_indirect_branch_register check. Don't check
3090 ix86_indirect_branch_register with GOT_memory_operand.
3092 * config/i386/predicates.md (GOT_memory_operand): Don't check
3093 ix86_indirect_branch_register here.
3094 (GOT32_symbol_operand): Likewise.
3096 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3098 * config/i386/predicates.md (constant_call_address_operand):
3099 Rewrite ix86_indirect_branch_register logic.
3100 (sibcall_insn_operand): Likewise.
3102 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3104 * config/i386/constraints.md (Bs): Replace
3105 ix86_indirect_branch_thunk_register with
3106 ix86_indirect_branch_register.
3108 * config/i386/i386.md (indirect_jump): Likewise.
3109 (tablejump): Likewise.
3110 (*sibcall_memory): Likewise.
3111 (*sibcall_value_memory): Likewise.
3112 Peepholes of indirect call and jump via memory: Likewise.
3113 * config/i386/i386.opt: Likewise.
3114 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3115 (GOT_memory_operand): Likewise.
3116 (call_insn_operand): Likewise.
3117 (sibcall_insn_operand): Likewise.
3118 (GOT32_symbol_operand): Likewise.
3120 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3123 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3124 type rather than type addr's type points to.
3125 (expand_omp_atomic_mutex): Likewise.
3126 (expand_omp_atomic): Likewise.
3128 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3131 * config/i386/i386.c (output_indirect_thunk_function): Use
3132 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3133 for __x86_return_thunk.
3135 2018-01-15 Richard Biener <rguenther@suse.de>
3138 * expmed.c (extract_bit_field_1): Fix typo.
3140 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3143 * config/arm/iterators.md (VF): New mode iterator.
3144 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3145 Remove integer-related logic from pattern.
3146 (neon_vabd<mode>_3): Likewise.
3148 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3151 * common.opt (fstrict-overflow): No longer an alias.
3152 (fwrapv-pointer): New option.
3153 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3154 also for pointer types based on flag_wrapv_pointer.
3155 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3156 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3157 opts->x_flag_wrapv got set.
3158 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3159 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3160 POINTER_TYPE_OVERFLOW_UNDEFINED.
3161 * match.pd: Likewise in address comparison pattern.
3162 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3164 2018-01-15 Richard Biener <rguenther@suse.de>
3167 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3168 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3169 Reset type names to their identifier if their TYPE_DECL doesn't
3170 have linkage (and thus is used for ODR and devirt).
3171 (save_debug_info_for_decl): Remove.
3172 (save_debug_info_for_type): Likewise.
3173 (add_tree_to_fld_list): Adjust.
3174 * tree-pretty-print.c (dump_generic_node): Make dumping of
3175 type names more robust.
3177 2018-01-15 Richard Biener <rguenther@suse.de>
3179 * BASE-VER: Bump to 8.0.1.
3181 2018-01-14 Martin Sebor <msebor@redhat.com>
3184 * builtins.c (check_access): Avoid warning when the no-warning bit
3187 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3189 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3190 * ira-color (allocno_hard_regs_compare): Likewise.
3192 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3195 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3196 Use .pushsection/.popsection.
3198 2018-01-14 Martin Sebor <msebor@redhat.com>
3201 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3203 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3205 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3206 entry from extra_headers.
3207 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3208 extra_headers, make the list bitwise identical to the i?86-*-* one.
3210 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3212 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3213 -mcmodel=large with -mindirect-branch=thunk,
3214 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3215 -mfunction-return=thunk-extern.
3216 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3217 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3218 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3220 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3222 * config/i386/i386.c (print_reg): Print the name of the full
3223 integer register without '%'.
3224 (ix86_print_operand): Handle 'V'.
3225 * doc/extend.texi: Document 'V' modifier.
3227 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3229 * config/i386/constraints.md (Bs): Disallow memory operand for
3230 -mindirect-branch-register.
3232 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3233 (GOT_memory_operand): Likewise.
3234 (call_insn_operand): Likewise.
3235 (sibcall_insn_operand): Likewise.
3236 (GOT32_symbol_operand): Likewise.
3237 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3238 for -mindirect-branch-register.
3239 (tablejump): Likewise.
3240 (*sibcall_memory): Likewise.
3241 (*sibcall_value_memory): Likewise.
3242 Disallow peepholes of indirect call and jump via memory for
3243 -mindirect-branch-register.
3244 (*call_pop): Replace m with Bw.
3245 (*call_value_pop): Likewise.
3246 (*sibcall_pop_memory): Replace m with Bs.
3247 * config/i386/i386.opt (mindirect-branch-register): New option.
3248 * doc/invoke.texi: Document -mindirect-branch-register option.
3250 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3252 * config/i386/i386-protos.h (ix86_output_function_return): New.
3253 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3254 set function_return_type.
3255 (indirect_thunk_name): Add ret_p to indicate thunk for function
3257 (output_indirect_thunk_function): Pass false to
3258 indirect_thunk_name.
3259 (ix86_output_indirect_branch_via_reg): Likewise.
3260 (ix86_output_indirect_branch_via_push): Likewise.
3261 (output_indirect_thunk_function): Create alias for function
3262 return thunk if regno < 0.
3263 (ix86_output_function_return): New function.
3264 (ix86_handle_fndecl_attribute): Handle function_return.
3265 (ix86_attribute_table): Add function_return.
3266 * config/i386/i386.h (machine_function): Add
3267 function_return_type.
3268 * config/i386/i386.md (simple_return_internal): Use
3269 ix86_output_function_return.
3270 (simple_return_internal_long): Likewise.
3271 * config/i386/i386.opt (mfunction-return=): New option.
3272 (indirect_branch): Mention -mfunction-return=.
3273 * doc/extend.texi: Document function_return function attribute.
3274 * doc/invoke.texi: Document -mfunction-return= option.
3276 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3278 * config/i386/i386-opts.h (indirect_branch): New.
3279 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3280 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3281 with local indirect jump when converting indirect call and jump.
3282 (ix86_set_indirect_branch_type): New.
3283 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3284 (indirectlabelno): New.
3285 (indirect_thunk_needed): Likewise.
3286 (indirect_thunk_bnd_needed): Likewise.
3287 (indirect_thunks_used): Likewise.
3288 (indirect_thunks_bnd_used): Likewise.
3289 (INDIRECT_LABEL): Likewise.
3290 (indirect_thunk_name): Likewise.
3291 (output_indirect_thunk): Likewise.
3292 (output_indirect_thunk_function): Likewise.
3293 (ix86_output_indirect_branch_via_reg): Likewise.
3294 (ix86_output_indirect_branch_via_push): Likewise.
3295 (ix86_output_indirect_branch): Likewise.
3296 (ix86_output_indirect_jmp): Likewise.
3297 (ix86_code_end): Call output_indirect_thunk_function if needed.
3298 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3300 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3301 (ix86_attribute_table): Add indirect_branch.
3302 * config/i386/i386.h (machine_function): Add indirect_branch_type
3303 and has_local_indirect_jump.
3304 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3306 (tablejump): Likewise.
3307 (*indirect_jump): Use ix86_output_indirect_jmp.
3308 (*tablejump_1): Likewise.
3309 (simple_return_indirect_internal): Likewise.
3310 * config/i386/i386.opt (mindirect-branch=): New option.
3311 (indirect_branch): New.
3314 (thunk-inline): Likewise.
3315 (thunk-extern): Likewise.
3316 * doc/extend.texi: Document indirect_branch function attribute.
3317 * doc/invoke.texi: Document -mindirect-branch= option.
3319 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3322 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3324 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3326 * ipa-inline.c (want_inline_small_function_p): Return false if
3327 inlining has already failed with CIF_FINAL_ERROR.
3328 (update_caller_keys): Call want_inline_small_function_p before
3330 (update_callee_keys): Likewise.
3332 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3334 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3336 (rs6000_quadword_masked_address_p): Likewise.
3337 (quad_aligned_load_p): Likewise.
3338 (quad_aligned_store_p): Likewise.
3339 (const_load_sequence_p): Add comment to describe the outer-most loop.
3340 (mimic_memory_attributes_and_flags): New function.
3341 (rs6000_gen_stvx): Likewise.
3342 (replace_swapped_aligned_store): Likewise.
3343 (rs6000_gen_lvx): Likewise.
3344 (replace_swapped_aligned_load): Likewise.
3345 (replace_swapped_load_constant): Capitalize argument name in
3346 comment describing this function.
3347 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3348 and stores that access quad-word aligned addresses and replace
3349 with stvx or lvx instructions when appropriate.
3350 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3351 New function prototype.
3352 (rs6000_quadword_masked_address_p): Likewise.
3353 (rs6000_gen_lvx): Likewise.
3354 (rs6000_gen_stvx): Likewise.
3355 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3356 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3357 when memory address is aligned.
3358 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3359 this split to select lvx instruction when memory address is aligned.
3360 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3361 instruction when memory address is aligned.
3362 (*vsx_le_perm_load_v16qi): Likewise.
3363 (four unnamed splitters): Modify to select the stvx instruction
3364 when memory is aligned.
3366 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3368 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3369 which appears in the queue multiple times.
3371 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3372 Alan Hayward <alan.hayward@arm.com>
3373 David Sherwood <david.sherwood@arm.com>
3375 * tree-vectorizer.h (vec_lower_bound): New structure.
3376 (_loop_vec_info): Add check_nonzero and lower_bounds.
3377 (LOOP_VINFO_CHECK_NONZERO): New macro.
3378 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
3379 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
3380 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
3381 fields. Make seg_len the distance travelled, not including the
3383 (dr_direction_indicator): Declare.
3384 (dr_zero_step_indicator): Likewise.
3385 (dr_known_forward_stride_p): Likewise.
3386 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
3388 (runtime_alias_check_p): Allow runtime alias checks with
3390 (operator ==): Compare access_size and align.
3391 (prune_runtime_alias_test_list): Rework for new distinction between
3392 the access_size and seg_len.
3393 (create_intersect_range_checks_index): Likewise. Cope with polynomial
3395 (get_segment_min_max): New function.
3396 (create_intersect_range_checks): Use it.
3397 (dr_step_indicator): New function.
3398 (dr_direction_indicator): Likewise.
3399 (dr_zero_step_indicator): Likewise.
3400 (dr_known_forward_stride_p): Likewise.
3401 * tree-loop-distribution.c (data_ref_segment_size): Return
3402 DR_STEP * (niters - 1).
3403 (compute_alias_check_pairs): Update call to the dr_with_seg_len
3405 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
3406 (vect_preserves_scalar_order_p): New function, split out from...
3407 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
3408 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
3409 (vect_vfa_access_size): New function.
3410 (vect_vfa_align): Likewise.
3411 (vect_compile_time_alias): Take access_size_a and access_b arguments.
3412 (dump_lower_bound): New function.
3413 (vect_check_lower_bound): Likewise.
3414 (vect_small_gap_p): Likewise.
3415 (vectorizable_with_step_bound_p): Likewise.
3416 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
3417 depencies if the vectorization factor is 1. Convert the checks
3418 for nonzero steps into checks on the bounds of DR_STEP. Try using
3419 a bunds check for variable steps if the minimum required step is
3420 relatively small. Update calls to the dr_with_seg_len
3421 constructor and to vect_compile_time_alias.
3422 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
3424 (vect_loop_versioning): Call it.
3425 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
3427 (vect_estimate_min_profitable_iters): Account for any bounds checks.
3429 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3430 Alan Hayward <alan.hayward@arm.com>
3431 David Sherwood <david.sherwood@arm.com>
3433 * doc/sourcebuild.texi (vect_scatter_store): Document.
3434 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
3436 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
3438 * genopinit.c (main): Add supports_vec_scatter_store and
3439 supports_vec_scatter_store_cached to target_optabs.
3440 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
3441 IFN_MASK_SCATTER_STORE.
3442 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
3444 * internal-fn.h (internal_store_fn_p): Declare.
3445 (internal_fn_stored_value_index): Likewise.
3446 * internal-fn.c (scatter_store_direct): New macro.
3447 (expand_scatter_store_optab_fn): New function.
3448 (direct_scatter_store_optab_supported_p): New macro.
3449 (internal_store_fn_p): New function.
3450 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
3451 IFN_MASK_SCATTER_STORE.
3452 (internal_fn_mask_index): Likewise.
3453 (internal_fn_stored_value_index): New function.
3454 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
3456 * optabs-query.h (supports_vec_scatter_store_p): Declare.
3457 * optabs-query.c (supports_vec_scatter_store_p): New function.
3458 * tree-vectorizer.h (vect_get_store_rhs): Declare.
3459 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
3460 true for scatter stores.
3461 (vect_gather_scatter_fn_p): Handle scatter stores too.
3462 (vect_check_gather_scatter): Consider using scatter stores if
3463 supports_vec_scatter_store_p.
3464 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
3466 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3467 internal_fn_stored_value_index.
3468 (check_load_store_masking): Handle scatter stores too.
3469 (vect_get_store_rhs): Make public.
3470 (vectorizable_call): Use internal_store_fn_p.
3471 (vectorizable_store): Handle scatter store internal functions.
3472 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
3473 when deciding whether the end of the group has been reached.
3474 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
3475 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
3476 (mask_scatter_store<mode>): New insns.
3478 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3479 Alan Hayward <alan.hayward@arm.com>
3480 David Sherwood <david.sherwood@arm.com>
3482 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
3483 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
3484 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
3486 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
3487 Use vect_truncate_gather_scatter_offset if we can't treat the
3488 operation as a normal gather load or scatter store.
3489 (get_group_load_store_type): Take the gather_scatter_info
3490 as argument. Try using a gather load or scatter store for
3491 single-element groups.
3492 (get_load_store_type): Update calls to get_group_load_store_type
3493 and vect_use_strided_gather_scatters_p.
3495 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3496 Alan Hayward <alan.hayward@arm.com>
3497 David Sherwood <david.sherwood@arm.com>
3499 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
3500 optional tree argument.
3501 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
3503 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
3504 but continue to use the current value as a fallback.
3505 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
3506 to compare the updates.
3507 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
3508 (get_load_store_type): Use it when handling a strided access.
3509 (vect_get_strided_load_store_ops): New function.
3510 (vect_get_data_ptr_increment): Likewise.
3511 (vectorizable_load): Handle strided gather loads. Always pass
3512 a step to vect_create_data_ref_ptr and bump_vector_ptr.
3514 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3515 Alan Hayward <alan.hayward@arm.com>
3516 David Sherwood <david.sherwood@arm.com>
3518 * doc/md.texi (gather_load@var{m}): Document.
3519 (mask_gather_load@var{m}): Likewise.
3520 * genopinit.c (main): Add supports_vec_gather_load and
3521 supports_vec_gather_load_cached to target_optabs.
3522 * optabs-tree.c (init_tree_optimization_optabs): Use
3523 ggc_cleared_alloc to allocate target_optabs.
3524 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
3525 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
3527 * internal-fn.h (internal_load_fn_p): Declare.
3528 (internal_gather_scatter_fn_p): Likewise.
3529 (internal_fn_mask_index): Likewise.
3530 (internal_gather_scatter_fn_supported_p): Likewise.
3531 * internal-fn.c (gather_load_direct): New macro.
3532 (expand_gather_load_optab_fn): New function.
3533 (direct_gather_load_optab_supported_p): New macro.
3534 (direct_internal_fn_optab): New function.
3535 (internal_load_fn_p): Likewise.
3536 (internal_gather_scatter_fn_p): Likewise.
3537 (internal_fn_mask_index): Likewise.
3538 (internal_gather_scatter_fn_supported_p): Likewise.
3539 * optabs-query.c (supports_at_least_one_mode_p): New function.
3540 (supports_vec_gather_load_p): Likewise.
3541 * optabs-query.h (supports_vec_gather_load_p): Declare.
3542 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
3543 and memory_type field.
3544 (NUM_PATTERNS): Bump to 15.
3545 * tree-vect-data-refs.c: Include internal-fn.h.
3546 (vect_gather_scatter_fn_p): New function.
3547 (vect_describe_gather_scatter_call): Likewise.
3548 (vect_check_gather_scatter): Try using internal functions for
3549 gather loads. Recognize existing calls to a gather load function.
3550 (vect_analyze_data_refs): Consider using gather loads if
3551 supports_vec_gather_load_p.
3552 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
3553 (vect_get_gather_scatter_offset_type): Likewise.
3554 (vect_convert_mask_for_vectype): Likewise.
3555 (vect_add_conversion_to_patterm): Likewise.
3556 (vect_try_gather_scatter_pattern): Likewise.
3557 (vect_recog_gather_scatter_pattern): New pattern recognizer.
3558 (vect_vect_recog_func_ptrs): Add it.
3559 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
3560 internal_fn_mask_index and internal_gather_scatter_fn_p.
3561 (check_load_store_masking): Take the gather_scatter_info as an
3562 argument and handle gather loads.
3563 (vect_get_gather_scatter_ops): New function.
3564 (vectorizable_call): Check internal_load_fn_p.
3565 (vectorizable_load): Likewise. Handle gather load internal
3567 (vectorizable_store): Update call to check_load_store_masking.
3568 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
3569 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
3570 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
3571 (aarch64_gather_scale_operand_d): New predicates.
3572 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
3573 (mask_gather_load<mode>): New insns.
3575 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3576 Alan Hayward <alan.hayward@arm.com>
3577 David Sherwood <david.sherwood@arm.com>
3579 * optabs.def (fold_left_plus_optab): New optab.
3580 * doc/md.texi (fold_left_plus_@var{m}): Document.
3581 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
3582 * internal-fn.c (fold_left_direct): Define.
3583 (expand_fold_left_optab_fn): Likewise.
3584 (direct_fold_left_optab_supported_p): Likewise.
3585 * fold-const-call.c (fold_const_fold_left): New function.
3586 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
3587 * tree-parloops.c (valid_reduction_p): New function.
3588 (gather_scalar_reductions): Use it.
3589 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
3590 (vect_finish_replace_stmt): Declare.
3591 * tree-vect-loop.c (fold_left_reduction_fn): New function.
3592 (needs_fold_left_reduction_p): New function, split out from...
3593 (vect_is_simple_reduction): ...here. Accept reductions that
3594 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
3595 (vect_force_simple_reduction): Also store the reduction type in
3596 the assignment's STMT_VINFO_REDUC_TYPE.
3597 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
3598 (merge_with_identity): New function.
3599 (vect_expand_fold_left): Likewise.
3600 (vectorize_fold_left_reduction): Likewise.
3601 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
3602 scalar phi in place for it. Check for target support and reject
3603 cases that would reassociate the operation. Defer the transform
3604 phase to vectorize_fold_left_reduction.
3605 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
3606 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
3607 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
3609 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3611 * tree-if-conv.c (predicate_mem_writes): Remove redundant
3612 call to ifc_temp_var.
3614 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3615 Alan Hayward <alan.hayward@arm.com>
3616 David Sherwood <david.sherwood@arm.com>
3618 * target.def (legitimize_address_displacement): Take the original
3619 offset as a poly_int.
3620 * targhooks.h (default_legitimize_address_displacement): Update
3622 * targhooks.c (default_legitimize_address_displacement): Likewise.
3623 * doc/tm.texi: Regenerate.
3624 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
3625 as an argument, moving assert of ad->disp == ad->disp_term to...
3626 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
3627 Try calling targetm.legitimize_address_displacement before expanding
3628 the address rather than afterwards, and adjust for the new interface.
3629 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
3630 Match the new hook interface. Handle SVE addresses.
3631 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
3634 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3636 * Makefile.in (OBJS): Add early-remat.o.
3637 * target.def (select_early_remat_modes): New hook.
3638 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
3639 * doc/tm.texi: Regenerate.
3640 * targhooks.h (default_select_early_remat_modes): Declare.
3641 * targhooks.c (default_select_early_remat_modes): New function.
3642 * timevar.def (TV_EARLY_REMAT): New timevar.
3643 * passes.def (pass_early_remat): New pass.
3644 * tree-pass.h (make_pass_early_remat): Declare.
3645 * early-remat.c: New file.
3646 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
3648 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
3650 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3651 Alan Hayward <alan.hayward@arm.com>
3652 David Sherwood <david.sherwood@arm.com>
3654 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
3655 vfm1 with a bound_epilog parameter.
3656 (vect_do_peeling): Update calls accordingly, and move the prologue
3657 call earlier in the function. Treat the base bound_epilog as 0 for
3658 fully-masked loops and retain vf - 1 for other loops. Add 1 to
3659 this base when peeling for gaps.
3660 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
3661 with fully-masked loops.
3662 (vect_estimate_min_profitable_iters): Handle the single peeled
3663 iteration in that case.
3665 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3666 Alan Hayward <alan.hayward@arm.com>
3667 David Sherwood <david.sherwood@arm.com>
3669 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
3670 single-element interleaving even if the size is not a power of 2.
3671 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
3672 accesses for single-element interleaving if the group size is
3675 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3676 Alan Hayward <alan.hayward@arm.com>
3677 David Sherwood <david.sherwood@arm.com>
3679 * doc/md.texi (fold_extract_last_@var{m}): Document.
3680 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
3681 * optabs.def (fold_extract_last_optab): New optab.
3682 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
3683 * internal-fn.c (fold_extract_direct): New macro.
3684 (expand_fold_extract_optab_fn): Likewise.
3685 (direct_fold_extract_optab_supported_p): Likewise.
3686 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
3687 * tree-vect-loop.c (vect_model_reduction_cost): Handle
3688 EXTRACT_LAST_REDUCTION.
3689 (get_initial_def_for_reduction): Do not create an initial vector
3690 for EXTRACT_LAST_REDUCTION reductions.
3691 (vectorizable_reduction): Leave the scalar phi in place for
3692 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
3693 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
3694 epilogue code for EXTRACT_LAST_REDUCTION and defer the
3695 transform phase to vectorizable_condition.
3696 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
3698 (vect_finish_stmt_generation): ...here.
3699 (vect_finish_replace_stmt): New function.
3700 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
3701 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
3703 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
3705 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3706 Alan Hayward <alan.hayward@arm.com>
3707 David Sherwood <david.sherwood@arm.com>
3709 * doc/md.texi (extract_last_@var{m}): Document.
3710 * optabs.def (extract_last_optab): New optab.
3711 * internal-fn.def (EXTRACT_LAST): New internal function.
3712 * internal-fn.c (cond_unary_direct): New macro.
3713 (expand_cond_unary_optab_fn): Likewise.
3714 (direct_cond_unary_optab_supported_p): Likewise.
3715 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
3716 loops using EXTRACT_LAST.
3717 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
3718 (extract_last_<mode>): ...this optab.
3719 (vec_extract<mode><Vel>): Update accordingly.
3721 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3722 Alan Hayward <alan.hayward@arm.com>
3723 David Sherwood <david.sherwood@arm.com>
3725 * target.def (empty_mask_is_expensive): New hook.
3726 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
3727 * doc/tm.texi: Regenerate.
3728 * targhooks.h (default_empty_mask_is_expensive): Declare.
3729 * targhooks.c (default_empty_mask_is_expensive): New function.
3730 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
3731 if the target says that empty masks are expensive.
3732 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
3734 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
3736 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3737 Alan Hayward <alan.hayward@arm.com>
3738 David Sherwood <david.sherwood@arm.com>
3740 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
3741 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
3742 (vect_use_loop_mask_for_alignment_p): New function.
3743 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
3744 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
3745 niters_skip argument. Make sure that the first niters_skip elements
3746 of the first iteration are inactive.
3747 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
3748 Update call to vect_set_loop_masks_directly.
3749 (get_misalign_in_elems): New function, split out from...
3750 (vect_gen_prolog_loop_niters): ...here.
3751 (vect_update_init_of_dr): Take a code argument that specifies whether
3752 the adjustment should be added or subtracted.
3753 (vect_update_init_of_drs): Likewise.
3754 (vect_prepare_for_masked_peels): New function.
3755 (vect_do_peeling): Skip prologue peeling if we're using a mask
3756 instead. Update call to vect_update_inits_of_drs.
3757 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3759 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
3760 alignment. Do not include the number of peeled iterations in
3761 the minimum threshold in that case.
3762 (vectorizable_induction): Adjust the start value down by
3763 LOOP_VINFO_MASK_SKIP_NITERS iterations.
3764 (vect_transform_loop): Call vect_prepare_for_masked_peels.
3765 Take the number of skipped iterations into account when calculating
3767 * tree-vect-stmts.c (vect_gen_while_not): New function.
3769 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3770 Alan Hayward <alan.hayward@arm.com>
3771 David Sherwood <david.sherwood@arm.com>
3773 * doc/sourcebuild.texi (vect_fully_masked): Document.
3774 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
3776 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
3778 (vect_analyze_loop_2): ...here. Don't check the vectorization
3779 factor against the number of loop iterations if the loop is
3782 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3783 Alan Hayward <alan.hayward@arm.com>
3784 David Sherwood <david.sherwood@arm.com>
3786 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
3787 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
3788 (dump_groups): Update accordingly.
3789 (iv_use::mem_type): New member variable.
3790 (address_p): New function.
3791 (record_use): Add a mem_type argument and initialize the new
3793 (record_group_use): Add a mem_type argument. Use address_p.
3794 Remove obsolete null checks of base_object. Update call to record_use.
3795 (find_interesting_uses_op): Update call to record_group_use.
3796 (find_interesting_uses_cond): Likewise.
3797 (find_interesting_uses_address): Likewise.
3798 (get_mem_type_for_internal_fn): New function.
3799 (find_address_like_use): Likewise.
3800 (find_interesting_uses_stmt): Try find_address_like_use before
3801 calling find_interesting_uses_op.
3802 (addr_offset_valid_p): Use the iv mem_type field as the type
3803 of the addressed memory.
3804 (add_autoinc_candidates): Likewise.
3805 (get_address_cost): Likewise.
3806 (split_small_address_groups_p): Use address_p.
3807 (split_address_groups): Likewise.
3808 (add_iv_candidate_for_use): Likewise.
3809 (autoinc_possible_for_pair): Likewise.
3810 (rewrite_groups): Likewise.
3811 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
3812 (determine_group_iv_cost): Update after split of USE_ADDRESS.
3813 (get_alias_ptr_type_for_ptr_address): New function.
3814 (rewrite_use_address): Rewrite address uses in calls that were
3815 identified by find_address_like_use.
3817 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3818 Alan Hayward <alan.hayward@arm.com>
3819 David Sherwood <david.sherwood@arm.com>
3821 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
3823 * gimple-expr.h (is_gimple_addressable: Likewise.
3824 * gimple-expr.c (is_gimple_address): Likewise.
3825 * internal-fn.c (expand_call_mem_ref): New function.
3826 (expand_mask_load_optab_fn): Use it.
3827 (expand_mask_store_optab_fn): Likewise.
3829 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3830 Alan Hayward <alan.hayward@arm.com>
3831 David Sherwood <david.sherwood@arm.com>
3833 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
3834 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
3835 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
3836 (cond_umax@var{mode}): Document.
3837 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
3838 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
3839 (cond_umin_optab, cond_umax_optab): New optabs.
3840 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
3841 (COND_IOR, COND_XOR): New internal functions.
3842 * internal-fn.h (get_conditional_internal_fn): Declare.
3843 * internal-fn.c (cond_binary_direct): New macro.
3844 (expand_cond_binary_optab_fn): Likewise.
3845 (direct_cond_binary_optab_supported_p): Likewise.
3846 (get_conditional_internal_fn): New function.
3847 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
3848 Cope with reduction statements that are vectorized as calls rather
3850 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
3851 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
3852 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
3853 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
3854 (UNSPEC_COND_EOR): New unspecs.
3855 (optab): Add mappings for them.
3856 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
3857 (sve_int_op, sve_fp_op): New int attributes.
3859 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3860 Alan Hayward <alan.hayward@arm.com>
3861 David Sherwood <david.sherwood@arm.com>
3863 * optabs.def (while_ult_optab): New optab.
3864 * doc/md.texi (while_ult@var{m}@var{n}): Document.
3865 * internal-fn.def (WHILE_ULT): New internal function.
3866 * internal-fn.h (direct_internal_fn_supported_p): New override
3867 that takes two types as argument.
3868 * internal-fn.c (while_direct): New macro.
3869 (expand_while_optab_fn): New function.
3870 (convert_optab_supported_p): Likewise.
3871 (direct_while_optab_supported_p): New macro.
3872 * wide-int.h (wi::udiv_ceil): New function.
3873 * tree-vectorizer.h (rgroup_masks): New structure.
3874 (vec_loop_masks): New typedef.
3875 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
3877 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
3878 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
3879 (vect_max_vf): New function.
3880 (slpeel_make_loop_iterate_ntimes): Delete.
3881 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
3882 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
3883 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
3884 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
3885 internal-fn.h, stor-layout.h and optabs-query.h.
3886 (vect_set_loop_mask): New function.
3887 (add_preheader_seq): Likewise.
3888 (add_header_seq): Likewise.
3889 (interleave_supported_p): Likewise.
3890 (vect_maybe_permute_loop_masks): Likewise.
3891 (vect_set_loop_masks_directly): Likewise.
3892 (vect_set_loop_condition_masked): Likewise.
3893 (vect_set_loop_condition_unmasked): New function, split out from
3894 slpeel_make_loop_iterate_ntimes.
3895 (slpeel_make_loop_iterate_ntimes): Rename to..
3896 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
3897 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
3898 (vect_do_peeling): Update call accordingly.
3899 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
3901 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
3902 mask_compare_type, can_fully_mask_p and fully_masked_p.
3903 (release_vec_loop_masks): New function.
3904 (_loop_vec_info): Use it to free the loop masks.
3905 (can_produce_all_loop_masks_p): New function.
3906 (vect_get_max_nscalars_per_iter): Likewise.
3907 (vect_verify_full_masking): Likewise.
3908 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
3909 retries, and free the mask rgroups before retrying. Check loop-wide
3910 reasons for disallowing fully-masked loops. Make the final decision
3911 about whether use a fully-masked loop or not.
3912 (vect_estimate_min_profitable_iters): Do not assume that peeling
3913 for the number of iterations will be needed for fully-masked loops.
3914 (vectorizable_reduction): Disable fully-masked loops.
3915 (vectorizable_live_operation): Likewise.
3916 (vect_halve_mask_nunits): New function.
3917 (vect_double_mask_nunits): Likewise.
3918 (vect_record_loop_mask): Likewise.
3919 (vect_get_loop_mask): Likewise.
3920 (vect_transform_loop): Handle the case in which the final loop
3921 iteration might handle a partial vector. Call vect_set_loop_condition
3922 instead of slpeel_make_loop_iterate_ntimes.
3923 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
3924 (check_load_store_masking): New function.
3925 (prepare_load_store_mask): Likewise.
3926 (vectorizable_store): Handle fully-masked loops.
3927 (vectorizable_load): Likewise.
3928 (supportable_widening_operation): Use vect_halve_mask_nunits for
3930 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
3931 (vect_gen_while): New function.
3932 * config/aarch64/aarch64.md (umax<mode>3): New expander.
3933 (aarch64_uqdec<mode>): New insn.
3935 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3936 Alan Hayward <alan.hayward@arm.com>
3937 David Sherwood <david.sherwood@arm.com>
3939 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
3940 (reduc_xor_scal_optab): New optabs.
3941 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
3942 (reduc_xor_scal_@var{m}): Document.
3943 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
3944 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
3946 * fold-const-call.c (fold_const_call): Handle them.
3947 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
3948 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
3949 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
3950 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
3951 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
3952 (UNSPEC_XORV): New unspecs.
3953 (optab): Add entries for them.
3954 (BITWISEV): New int iterator.
3955 (bit_reduc_op): New int attributes.
3957 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3958 Alan Hayward <alan.hayward@arm.com>
3959 David Sherwood <david.sherwood@arm.com>
3961 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
3962 * internal-fn.def (VEC_SHL_INSERT): New internal function.
3963 * optabs.def (vec_shl_insert_optab): New optab.
3964 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
3965 (duplicate_and_interleave): Likewise.
3966 * tree-vect-loop.c: Include internal-fn.h.
3967 (neutral_op_for_slp_reduction): New function, split out from
3968 get_initial_defs_for_reduction.
3969 (get_initial_def_for_reduction): Handle option 2 for variable-length
3970 vectors by loading the neutral value into a vector and then shifting
3971 the initial value into element 0.
3972 (get_initial_defs_for_reduction): Replace the code argument with
3973 the neutral value calculated by neutral_op_for_slp_reduction.
3974 Use gimple_build_vector for constant-length vectors.
3975 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
3976 but the first group_size elements have a neutral value.
3977 Use duplicate_and_interleave otherwise.
3978 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
3979 Update call to get_initial_defs_for_reduction. Handle SLP
3980 reductions for variable-length vectors by creating one vector
3981 result for each scalar result, with the elements associated
3982 with other scalar results stubbed out with the neutral value.
3983 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
3984 Require IFN_VEC_SHL_INSERT for double reductions on
3985 variable-length vectors, or SLP reductions that have
3986 a neutral value. Require can_duplicate_and_interleave_p
3987 support for variable-length unchained SLP reductions if there
3988 is no neutral value, such as for MIN/MAX reductions. Also require
3989 the number of vector elements to be a multiple of the number of
3990 SLP statements when doing variable-length unchained SLP reductions.
3991 Update call to vect_create_epilog_for_reduction.
3992 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
3993 and remove initial values.
3994 (duplicate_and_interleave): Make public.
3995 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
3996 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
3998 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3999 Alan Hayward <alan.hayward@arm.com>
4000 David Sherwood <david.sherwood@arm.com>
4002 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4003 (can_duplicate_and_interleave_p): New function.
4004 (vect_get_and_check_slp_defs): Take the vector of statements
4005 rather than just the current one. Remove excess parentheses.
4006 Restriction rejectinon of vect_constant_def and vect_external_def
4007 for variable-length vectors to boolean types, or types for which
4008 can_duplicate_and_interleave_p is false.
4009 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4010 (duplicate_and_interleave): New function.
4011 (vect_get_constant_vectors): Use gimple_build_vector for
4012 constant-length vectors and suitable variable-length constant
4013 vectors. Use duplicate_and_interleave for other variable-length
4014 vectors. Don't defer the update when inserting new statements.
4016 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4017 Alan Hayward <alan.hayward@arm.com>
4018 David Sherwood <david.sherwood@arm.com>
4020 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4021 min_profitable_iters doesn't go negative.
4023 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4024 Alan Hayward <alan.hayward@arm.com>
4025 David Sherwood <david.sherwood@arm.com>
4027 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4028 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4029 * optabs.def (vec_mask_load_lanes_optab): New optab.
4030 (vec_mask_store_lanes_optab): Likewise.
4031 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4032 (MASK_STORE_LANES): Likewise.
4033 * internal-fn.c (mask_load_lanes_direct): New macro.
4034 (mask_store_lanes_direct): Likewise.
4035 (expand_mask_load_optab_fn): Handle masked operations.
4036 (expand_mask_load_lanes_optab_fn): New macro.
4037 (expand_mask_store_optab_fn): Handle masked operations.
4038 (expand_mask_store_lanes_optab_fn): New macro.
4039 (direct_mask_load_lanes_optab_supported_p): Likewise.
4040 (direct_mask_store_lanes_optab_supported_p): Likewise.
4041 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4043 (vect_load_lanes_supported): Likewise.
4044 * tree-vect-data-refs.c (strip_conversion): New function.
4045 (can_group_stmts_p): Likewise.
4046 (vect_analyze_data_ref_accesses): Use it instead of checking
4047 for a pair of assignments.
4048 (vect_store_lanes_supported): Take a masked_p parameter.
4049 (vect_load_lanes_supported): Likewise.
4050 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4051 vect_store_lanes_supported and vect_load_lanes_supported.
4052 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4053 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4054 parameter. Don't allow gaps for masked accesses.
4055 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4056 and vect_load_lanes_supported.
4057 (get_load_store_type): Take a masked_p parameter and update
4058 call to get_group_load_store_type.
4059 (vectorizable_store): Update call to get_load_store_type.
4060 Handle IFN_MASK_STORE_LANES.
4061 (vectorizable_load): Update call to get_load_store_type.
4062 Handle IFN_MASK_LOAD_LANES.
4064 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4065 Alan Hayward <alan.hayward@arm.com>
4066 David Sherwood <david.sherwood@arm.com>
4068 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4070 * config/aarch64/aarch64-protos.h
4071 (aarch64_sve_struct_memory_operand_p): Declare.
4072 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4073 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4074 (VPRED, vpred): Handle SVE structure modes.
4075 * config/aarch64/constraints.md (Utx): New constraint.
4076 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4077 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4078 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4079 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4080 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4081 structure modes. Split into pieces after RA.
4082 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4083 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4085 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4086 SVE structure modes.
4087 (aarch64_classify_address): Likewise.
4088 (sizetochar): Move earlier in file.
4089 (aarch64_print_operand): Handle SVE register lists.
4090 (aarch64_array_mode): New function.
4091 (aarch64_sve_struct_memory_operand_p): Likewise.
4092 (TARGET_ARRAY_MODE): Redefine.
4094 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4095 Alan Hayward <alan.hayward@arm.com>
4096 David Sherwood <david.sherwood@arm.com>
4098 * target.def (array_mode): New target hook.
4099 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4100 * doc/tm.texi: Regenerate.
4101 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4102 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4103 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4105 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4108 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4109 Alan Hayward <alan.hayward@arm.com>
4110 David Sherwood <david.sherwood@arm.com>
4112 * fold-const.c (fold_binary_loc): Check the argument types
4113 rather than the result type when testing for a vector operation.
4115 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4117 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4118 * doc/tm.texi: Regenerate.
4120 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4121 Alan Hayward <alan.hayward@arm.com>
4122 David Sherwood <david.sherwood@arm.com>
4124 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4125 (sve): Document new AArch64 extension.
4126 * doc/md.texi (w): Extend the description of the AArch64
4127 constraint to include SVE vectors.
4128 (Upl, Upa): Document new AArch64 predicate constraints.
4129 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4131 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4132 (msve-vector-bits=): New option.
4133 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4134 SVE when these are disabled.
4135 (sve): New extension.
4136 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4137 modes. Adjust their number of units based on aarch64_sve_vg.
4138 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4139 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4140 aarch64_addr_query_type.
4141 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4142 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4143 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4144 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4145 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4146 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4147 (aarch64_simd_imm_zero_p): Delete.
4148 (aarch64_check_zero_based_sve_index_immediate): Declare.
4149 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4150 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4151 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4152 (aarch64_sve_float_mul_immediate_p): Likewise.
4153 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4155 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4156 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4157 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4158 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4159 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4160 (aarch64_regmode_natural_size): Likewise.
4161 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4162 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4164 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4165 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4166 for VG and the SVE predicate registers.
4167 (V_ALIASES): Add a "z"-prefixed alias.
4168 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4169 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4170 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4171 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4172 (REG_CLASS_NAMES): Add entries for them.
4173 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4174 and the predicate registers.
4175 (aarch64_sve_vg): Declare.
4176 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4177 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4178 (REGMODE_NATURAL_SIZE): Define.
4179 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4181 * config/aarch64/aarch64.c: Include cfgrtl.h.
4182 (simd_immediate_info): Add a constructor for series vectors,
4183 and an associated step field.
4184 (aarch64_sve_vg): New variable.
4185 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4186 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4187 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4188 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4189 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4190 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4191 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4192 (aarch64_get_mask_mode): New functions.
4193 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4194 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4195 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4196 predicate modes and predicate registers. Explicitly restrict
4197 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4198 to store a vector mode if it is recognized by
4199 aarch64_classify_vector_mode.
4200 (aarch64_regmode_natural_size): New function.
4201 (aarch64_hard_regno_caller_save_mode): Return the original mode
4203 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4204 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4205 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4206 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4208 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4209 does not overlap dest if the function is frame-related. Handle
4211 (aarch64_split_add_offset): New function.
4212 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4213 them aarch64_add_offset.
4214 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4215 and update call to aarch64_sub_sp.
4216 (aarch64_add_cfa_expression): New function.
4217 (aarch64_expand_prologue): Pass extra temporary registers to the
4218 functions above. Handle the case in which we need to emit new
4219 DW_CFA_expressions for registers that were originally saved
4220 relative to the stack pointer, but now have to be expressed
4221 relative to the frame pointer.
4222 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4224 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4225 IP0 and IP1 values for SVE frames.
4226 (aarch64_expand_vec_series): New function.
4227 (aarch64_expand_sve_widened_duplicate): Likewise.
4228 (aarch64_expand_sve_const_vector): Likewise.
4229 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4230 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4231 into the register, rather than emitting a SET directly.
4232 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4233 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4234 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4235 (offset_9bit_signed_scaled_p): New functions.
4236 (aarch64_replicate_bitmask_imm): New function.
4237 (aarch64_bitmask_imm): Use it.
4238 (aarch64_cannot_force_const_mem): Reject expressions involving
4239 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4240 (aarch64_classify_index): Handle SVE indices, by requiring
4241 a plain register index with a scale that matches the element size.
4242 (aarch64_classify_address): Handle SVE addresses. Assert that
4243 the mode of the address is VOIDmode or an integer mode.
4244 Update call to aarch64_classify_symbol.
4245 (aarch64_classify_symbolic_expression): Update call to
4246 aarch64_classify_symbol.
4247 (aarch64_const_vec_all_in_range_p): New function.
4248 (aarch64_print_vector_float_operand): Likewise.
4249 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4250 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4251 and the FP immediates 1.0 and 0.5.
4252 (aarch64_print_address_internal): Handle SVE addresses.
4253 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4254 (aarch64_regno_regclass): Handle predicate registers.
4255 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4257 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4258 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4259 (aarch64_convert_sve_vector_bits): New function.
4260 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4261 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4263 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4264 Handle SVE vector and predicate modes. Accept VL-based constants
4265 that need only one temporary register, and VL offsets that require
4266 no temporary registers.
4267 (aarch64_conditional_register_usage): Mark the predicate registers
4268 as fixed if SVE isn't available.
4269 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4270 Return true for SVE vector and predicate modes.
4271 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4272 rather than an unsigned int. Handle SVE modes.
4273 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4275 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4277 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4278 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4279 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4280 (aarch64_sve_float_mul_immediate_p): New functions.
4281 (aarch64_sve_valid_immediate): New function.
4282 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4283 Explicitly reject structure modes. Check for INDEX constants.
4284 Handle PTRUE and PFALSE constants.
4285 (aarch64_check_zero_based_sve_index_immediate): New function.
4286 (aarch64_simd_imm_zero_p): Delete.
4287 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4288 vector modes. Accept constants in the range of CNT[BHWD].
4289 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4290 ask for an Advanced SIMD mode.
4291 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4292 (aarch64_simd_vector_alignment): Handle SVE predicates.
4293 (aarch64_vectorize_preferred_vector_alignment): New function.
4294 (aarch64_simd_vector_alignment_reachable): Use it instead of
4296 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4297 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4299 (MAX_VECT_LEN): Delete.
4300 (expand_vec_perm_d): Add a vec_flags field.
4301 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4302 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4303 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4305 (aarch64_evpc_rev): Rename to...
4306 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4307 (aarch64_evpc_rev_global): New function.
4308 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4309 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4311 (aarch64_evpc_sve_tbl): New function.
4312 (aarch64_expand_vec_perm_const_1): Update after rename of
4313 aarch64_evpc_rev. Handle SVE permutes too, trying
4314 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4315 than aarch64_evpc_tbl.
4316 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4317 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4318 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4319 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4320 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4321 (aarch64_expand_sve_vcond): New functions.
4322 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4323 of aarch64_vector_mode_p.
4324 (aarch64_dwarf_poly_indeterminate_value): New function.
4325 (aarch64_compute_pressure_classes): Likewise.
4326 (aarch64_can_change_mode_class): Likewise.
4327 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4328 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4329 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4330 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4331 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4332 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4333 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4334 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4336 (Dn, Dl, Dr): Accept const as well as const_vector.
4337 (Dz): Likewise. Compare against CONST0_RTX.
4338 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4339 of "vector" where appropriate.
4340 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4341 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4342 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4343 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4344 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4345 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4346 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4347 (v_int_equiv): Extend to SVE modes.
4348 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4350 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4351 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4352 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4353 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4354 (SVE_COND_FP_CMP): New int iterators.
4355 (perm_hilo): Handle the new unpack unspecs.
4356 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4358 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4359 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4360 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4361 (aarch64_equality_operator, aarch64_constant_vector_operand)
4362 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4363 (aarch64_sve_nonimmediate_operand): Likewise.
4364 (aarch64_sve_general_operand): Likewise.
4365 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4366 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4367 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4368 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4369 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4370 (aarch64_sve_float_arith_immediate): Likewise.
4371 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4372 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4373 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4374 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4375 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4376 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4377 (aarch64_sve_float_arith_operand): Likewise.
4378 (aarch64_sve_float_arith_with_sub_operand): Likewise.
4379 (aarch64_sve_float_mul_operand): Likewise.
4380 (aarch64_sve_vec_perm_operand): Likewise.
4381 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
4382 (aarch64_mov_operand): Accept const_poly_int and const_vector.
4383 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
4384 as well as const_vector.
4385 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
4386 in file. Use CONST0_RTX and CONSTM1_RTX.
4387 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
4388 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
4389 Use aarch64_simd_imm_zero.
4390 * config/aarch64/aarch64-sve.md: New file.
4391 * config/aarch64/aarch64.md: Include it.
4392 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
4393 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
4394 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
4395 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
4396 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
4397 (sve): New attribute.
4398 (enabled): Disable instructions with the sve attribute unless
4400 (movqi, movhi): Pass CONST_POLY_INT operaneds through
4401 aarch64_expand_mov_immediate.
4402 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
4403 CNT[BHSD] immediates.
4404 (movti): Split CONST_POLY_INT moves into two halves.
4405 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
4406 Split additions that need a temporary here if the destination
4407 is the stack pointer.
4408 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
4409 (*add<mode>3_poly_1): New instruction.
4410 (set_clobber_cc): New expander.
4412 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4414 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
4415 parameter and use it instead of GET_MODE_SIZE (innermode). Use
4416 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
4417 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
4418 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
4419 Change innermode from fixed_mode_size to machine_mode.
4420 (simplify_subreg): Update call accordingly. Handle a constant-sized
4421 subreg of a variable-length CONST_VECTOR.
4423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4424 Alan Hayward <alan.hayward@arm.com>
4425 David Sherwood <david.sherwood@arm.com>
4427 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
4428 (add_offset_to_base): New function, split out from...
4429 (create_mem_ref): ...here. When handling a scale other than 1,
4430 check first whether the address is valid without the offset.
4431 Add it into the base if so, leaving the index and scale as-is.
4433 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4436 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
4437 fold_for_warn before checking if arg2 is INTEGER_CST.
4439 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
4441 * config/rs6000/predicates.md (load_multiple_operation): Delete.
4442 (store_multiple_operation): Delete.
4443 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
4444 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
4445 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
4446 guarded by TARGET_STRING.
4447 (rs6000_output_load_multiple): Delete.
4448 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
4449 OPTION_MASK_STRING / TARGET_STRING handling.
4450 (print_operand) <'N', 'O'>: Add comment that these are unused now.
4451 (const rs6000_opt_masks) <"string">: Change mask to 0.
4452 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
4453 (MASK_STRING): Delete.
4454 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
4456 (load_multiple): Delete.
4463 (store_multiple): Delete.
4470 (movmemsi_8reg): Delete.
4471 (corresponding unnamed define_insn): Delete.
4472 (movmemsi_6reg): Delete.
4473 (corresponding unnamed define_insn): Delete.
4474 (movmemsi_4reg): Delete.
4475 (corresponding unnamed define_insn): Delete.
4476 (movmemsi_2reg): Delete.
4477 (corresponding unnamed define_insn): Delete.
4478 (movmemsi_1reg): Delete.
4479 (corresponding unnamed define_insn): Delete.
4480 * config/rs6000/rs6000.opt (mno-string): New.
4481 (mstring): Replace by deprecation warning stub.
4482 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
4484 2018-01-12 Jakub Jelinek <jakub@redhat.com>
4486 * regrename.c (regrename_do_replace): If replacing the same
4487 reg multiple times, try to reuse last created gen_raw_REG.
4490 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
4491 main to workaround a bug in GDB.
4493 2018-01-12 Tom de Vries <tom@codesourcery.com>
4496 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
4498 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
4500 PR rtl-optimization/80481
4501 * ira-color.c (get_cap_member): New function.
4502 (allocnos_conflict_by_live_ranges_p): Use it.
4503 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
4504 (setup_slot_coalesced_allocno_live_ranges): Ditto.
4506 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
4509 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
4510 (*saddl_se_1): Ditto.
4512 (*ssubl_se_1): Ditto.
4514 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4516 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
4517 rather than wi::to_widest for DR_INITs.
4518 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
4519 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
4520 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
4522 (vect_analyze_group_access_1): Note that here.
4524 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4526 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
4527 polynomial type sizes.
4529 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
4531 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
4532 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
4533 (gimple_add_tmp_var): Likewise.
4535 2018-01-12 Martin Liska <mliska@suse.cz>
4537 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
4538 (gimple_alloc_sizes): Likewise.
4539 (dump_gimple_statistics): Use PRIu64 in printf format.
4540 * gimple.h: Change uint64_t to int.
4542 2018-01-12 Martin Liska <mliska@suse.cz>
4544 * tree-core.h: Use uint64_t instead of int.
4545 * tree.c (tree_node_counts): Likewise.
4546 (tree_node_sizes): Likewise.
4547 (dump_tree_statistics): Use PRIu64 in printf format.
4549 2018-01-12 Martin Liska <mliska@suse.cz>
4551 * Makefile.in: As qsort_chk is implemented in vec.c, add
4552 vec.o to linkage of gencfn-macros.
4553 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
4554 passing the info to record_node_allocation_statistics.
4555 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
4557 * ggc-common.c (struct ggc_usage): Add operator== and use
4558 it in operator< and compare function.
4559 * mem-stats.h (struct mem_usage): Likewise.
4560 * vec.c (struct vec_usage): Remove operator< and compare
4561 function. Can be simply inherited.
4563 2018-01-12 Martin Jambor <mjambor@suse.cz>
4566 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
4567 * tree-ssa-math-opts.c: Include domwalk.h.
4568 (convert_mult_to_fma_1): New function.
4569 (fma_transformation_info): New type.
4570 (fma_deferring_state): Likewise.
4571 (cancel_fma_deferring): New function.
4572 (result_of_phi): Likewise.
4573 (last_fma_candidate_feeds_initial_phi): Likewise.
4574 (convert_mult_to_fma): Added deferring logic, split actual
4575 transformation to convert_mult_to_fma_1.
4576 (math_opts_dom_walker): New type.
4577 (math_opts_dom_walker::after_dom_children): New method, body moved
4578 here from pass_optimize_widening_mul::execute, added deferring logic
4580 (pass_optimize_widening_mul::execute): Moved most of code to
4581 math_opts_dom_walker::after_dom_children.
4582 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
4583 * config/i386/i386.c (ix86_option_override_internal): Added
4584 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
4586 2018-01-12 Richard Biener <rguenther@suse.de>
4589 * dwarf2out.c (gen_variable_die): Do not reset old_die for
4590 inline instance vars.
4592 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
4595 * config/rx/rx.c (rx_is_restricted_memory_address):
4598 2018-01-12 Richard Biener <rguenther@suse.de>
4600 PR tree-optimization/80846
4601 * target.def (split_reduction): New target hook.
4602 * targhooks.c (default_split_reduction): New function.
4603 * targhooks.h (default_split_reduction): Declare.
4604 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
4605 target requests first reduce vectors by combining low and high
4607 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
4608 (get_vectype_for_scalar_type_and_size): Export.
4609 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
4610 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
4611 * doc/tm.texi: Regenerate.
4612 * config/i386/i386.c (ix86_split_reduction): Implement
4613 TARGET_VECTORIZE_SPLIT_REDUCTION.
4615 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4618 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
4619 in PIC mode except for TARGET_VXWORKS_RTP.
4620 * config/sparc/sparc.c: Include cfgrtl.h.
4621 (TARGET_INIT_PIC_REG): Define.
4622 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
4623 (sparc_pic_register_p): New predicate.
4624 (sparc_legitimate_address_p): Use it.
4625 (sparc_legitimize_pic_address): Likewise.
4626 (sparc_delegitimize_address): Likewise.
4627 (sparc_mode_dependent_address_p): Likewise.
4628 (gen_load_pcrel_sym): Remove 4th parameter.
4629 (load_got_register): Adjust call to above. Remove obsolete stuff.
4630 (sparc_expand_prologue): Do not call load_got_register here.
4631 (sparc_flat_expand_prologue): Likewise.
4632 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
4633 (sparc_use_pseudo_pic_reg): New function.
4634 (sparc_init_pic_reg): Likewise.
4635 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4636 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
4638 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
4640 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
4641 Add item for branch_cost.
4643 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4645 PR rtl-optimization/83565
4646 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
4647 not extend the result to a larger mode for rotate operations.
4648 (num_sign_bit_copies1): Likewise.
4650 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4653 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
4655 Use values-Xc.o for -pedantic.
4656 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
4658 2018-01-12 Martin Liska <mliska@suse.cz>
4661 * ipa-devirt.c (final_warning_record::grow_type_warnings):
4663 (possible_polymorphic_call_targets): Use it.
4664 (ipa_devirt): Likewise.
4666 2018-01-12 Martin Liska <mliska@suse.cz>
4668 * profile-count.h (enum profile_quality): Use 0 as invalid
4669 enum value of profile_quality.
4671 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
4673 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
4674 -mext-string options.
4676 2018-01-12 Richard Biener <rguenther@suse.de>
4678 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
4679 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
4680 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
4682 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
4684 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4686 * configure.ac (--with-long-double-format): Add support for the
4687 configuration option to change the default long double format on
4689 * config.gcc (powerpc*-linux*-*): Likewise.
4690 * configure: Regenerate.
4691 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
4692 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
4693 used without modification.
4695 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4697 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
4698 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
4699 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
4700 MISC_BUILTIN_SPEC_BARRIER.
4701 (rs6000_init_builtins): Likewise.
4702 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
4704 (speculation_barrier): New define_insn.
4705 * doc/extend.texi: Document __builtin_speculation_barrier.
4707 2018-01-11 Jakub Jelinek <jakub@redhat.com>
4710 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
4711 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
4712 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
4714 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
4715 integral modes instead of "ss" and "sd".
4716 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
4717 vectors with 32-bit and 64-bit elements.
4718 (vecdupssescalarmodesuffix): New mode attribute.
4719 (vec_dup<mode>): Use it.
4721 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
4724 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
4725 frame if argument is passed on stack.
4727 2018-01-11 Jakub Jelinek <jakub@redhat.com>
4730 * ree.c (combine_reaching_defs): Optimize also
4731 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
4732 reg2=any_extend(exp); reg1=reg2;, formatting fix.
4734 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
4737 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
4739 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
4742 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
4743 after they are computed.
4745 2018-01-11 Bin Cheng <bin.cheng@arm.com>
4747 PR tree-optimization/83695
4748 * gimple-loop-linterchange.cc
4749 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
4750 reset cached scev information after interchange.
4751 (pass_linterchange::execute): Remove call to scev_reset_htab.
4753 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4755 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
4756 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
4757 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
4758 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
4759 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
4760 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
4761 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
4762 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
4763 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
4764 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
4765 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
4766 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
4767 (V_lane_reg): Likewise.
4768 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
4770 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
4771 (vfmal_lane_low<mode>_intrinsic,
4772 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
4773 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
4774 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
4775 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
4776 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
4777 vfmsl_lane_high<mode>_intrinsic): New define_insns.
4779 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4781 * config/arm/arm-cpus.in (fp16fml): New feature.
4782 (ALL_SIMD): Add fp16fml.
4783 (armv8.2-a): Add fp16fml as an option.
4784 (armv8.3-a): Likewise.
4785 (armv8.4-a): Add fp16fml as part of fp16.
4786 * config/arm/arm.h (TARGET_FP16FML): Define.
4787 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
4789 * config/arm/arm-modes.def (V2HF): Define.
4790 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
4791 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
4792 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
4793 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
4794 vfmsl_low, vfmsl_high): New set of builtins.
4795 * config/arm/iterators.md (PLUSMINUS): New code iterator.
4796 (vfml_op): New code attribute.
4797 (VFMLHALVES): New int iterator.
4798 (VFML, VFMLSEL): New mode attributes.
4799 (V_reg): Define mapping for V2HF.
4800 (V_hi, V_lo): New mode attributes.
4801 (VF_constraint): Likewise.
4802 (vfml_half, vfml_half_selector): New int attributes.
4803 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
4805 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
4806 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
4808 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
4809 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
4810 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
4811 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
4813 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
4814 Document new effective target and option set.
4816 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4818 * config/arm/arm-cpus.in (armv8_4): New feature.
4819 (ARMv8_4a): New fgroup.
4820 (armv8.4-a): New arch.
4821 * config/arm/arm-tables.opt: Regenerate.
4822 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
4823 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
4824 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
4825 Add matching rules for -march=armv8.4-a and extensions.
4826 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
4828 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
4831 * config/rx/rx.md (BW): New mode attribute.
4832 (sync_lock_test_and_setsi): Add mode suffix to insn output.
4834 2018-01-11 Richard Biener <rguenther@suse.de>
4836 PR tree-optimization/83435
4837 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
4838 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
4839 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
4841 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4842 Alan Hayward <alan.hayward@arm.com>
4843 David Sherwood <david.sherwood@arm.com>
4845 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
4847 (aarch64_classify_address): Initialize it. Track polynomial offsets.
4848 (aarch64_print_address_internal): Use it to check for a zero offset.
4850 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4851 Alan Hayward <alan.hayward@arm.com>
4852 David Sherwood <david.sherwood@arm.com>
4854 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
4855 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
4856 Return a poly_int64 rather than a HOST_WIDE_INT.
4857 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
4858 rather than a HOST_WIDE_INT.
4859 * config/aarch64/aarch64.h (aarch64_frame): Protect with
4860 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
4861 hard_fp_offset, frame_size, initial_adjust, callee_offset and
4862 final_offset from HOST_WIDE_INT to poly_int64.
4863 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
4864 to_constant when getting the number of units in an Advanced SIMD
4866 (aarch64_builtin_vectorized_function): Check for a constant number
4868 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
4870 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
4871 attribute instead of GET_MODE_NUNITS.
4872 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
4873 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
4874 GET_MODE_SIZE for fixed-size registers.
4875 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
4876 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
4877 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
4878 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
4879 (aarch64_print_operand, aarch64_print_address_internal)
4880 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
4881 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
4882 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
4883 Handle polynomial GET_MODE_SIZE.
4884 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
4885 wider than SImode without modification.
4886 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
4887 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
4888 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
4889 passing and returning SVE modes.
4890 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
4891 rather than GEN_INT.
4892 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
4893 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
4894 (aarch64_allocate_and_probe_stack_space): Likewise.
4895 (aarch64_layout_frame): Cope with polynomial offsets.
4896 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
4897 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
4899 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
4900 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
4901 poly_int64 rather than a HOST_WIDE_INT.
4902 (aarch64_get_separate_components, aarch64_process_components)
4903 (aarch64_expand_prologue, aarch64_expand_epilogue)
4904 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
4905 (aarch64_anchor_offset): New function, split out from...
4906 (aarch64_legitimize_address): ...here.
4907 (aarch64_builtin_vectorization_cost): Handle polynomial
4908 TYPE_VECTOR_SUBPARTS.
4909 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
4911 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
4912 number of elements from the PARALLEL rather than the mode.
4913 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
4914 rather than GET_MODE_BITSIZE.
4915 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
4916 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
4917 (aarch64_expand_vec_perm_const_1): Handle polynomial
4918 d->perm.length () and d->perm elements.
4919 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
4920 Apply to_constant to d->perm elements.
4921 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
4922 polynomial CONST_VECTOR_NUNITS.
4923 (aarch64_move_pointer): Take amount as a poly_int64 rather
4925 (aarch64_progress_pointer): Avoid temporary variable.
4926 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
4927 the mode attribute instead of GET_MODE.
4929 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4930 Alan Hayward <alan.hayward@arm.com>
4931 David Sherwood <david.sherwood@arm.com>
4933 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
4934 x exists before using it.
4935 (aarch64_add_constant_internal): Rename to...
4936 (aarch64_add_offset_1): ...this. Replace regnum with separate
4937 src and dest rtxes. Handle the case in which they're different,
4938 including when the offset is zero. Replace scratchreg with an rtx.
4939 Use 2 additions if there is no spare register into which we can
4940 move a 16-bit constant.
4941 (aarch64_add_constant): Delete.
4942 (aarch64_add_offset): Replace reg with separate src and dest
4943 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
4944 Use aarch64_add_offset_1.
4945 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
4946 an rtx rather than an int. Take the delta as a poly_int64
4947 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
4948 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
4949 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
4950 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
4951 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
4953 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
4954 aarch64_add_constant.
4956 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4958 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
4959 Use scalar_float_mode.
4961 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4963 * config/aarch64/aarch64-simd.md
4964 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
4965 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
4966 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
4967 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
4968 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
4969 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
4970 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
4971 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
4972 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
4973 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
4975 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4978 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
4979 targ_options->x_arm_arch_string is non NULL.
4981 2018-01-11 Tamar Christina <tamar.christina@arm.com>
4983 * config/aarch64/aarch64.h
4984 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
4986 2018-01-11 Sudakshina Das <sudi.das@arm.com>
4989 * expmed.c (emit_store_flag_force): Swap if const op0
4990 and change VOIDmode to mode of op0.
4992 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
4994 PR rtl-optimization/83761
4995 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
4996 than bytes to mode_for_size.
4998 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5001 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5002 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5005 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5008 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5009 when in layout mode.
5010 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5011 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5014 2018-01-10 Michael Collison <michael.collison@arm.com>
5016 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5017 * config/aarch64/aarch64-option-extension.def: Add
5018 AARCH64_OPT_EXTENSION of 'fp16fml'.
5019 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5020 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5021 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5022 * config/aarch64/constraints.md (Ui7): New constraint.
5023 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5024 (VFMLA_SEL_W): Ditto.
5027 (VFMLA16_LOW): New int iterator.
5028 (VFMLA16_HIGH): Ditto.
5029 (UNSPEC_FMLAL): New unspec.
5030 (UNSPEC_FMLSL): Ditto.
5031 (UNSPEC_FMLAL2): Ditto.
5032 (UNSPEC_FMLSL2): Ditto.
5033 (f16mac): New code attribute.
5034 * config/aarch64/aarch64-simd-builtins.def
5035 (aarch64_fmlal_lowv2sf): Ditto.
5036 (aarch64_fmlsl_lowv2sf): Ditto.
5037 (aarch64_fmlalq_lowv4sf): Ditto.
5038 (aarch64_fmlslq_lowv4sf): Ditto.
5039 (aarch64_fmlal_highv2sf): Ditto.
5040 (aarch64_fmlsl_highv2sf): Ditto.
5041 (aarch64_fmlalq_highv4sf): Ditto.
5042 (aarch64_fmlslq_highv4sf): Ditto.
5043 (aarch64_fmlal_lane_lowv2sf): Ditto.
5044 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5045 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5046 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5047 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5048 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5049 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5050 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5051 (aarch64_fmlal_lane_highv2sf): Ditto.
5052 (aarch64_fmlsl_lane_highv2sf): Ditto.
5053 (aarch64_fmlal_laneq_highv2sf): Ditto.
5054 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5055 (aarch64_fmlalq_lane_highv4sf): Ditto.
5056 (aarch64_fmlsl_lane_highv4sf): Ditto.
5057 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5058 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5059 * config/aarch64/aarch64-simd.md:
5060 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5061 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5062 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5063 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5064 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5065 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5066 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5067 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5068 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5069 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5070 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5071 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5072 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5073 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5074 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5075 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5076 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5077 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5078 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5079 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5080 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5081 (vfmlsl_low_u32): Ditto.
5082 (vfmlalq_low_u32): Ditto.
5083 (vfmlslq_low_u32): Ditto.
5084 (vfmlal_high_u32): Ditto.
5085 (vfmlsl_high_u32): Ditto.
5086 (vfmlalq_high_u32): Ditto.
5087 (vfmlslq_high_u32): Ditto.
5088 (vfmlal_lane_low_u32): Ditto.
5089 (vfmlsl_lane_low_u32): Ditto.
5090 (vfmlal_laneq_low_u32): Ditto.
5091 (vfmlsl_laneq_low_u32): Ditto.
5092 (vfmlalq_lane_low_u32): Ditto.
5093 (vfmlslq_lane_low_u32): Ditto.
5094 (vfmlalq_laneq_low_u32): Ditto.
5095 (vfmlslq_laneq_low_u32): Ditto.
5096 (vfmlal_lane_high_u32): Ditto.
5097 (vfmlsl_lane_high_u32): Ditto.
5098 (vfmlal_laneq_high_u32): Ditto.
5099 (vfmlsl_laneq_high_u32): Ditto.
5100 (vfmlalq_lane_high_u32): Ditto.
5101 (vfmlslq_lane_high_u32): Ditto.
5102 (vfmlalq_laneq_high_u32): Ditto.
5103 (vfmlslq_laneq_high_u32): Ditto.
5104 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5105 (AARCH64_FL_FOR_ARCH8_4): New.
5106 (AARCH64_ISA_F16FML): New ISA flag.
5107 (TARGET_F16FML): New feature flag for fp16fml.
5108 (doc/invoke.texi): Document new fp16fml option.
5110 2018-01-10 Michael Collison <michael.collison@arm.com>
5112 * config/aarch64/aarch64-builtins.c:
5113 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5114 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5115 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5116 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5117 (AARCH64_ISA_SHA3): New ISA flag.
5118 (TARGET_SHA3): New feature flag for sha3.
5119 * config/aarch64/iterators.md (sha512_op): New int attribute.
5120 (CRYPTO_SHA512): New int iterator.
5121 (UNSPEC_SHA512H): New unspec.
5122 (UNSPEC_SHA512H2): Ditto.
5123 (UNSPEC_SHA512SU0): Ditto.
5124 (UNSPEC_SHA512SU1): Ditto.
5125 * config/aarch64/aarch64-simd-builtins.def
5126 (aarch64_crypto_sha512hqv2di): New builtin.
5127 (aarch64_crypto_sha512h2qv2di): Ditto.
5128 (aarch64_crypto_sha512su0qv2di): Ditto.
5129 (aarch64_crypto_sha512su1qv2di): Ditto.
5130 (aarch64_eor3qv8hi): Ditto.
5131 (aarch64_rax1qv2di): Ditto.
5132 (aarch64_xarqv2di): Ditto.
5133 (aarch64_bcaxqv8hi): Ditto.
5134 * config/aarch64/aarch64-simd.md:
5135 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5136 (aarch64_crypto_sha512su0qv2di): Ditto.
5137 (aarch64_crypto_sha512su1qv2di): Ditto.
5138 (aarch64_eor3qv8hi): Ditto.
5139 (aarch64_rax1qv2di): Ditto.
5140 (aarch64_xarqv2di): Ditto.
5141 (aarch64_bcaxqv8hi): Ditto.
5142 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5143 (vsha512h2q_u64): Ditto.
5144 (vsha512su0q_u64): Ditto.
5145 (vsha512su1q_u64): Ditto.
5146 (veor3q_u16): Ditto.
5147 (vrax1q_u64): Ditto.
5149 (vbcaxq_u16): Ditto.
5150 * config/arm/types.md (crypto_sha512): New type attribute.
5151 (crypto_sha3): Ditto.
5152 (doc/invoke.texi): Document new sha3 option.
5154 2018-01-10 Michael Collison <michael.collison@arm.com>
5156 * config/aarch64/aarch64-builtins.c:
5157 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5158 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5159 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5160 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5161 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5162 (AARCH64_ISA_SM4): New ISA flag.
5163 (TARGET_SM4): New feature flag for sm4.
5164 * config/aarch64/aarch64-simd-builtins.def
5165 (aarch64_sm3ss1qv4si): Ditto.
5166 (aarch64_sm3tt1aq4si): Ditto.
5167 (aarch64_sm3tt1bq4si): Ditto.
5168 (aarch64_sm3tt2aq4si): Ditto.
5169 (aarch64_sm3tt2bq4si): Ditto.
5170 (aarch64_sm3partw1qv4si): Ditto.
5171 (aarch64_sm3partw2qv4si): Ditto.
5172 (aarch64_sm4eqv4si): Ditto.
5173 (aarch64_sm4ekeyqv4si): Ditto.
5174 * config/aarch64/aarch64-simd.md:
5175 (aarch64_sm3ss1qv4si): Ditto.
5176 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5177 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5178 (aarch64_sm4eqv4si): Ditto.
5179 (aarch64_sm4ekeyqv4si): Ditto.
5180 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5181 (sm3part_op): Ditto.
5182 (CRYPTO_SM3TT): Ditto.
5183 (CRYPTO_SM3PART): Ditto.
5184 (UNSPEC_SM3SS1): New unspec.
5185 (UNSPEC_SM3TT1A): Ditto.
5186 (UNSPEC_SM3TT1B): Ditto.
5187 (UNSPEC_SM3TT2A): Ditto.
5188 (UNSPEC_SM3TT2B): Ditto.
5189 (UNSPEC_SM3PARTW1): Ditto.
5190 (UNSPEC_SM3PARTW2): Ditto.
5191 (UNSPEC_SM4E): Ditto.
5192 (UNSPEC_SM4EKEY): Ditto.
5193 * config/aarch64/constraints.md (Ui2): New constraint.
5194 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5195 * config/arm/types.md (crypto_sm3): New type attribute.
5196 (crypto_sm4): Ditto.
5197 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5198 (vsm3tt1aq_u32): Ditto.
5199 (vsm3tt1bq_u32): Ditto.
5200 (vsm3tt2aq_u32): Ditto.
5201 (vsm3tt2bq_u32): Ditto.
5202 (vsm3partw1q_u32): Ditto.
5203 (vsm3partw2q_u32): Ditto.
5204 (vsm4eq_u32): Ditto.
5205 (vsm4ekeyq_u32): Ditto.
5206 (doc/invoke.texi): Document new sm4 option.
5208 2018-01-10 Michael Collison <michael.collison@arm.com>
5210 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5211 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5212 (AARCH64_FL_FOR_ARCH8_4): New.
5213 (AARCH64_FL_V8_4): New flag.
5214 (doc/invoke.texi): Document new armv8.4-a option.
5216 2018-01-10 Michael Collison <michael.collison@arm.com>
5218 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5219 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5220 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5221 * config/aarch64/aarch64-option-extension.def: Add
5222 AARCH64_OPT_EXTENSION of 'sha2'.
5223 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5224 (crypto): Disable sha2 and aes if crypto disabled.
5225 (crypto): Enable aes and sha2 if enabled.
5226 (simd): Disable sha2 and aes if simd disabled.
5227 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5229 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5230 (TARGET_SHA2): New feature flag for sha2.
5231 (TARGET_AES): New feature flag for aes.
5232 * config/aarch64/aarch64-simd.md:
5233 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5234 conditional on TARGET_AES.
5235 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5236 (aarch64_crypto_sha1hsi): Make pattern conditional
5238 (aarch64_crypto_sha1hv4si): Ditto.
5239 (aarch64_be_crypto_sha1hv4si): Ditto.
5240 (aarch64_crypto_sha1su1v4si): Ditto.
5241 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5242 (aarch64_crypto_sha1su0v4si): Ditto.
5243 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5244 (aarch64_crypto_sha256su0v4si): Ditto.
5245 (aarch64_crypto_sha256su1v4si): Ditto.
5246 (doc/invoke.texi): Document new aes and sha2 options.
5248 2018-01-10 Martin Sebor <msebor@redhat.com>
5250 PR tree-optimization/83781
5251 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5254 2018-01-11 Martin Sebor <msebor@gmail.com>
5255 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5257 PR tree-optimization/83501
5258 PR tree-optimization/81703
5260 * tree-ssa-strlen.c (get_string_cst): Rename...
5261 (get_string_len): ...to this. Handle global constants.
5262 (handle_char_store): Adjust.
5264 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5265 Jim Wilson <jimw@sifive.com>
5267 * config/riscv/riscv-protos.h (riscv_output_return): New.
5268 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5269 (riscv_attribute_table, riscv_output_return),
5270 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5271 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5272 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5274 (riscv_expand_prologue): Add early return for naked function.
5275 (riscv_expand_epilogue): Likewise.
5276 (riscv_function_ok_for_sibcall): Return false for naked function.
5277 (riscv_set_current_function): New.
5278 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5279 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5280 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5281 * doc/extend.texi (RISC-V Function Attributes): New.
5283 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5285 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5286 check for 128-bit long double before checking TCmode.
5287 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5288 128-bit long doubles before checking TFmode or TCmode.
5289 (FLOAT128_IBM_P): Likewise.
5291 2018-01-10 Martin Sebor <msebor@redhat.com>
5293 PR tree-optimization/83671
5294 * builtins.c (c_strlen): Unconditionally return zero for the empty
5296 Use -Warray-bounds for warnings.
5297 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5298 for non-constant array indices with COMPONENT_REF, arrays of
5299 arrays, and pointers to arrays.
5300 (gimple_fold_builtin_strlen): Determine and set length range for
5301 non-constant character arrays.
5303 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5306 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5309 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5311 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5313 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5316 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5317 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5318 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5319 indexed_or_indirect_operand predicate.
5320 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5321 (*vsx_le_perm_load_v8hi): Likewise.
5322 (*vsx_le_perm_load_v16qi): Likewise.
5323 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5324 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5325 (*vsx_le_perm_store_v8hi): Likewise.
5326 (*vsx_le_perm_store_v16qi): Likewise.
5327 (eight unnamed splitters): Likewise.
5329 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5331 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5332 * config/rs6000/emmintrin.h: Likewise.
5333 * config/rs6000/mmintrin.h: Likewise.
5334 * config/rs6000/xmmintrin.h: Likewise.
5336 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5339 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5341 * tree.c (tree_nop_conversion): Return true for location wrapper
5343 (maybe_wrap_with_location): New function.
5344 (selftest::check_strip_nops): New function.
5345 (selftest::test_location_wrappers): New function.
5346 (selftest::tree_c_tests): Call it.
5347 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5348 (maybe_wrap_with_location): New decl.
5349 (EXPR_LOCATION_WRAPPER_P): New macro.
5350 (location_wrapper_p): New inline function.
5351 (tree_strip_any_location_wrapper): New inline function.
5353 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5356 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5357 stack_realign_offset for the largest alignment of stack slot
5359 (ix86_find_max_used_stack_alignment): New function.
5360 (ix86_finalize_stack_frame_flags): Use it. Set
5361 max_used_stack_alignment if we don't realign stack.
5362 * config/i386/i386.h (machine_function): Add
5363 max_used_stack_alignment.
5365 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5367 * config/arm/arm.opt (-mbranch-cost): New option.
5368 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5371 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5374 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5375 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5377 2018-01-10 Richard Biener <rguenther@suse.de>
5380 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
5381 early out so it also covers the case where we have a non-NULL
5384 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5386 PR tree-optimization/83753
5387 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
5388 for non-strided grouped accesses if the number of elements is 1.
5390 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5393 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
5394 * i386.h (TARGET_USE_GATHER): Define.
5395 * x86-tune.def (X86_TUNE_USE_GATHER): New.
5397 2018-01-10 Martin Liska <mliska@suse.cz>
5400 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
5401 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
5403 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
5404 CLEANUP_NO_PARTITIONING is not set.
5406 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
5408 * doc/rtl.texi: Remove documentation of (const ...) wrappers
5409 for vectors, as a partial revert of r254296.
5410 * rtl.h (const_vec_p): Delete.
5411 (const_vec_duplicate_p): Don't test for vector CONSTs.
5412 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
5413 * expmed.c (make_tree): Likewise.
5416 * common.md (E, F): Use CONSTANT_P instead of checking for
5418 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
5419 checking for CONST_VECTOR.
5421 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5424 * predict.c (force_edge_cold): Handle in more sane way edges
5427 2018-01-09 Carl Love <cel@us.ibm.com>
5429 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
5431 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
5432 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
5433 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
5434 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
5435 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
5436 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
5437 * config/rs6000/rs6000-protos.h: Add extern defition for
5438 rs6000_generate_float2_double_code.
5439 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
5441 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
5442 (float2_v2df): Add define_expand.
5444 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
5447 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
5448 op_mode in the force_to_mode call.
5450 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5452 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
5453 instead of checking each element individually.
5454 (aarch64_evpc_uzp): Likewise.
5455 (aarch64_evpc_zip): Likewise.
5456 (aarch64_evpc_ext): Likewise.
5457 (aarch64_evpc_rev): Likewise.
5458 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
5459 instead of checking each element individually. Return true without
5461 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
5462 whether all selected elements come from the same input, instead of
5463 checking each element individually. Remove calls to gen_rtx_REG,
5464 start_sequence and end_sequence and instead assert that no rtl is
5467 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5469 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
5470 order of HIGH and CONST checks.
5472 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
5474 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
5475 if the destination isn't an SSA_NAME.
5477 2018-01-09 Richard Biener <rguenther@suse.de>
5479 PR tree-optimization/83668
5480 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
5482 (canonicalize_loop_form): ... here, renamed from ...
5483 (canonicalize_loop_closed_ssa_form): ... this and amended to
5484 swap successor edges for loop exit blocks to make us use
5485 the RPO order we need for initial schedule generation.
5487 2018-01-09 Joseph Myers <joseph@codesourcery.com>
5489 PR tree-optimization/64811
5490 * match.pd: When optimizing comparisons with Inf, avoid
5491 introducing or losing exceptions from comparisons with NaN.
5493 2018-01-09 Martin Liska <mliska@suse.cz>
5496 * asan.c (shadow_mem_size): Add gcc_assert.
5498 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
5500 Don't save registers in main().
5503 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
5504 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
5505 * config/avr/avr.c (avr_set_current_function): Don't error if
5506 naked, OS_task or OS_main are specified at the same time.
5507 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
5509 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
5511 * common/config/avr/avr-common.c (avr_option_optimization_table):
5512 Switch on -mmain-is-OS_task for optimizing compilations.
5514 2018-01-09 Richard Biener <rguenther@suse.de>
5516 PR tree-optimization/83572
5517 * graphite.c: Include cfganal.h.
5518 (graphite_transform_loops): Connect infinite loops to exit
5519 and remove fake edges at the end.
5521 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5523 * ipa-inline.c (edge_badness): Revert accidental checkin.
5525 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
5528 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
5529 symbols; not inline clones.
5531 2018-01-09 Jakub Jelinek <jakub@redhat.com>
5534 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
5535 hard registers. Formatting fixes.
5537 PR preprocessor/83722
5538 * gcc.c (try_generate_repro): Pass
5539 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
5540 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
5543 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
5544 Kito Cheng <kito.cheng@gmail.com>
5546 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
5547 (riscv_leaf_function_p): Delete.
5548 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
5550 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5552 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
5554 (do_ifelse): New function.
5555 (do_isel): New function.
5556 (do_sub3): New function.
5557 (do_add3): New function.
5558 (do_load_mask_compare): New function.
5559 (do_overlap_load_compare): New function.
5560 (expand_compare_loop): New function.
5561 (expand_block_compare): Call expand_compare_loop() when appropriate.
5562 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
5564 (-mblock-compare-inline-loop-limit): New option.
5566 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5569 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
5570 Reverse order of second and third operands in first alternative.
5571 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
5572 of first and second elements in UNSPEC_VPERMR vector.
5573 (altivec_expand_vec_perm_le): Likewise.
5575 2018-01-08 Jeff Law <law@redhat.com>
5577 PR rtl-optimizatin/81308
5578 * tree-switch-conversion.c (cfg_altered): New file scoped static.
5579 (process_switch): If group_case_labels makes a change, then set
5581 (pass_convert_switch::execute): If a switch is converted, then
5582 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
5584 PR rtl-optimization/81308
5585 * recog.c (split_all_insns): Conditionally cleanup the CFG after
5588 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
5590 PR target/83663 - Revert r255946
5591 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
5592 generation for cases where splatting a value is not useful.
5593 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
5594 across a vec_duplicate and a paradoxical subreg forming a vector
5595 mode to a vec_concat.
5597 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5599 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
5600 -march=armv8.3-a variants.
5601 * config/arm/t-multilib: Likewise.
5602 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
5604 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5606 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
5608 (cceq_ior_compare_complement): Give it a name so I can use it, and
5609 change boolean_or_operator predicate to boolean_operator so it can
5610 be used to generate a crand.
5611 (eqne): New code iterator.
5612 (bd/bd_neg): New code_attrs.
5613 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
5614 a single define_insn.
5615 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
5616 decrement (bdnzt/bdnzf/bdzt/bdzf).
5617 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
5618 with the new names of the branch decrement patterns, and added the
5619 names of the branch decrement conditional patterns.
5621 2018-01-08 Richard Biener <rguenther@suse.de>
5623 PR tree-optimization/83563
5624 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
5627 2018-01-08 Richard Biener <rguenther@suse.de>
5630 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
5632 2018-01-08 Richard Biener <rguenther@suse.de>
5634 PR tree-optimization/83685
5635 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
5636 references to abnormals.
5638 2018-01-08 Richard Biener <rguenther@suse.de>
5641 * dwarf2out.c (output_indirect_strings): Handle empty
5642 skeleton_debug_str_hash.
5643 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
5645 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5647 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
5648 (emit_store_direct): Likewise.
5649 (arc_trampoline_adjust_address): Likewise.
5650 (arc_asm_trampoline_template): New function.
5651 (arc_initialize_trampoline): Use asm_trampoline_template.
5652 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
5653 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
5654 * config/arc/arc.md (flush_icache): Delete pattern.
5656 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
5658 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
5659 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
5662 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5665 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
5666 by not USED_FOR_TARGET.
5667 (make_pass_resolve_sw_modes): Likewise.
5669 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
5671 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
5674 2018-01-08 Richard Biener <rguenther@suse.de>
5677 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
5679 2018-01-08 Richard Biener <rguenther@suse.de>
5682 * match.pd ((t * 2) / 2) -> t): Add missing :c.
5684 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
5687 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
5688 basic blocks with a small number of successors.
5689 (convert_control_dep_chain_into_preds): Improve handling of
5691 (dump_predicates): Split apart into...
5692 (dump_pred_chain): ...here...
5693 (dump_pred_info): ...and here.
5694 (can_one_predicate_be_invalidated_p): Add debugging printfs.
5695 (can_chain_union_be_invalidated_p): Improve check for invalidation
5697 (uninit_uses_cannot_happen): Avoid unnecessary if
5698 convert_control_dep_chain_into_preds yielded nothing.
5700 2018-01-06 Martin Sebor <msebor@redhat.com>
5702 PR tree-optimization/83640
5703 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
5704 subtracting negative offset from size.
5705 (builtin_access::overlap): Adjust offset bounds of the access to fall
5706 within the size of the object if possible.
5708 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
5710 PR rtl-optimization/83699
5711 * expmed.c (extract_bit_field_1): Restrict the vector usage of
5712 extract_bit_field_as_subreg to cases in which the extracted
5713 value is also a vector.
5715 * lra-constraints.c (process_alt_operands): Test for the equivalence
5716 substitutions when detecting a possible reload cycle.
5718 2018-01-06 Jakub Jelinek <jakub@redhat.com>
5721 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
5722 by default if flag_selective_schedling{,2}. Formatting fixes.
5724 PR rtl-optimization/83682
5725 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
5726 if it has non-VECTOR_MODE element mode.
5727 (vec_duplicate_p): Likewise.
5730 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
5731 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
5733 2018-01-05 Jakub Jelinek <jakub@redhat.com>
5736 * config/i386/i386-builtin.def
5737 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
5738 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
5739 Require also OPTION_MASK_ISA_AVX512F in addition to
5740 OPTION_MASK_ISA_GFNI.
5741 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
5742 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
5743 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
5744 to OPTION_MASK_ISA_GFNI.
5745 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
5746 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
5747 OPTION_MASK_ISA_AVX512BW.
5748 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
5749 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
5750 addition to OPTION_MASK_ISA_GFNI.
5751 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
5752 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
5753 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
5754 to OPTION_MASK_ISA_GFNI.
5755 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
5756 a requirement for all ISAs rather than any of them with a few
5758 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
5760 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
5761 bitmasks to be enabled with 3 exceptions, instead of requiring any
5762 enabled ISA with lots of exceptions.
5763 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
5764 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
5765 Change avx512bw in isa attribute to avx512f.
5766 * config/i386/sgxintrin.h: Add license boilerplate.
5767 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
5768 to __AVX512F__ and __AVX512VL to __AVX512VL__.
5769 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
5770 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
5772 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
5773 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
5774 temporarily sse2 rather than sse if not enabled already.
5777 * config/i386/sse.md (VI248_VLBW): Rename to ...
5778 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
5779 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
5780 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
5781 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
5782 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
5783 mode iterator instead of VI248_VLBW.
5785 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
5787 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
5788 (record_modified): Skip clobbers; add debug output.
5789 (param_change_prob): Use sreal frequencies.
5791 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
5793 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
5794 punt for user-aligned variables.
5796 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
5798 * tree-chrec.c (chrec_contains_symbols): Return true for
5801 2018-01-05 Sudakshina Das <sudi.das@arm.com>
5804 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
5805 of (x|y) == x for BICS pattern.
5807 2018-01-05 Jakub Jelinek <jakub@redhat.com>
5809 PR tree-optimization/83605
5810 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
5811 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
5814 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
5816 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
5817 * config/epiphany/rtems.h: New file.
5819 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5820 Uros Bizjak <ubizjak@gmail.com>
5823 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
5824 QIreg_operand instead of register_operand predicate.
5825 * config/i386/i386.c (ix86_rop_should_change_byte_p,
5826 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
5827 comments instead of -fmitigate[-_]rop.
5829 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5832 * cgraphunit.c (symbol_table::compile): Switch to text_section
5833 before calling assembly_start debug hook.
5834 * run-rtl-passes.c (run_rtl_passes): Likewise.
5837 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5839 * tree-vrp.c (extract_range_from_binary_expr_1): Check
5840 range_int_cst_p rather than !symbolic_range_p before calling
5841 extract_range_from_multiplicative_op_1.
5843 2018-01-04 Jeff Law <law@redhat.com>
5845 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
5846 redundant test in assertion.
5848 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5850 * doc/rtl.texi: Document machine_mode wrapper classes.
5852 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5854 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
5857 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5859 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
5860 the VEC_PERM_EXPR fold to fail.
5862 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5865 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
5866 to switched_sections.
5868 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5871 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
5874 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
5877 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
5878 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
5880 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5883 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
5884 is BLKmode and bitpos not zero or mode change is needed.
5886 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
5889 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
5892 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
5895 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
5896 instead of MULT rtx. Update all corresponding splitters.
5898 (*ssub<modesuffix>): Ditto.
5900 (*cmp_sadd_di): Update split patterns.
5901 (*cmp_sadd_si): Ditto.
5902 (*cmp_sadd_sidi): Ditto.
5903 (*cmp_ssub_di): Ditto.
5904 (*cmp_ssub_si): Ditto.
5905 (*cmp_ssub_sidi): Ditto.
5906 * config/alpha/predicates.md (const23_operand): New predicate.
5907 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
5908 Look for ASHIFT, not MULT inner operand.
5909 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
5911 2018-01-04 Martin Liska <mliska@suse.cz>
5913 PR gcov-profile/83669
5914 * gcov.c (output_intermediate_file): Add version to intermediate
5916 * doc/gcov.texi: Document new field 'version' in intermediate
5917 file format. Fix location of '-k' option of gcov command.
5919 2018-01-04 Martin Liska <mliska@suse.cz>
5922 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
5924 2018-01-04 Jakub Jelinek <jakub@redhat.com>
5926 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
5928 2018-01-03 Martin Sebor <msebor@redhat.com>
5930 PR tree-optimization/83655
5931 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
5932 checking calls with invalid arguments.
5934 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5936 * tree-vect-stmts.c (vect_get_store_rhs): New function.
5937 (vectorizable_mask_load_store): Delete.
5938 (vectorizable_call): Return false for masked loads and stores.
5939 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
5940 instead of gimple_assign_rhs1.
5941 (vectorizable_load): Handle IFN_MASK_LOAD.
5942 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
5944 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5946 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
5948 (vectorizable_mask_load_store): ...here.
5949 (vectorizable_load): ...and here.
5951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5953 * tree-vect-stmts.c (vect_build_all_ones_mask)
5954 (vect_build_zero_merge_argument): New functions, split out from...
5955 (vectorizable_load): ...here.
5957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5959 * tree-vect-stmts.c (vect_check_store_rhs): New function,
5961 (vectorizable_mask_load_store): ...here.
5962 (vectorizable_store): ...and here.
5964 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5966 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
5968 (vectorizable_mask_load_store): ...here.
5970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5972 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
5973 (vect_model_store_cost): Take a vec_load_store_type instead of a
5975 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
5976 (vect_model_store_cost): Take a vec_load_store_type instead of a
5978 (vectorizable_mask_load_store): Update accordingly.
5979 (vectorizable_store): Likewise.
5980 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
5982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5984 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
5985 IFN_MASK_LOAD calls here rather than...
5986 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
5988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5989 Alan Hayward <alan.hayward@arm.com>
5990 David Sherwood <david.sherwood@arm.com>
5992 * expmed.c (extract_bit_field_1): For vector extracts,
5993 fall back to extract_bit_field_as_subreg if vec_extract
5996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
5997 Alan Hayward <alan.hayward@arm.com>
5998 David Sherwood <david.sherwood@arm.com>
6000 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6001 they are variable or constant sized.
6002 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6003 slots for constant-sized data.
6005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6006 Alan Hayward <alan.hayward@arm.com>
6007 David Sherwood <david.sherwood@arm.com>
6009 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6010 handling COND_EXPRs with boolean comparisons, try to find a better
6011 basis for the mask type than the boolean itself.
6013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6015 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6016 is calculated and how it can be overridden.
6017 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6018 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6020 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6024 Alan Hayward <alan.hayward@arm.com>
6025 David Sherwood <david.sherwood@arm.com>
6027 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6028 Remove the mode argument.
6029 (aarch64_simd_valid_immediate): Remove the mode and inverse
6031 * config/aarch64/iterators.md (bitsize): New iterator.
6032 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6033 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6034 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6035 aarch64_simd_valid_immediate.
6036 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6037 (aarch64_reg_or_bic_imm): Likewise.
6038 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6039 with an insn_type enum and msl with a modifier_type enum.
6040 Replace element_width with a scalar_mode. Change the shift
6041 to unsigned int. Add constructors for scalar_float_mode and
6042 scalar_int_mode elements.
6043 (aarch64_vect_float_const_representable_p): Delete.
6044 (aarch64_can_const_movi_rtx_p)
6045 (aarch64_simd_scalar_immediate_valid_for_move)
6046 (aarch64_simd_make_constant): Update call to
6047 aarch64_simd_valid_immediate.
6048 (aarch64_advsimd_valid_immediate_hs): New function.
6049 (aarch64_advsimd_valid_immediate): Likewise.
6050 (aarch64_simd_valid_immediate): Remove mode and inverse
6051 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6052 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6053 and aarch64_float_const_representable_p on the result.
6054 (aarch64_output_simd_mov_immediate): Remove mode argument.
6055 Update call to aarch64_simd_valid_immediate and use of
6056 simd_immediate_info.
6057 (aarch64_output_scalar_simd_mov_immediate): Update call
6060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6061 Alan Hayward <alan.hayward@arm.com>
6062 David Sherwood <david.sherwood@arm.com>
6064 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6065 (mode_nunits): Likewise CONST_MODE_NUNITS.
6066 * machmode.def (ADJUST_NUNITS): Document.
6067 * genmodes.c (mode_data::need_nunits_adj): New field.
6068 (blank_mode): Update accordingly.
6069 (adj_nunits): New variable.
6070 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6072 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6073 listed in adj_nunits.
6074 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6075 listed in adj_nunits. Don't emit case statements for such modes.
6076 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6077 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6078 nothing if adj_nunits is nonnull.
6079 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6080 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6081 (emit_mode_fbit): Update use of print_maybe_const_decl.
6082 (emit_move_size): Likewise. Treat the array as non-const
6084 (emit_mode_adjustments): Handle adj_nunits.
6086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6088 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6089 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6090 (VECTOR_MODES): Use it.
6091 (make_vector_modes): Take the prefix as an argument.
6093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6094 Alan Hayward <alan.hayward@arm.com>
6095 David Sherwood <david.sherwood@arm.com>
6097 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6098 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6099 for MODE_VECTOR_BOOL.
6100 * machmode.def (VECTOR_BOOL_MODE): Document.
6101 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6102 (make_vector_bool_mode): New function.
6103 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6105 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6106 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6108 * stor-layout.c (int_mode_for_mode): Likewise.
6109 * tree.c (build_vector_type_for_mode): Likewise.
6110 * varasm.c (output_constant_pool_2): Likewise.
6111 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6112 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6113 for MODE_VECTOR_BOOL.
6114 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6115 of mode class checks.
6116 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6117 instead of a list of mode class checks.
6118 (expand_vector_scalar_condition): Likewise.
6119 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6122 Alan Hayward <alan.hayward@arm.com>
6123 David Sherwood <david.sherwood@arm.com>
6125 * machmode.h (mode_size): Change from unsigned short to
6127 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6128 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6129 or if measurement_type is not polynomial.
6130 (fixed_size_mode::includes_p): Check for constant-sized modes.
6131 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6132 return a poly_uint16 rather than an unsigned short.
6133 (emit_mode_size): Change the type of mode_size from unsigned short
6134 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6135 (emit_mode_adjustments): Cope with polynomial vector sizes.
6136 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6138 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6140 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6141 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6142 * caller-save.c (setup_save_areas): Likewise.
6143 (replace_reg_with_saved_mem): Likewise.
6144 * calls.c (emit_library_call_value_1): Likewise.
6145 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6146 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6147 (gen_lowpart_for_combine): Likewise.
6148 * convert.c (convert_to_integer_1): Likewise.
6149 * cse.c (equiv_constant, cse_insn): Likewise.
6150 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6151 (cselib_subst_to_values): Likewise.
6152 * dce.c (word_dce_process_block): Likewise.
6153 * df-problems.c (df_word_lr_mark_ref): Likewise.
6154 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6155 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6156 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6157 (rtl_for_decl_location): Likewise.
6158 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6159 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6160 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6161 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6162 (expand_expr_real_1): Likewise.
6163 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6164 (pad_below): Likewise.
6165 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6166 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6167 * ira.c (get_subreg_tracking_sizes): Likewise.
6168 * ira-build.c (ira_create_allocno_objects): Likewise.
6169 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6170 (ira_sort_regnos_for_alter_reg): Likewise.
6171 * ira-costs.c (record_operand_costs): Likewise.
6172 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6173 (resolve_simple_move): Likewise.
6174 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6175 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6176 (lra_constraints): Likewise.
6177 (CONST_POOL_OK_P): Reject variable-sized modes.
6178 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6179 (add_pseudo_to_slot, lra_spill): Likewise.
6180 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6181 * optabs-query.c (get_best_extraction_insn): Likewise.
6182 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6183 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6184 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6185 * recog.c (offsettable_address_addr_space_p): Likewise.
6186 * regcprop.c (maybe_mode_change): Likewise.
6187 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6188 * regrename.c (build_def_use): Likewise.
6189 * regstat.c (dump_reg_info): Likewise.
6190 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6191 (find_reloads, find_reloads_subreg_address): Likewise.
6192 * reload1.c (eliminate_regs_1): Likewise.
6193 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6194 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6195 (simplify_binary_operation_1, simplify_subreg): Likewise.
6196 * targhooks.c (default_function_arg_padding): Likewise.
6197 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6198 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6199 (verify_gimple_assign_ternary): Likewise.
6200 * tree-inline.c (estimate_move_cost): Likewise.
6201 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6202 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6203 (get_address_cost_ainc): Likewise.
6204 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6205 (vect_supportable_dr_alignment): Likewise.
6206 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6207 (vectorizable_reduction): Likewise.
6208 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6209 (vectorizable_operation, vectorizable_load): Likewise.
6210 * tree.c (build_same_sized_truth_vector_type): Likewise.
6211 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6212 * var-tracking.c (emit_note_insn_var_location): Likewise.
6213 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6214 (ADDR_VEC_ALIGN): Likewise.
6216 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6217 Alan Hayward <alan.hayward@arm.com>
6218 David Sherwood <david.sherwood@arm.com>
6220 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6222 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6223 or if measurement_type is polynomial.
6224 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6225 * combine.c (make_extraction): Likewise.
6226 * dse.c (find_shift_sequence): Likewise.
6227 * dwarf2out.c (mem_loc_descriptor): Likewise.
6228 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6229 (extract_bit_field, extract_low_bits): Likewise.
6230 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6231 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6232 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6233 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6234 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6235 * reload.c (find_reloads): Likewise.
6236 * reload1.c (alter_reg): Likewise.
6237 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6238 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6239 * tree-if-conv.c (predicate_mem_writes): Likewise.
6240 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6241 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6242 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6243 * valtrack.c (dead_debug_insert_temp): Likewise.
6244 * varasm.c (mergeable_constant_section): Likewise.
6245 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6248 Alan Hayward <alan.hayward@arm.com>
6249 David Sherwood <david.sherwood@arm.com>
6251 * expr.c (expand_assignment): Cope with polynomial mode sizes
6252 when assigning to a CONCAT.
6254 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6255 Alan Hayward <alan.hayward@arm.com>
6256 David Sherwood <david.sherwood@arm.com>
6258 * machmode.h (mode_precision): Change from unsigned short to
6260 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6262 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6263 or if measurement_type is not polynomial.
6264 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6265 in which the mode is already known to be a scalar_int_mode.
6266 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6267 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6269 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6270 for GET_MODE_PRECISION.
6271 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6272 for GET_MODE_PRECISION.
6273 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6275 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6276 (expand_field_assignment, make_extraction): Likewise.
6277 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6278 (get_last_value): Likewise.
6279 * convert.c (convert_to_integer_1): Likewise.
6280 * cse.c (cse_insn): Likewise.
6281 * expr.c (expand_expr_real_1): Likewise.
6282 * lra-constraints.c (simplify_operand_subreg): Likewise.
6283 * optabs-query.c (can_atomic_load_p): Likewise.
6284 * optabs.c (expand_atomic_load): Likewise.
6285 (expand_atomic_store): Likewise.
6286 * ree.c (combine_reaching_defs): Likewise.
6287 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6288 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6289 * tree.h (type_has_mode_precision_p): Likewise.
6290 * ubsan.c (instrument_si_overflow): Likewise.
6292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6293 Alan Hayward <alan.hayward@arm.com>
6294 David Sherwood <david.sherwood@arm.com>
6296 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6297 polynomial numbers of units.
6298 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6299 (valid_vector_subparts_p): New function.
6300 (build_vector_type): Remove temporary shim and take the number
6301 of units as a poly_uint64 rather than an int.
6302 (build_opaque_vector_type): Take the number of units as a
6303 poly_uint64 rather than an int.
6304 * tree.c (build_vector_from_ctor): Handle polynomial
6305 TYPE_VECTOR_SUBPARTS.
6306 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6307 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6308 (build_vector_from_val): If the number of units is variable,
6309 use build_vec_duplicate_cst for constant operands and
6310 VEC_DUPLICATE_EXPR otherwise.
6311 (make_vector_type): Remove temporary is_constant ().
6312 (build_vector_type, build_opaque_vector_type): Take the number of
6313 units as a poly_uint64 rather than an int.
6314 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6316 * cfgexpand.c (expand_debug_expr): Likewise.
6317 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6318 (store_constructor, expand_expr_real_1): Likewise.
6319 (const_scalar_mask_from_tree): Likewise.
6320 * fold-const-call.c (fold_const_reduction): Likewise.
6321 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6322 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6323 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6324 (fold_relational_const): Likewise.
6325 (native_interpret_vector): Likewise. Change the size from an
6326 int to an unsigned int.
6327 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6328 TYPE_VECTOR_SUBPARTS.
6329 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6330 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6331 duplicating a non-constant operand into a variable-length vector.
6332 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6333 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6334 * ipa-icf.c (sem_variable::equals): Likewise.
6335 * match.pd: Likewise.
6336 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6337 * print-tree.c (print_node): Likewise.
6338 * stor-layout.c (layout_type): Likewise.
6339 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6340 * tree-cfg.c (verify_gimple_comparison): Likewise.
6341 (verify_gimple_assign_binary): Likewise.
6342 (verify_gimple_assign_ternary): Likewise.
6343 (verify_gimple_assign_single): Likewise.
6344 * tree-pretty-print.c (dump_generic_node): Likewise.
6345 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6346 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6347 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6348 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6349 (vect_shift_permute_load_chain): Likewise.
6350 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6351 (expand_vector_condition, optimize_vector_constructor): Likewise.
6352 (lower_vec_perm, get_compute_type): Likewise.
6353 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6354 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6355 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6356 (vect_recog_mask_conversion_pattern): Likewise.
6357 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6358 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6359 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6360 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6361 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6362 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6363 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6364 (supportable_widening_operation): Likewise.
6365 (supportable_narrowing_operation): Likewise.
6366 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6368 * varasm.c (output_constant): Likewise.
6370 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6371 Alan Hayward <alan.hayward@arm.com>
6372 David Sherwood <david.sherwood@arm.com>
6374 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6375 so that both the length == 3 and length != 3 cases set up their
6376 own permute vectors. Add comments explaining why we know the
6377 number of elements is constant.
6378 (vect_permute_load_chain): Likewise.
6380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6381 Alan Hayward <alan.hayward@arm.com>
6382 David Sherwood <david.sherwood@arm.com>
6384 * machmode.h (mode_nunits): Change from unsigned char to
6386 (ONLY_FIXED_SIZE_MODES): New macro.
6387 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
6388 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
6389 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
6391 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
6392 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
6393 or if measurement_type is not polynomial.
6394 * genmodes.c (ZERO_COEFFS): New macro.
6395 (emit_mode_nunits_inline): Make mode_nunits_inline return a
6397 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
6398 Use ZERO_COEFFS when emitting initializers.
6399 * data-streamer.h (bp_pack_poly_value): New function.
6400 (bp_unpack_poly_value): Likewise.
6401 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6402 for GET_MODE_NUNITS.
6403 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6404 for GET_MODE_NUNITS.
6405 * tree.c (make_vector_type): Remove temporary shim and make
6406 the real function take the number of units as a poly_uint64
6408 (build_vector_type_for_mode): Handle polynomial nunits.
6409 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
6410 * emit-rtl.c (const_vec_series_p_1): Likewise.
6411 (gen_rtx_CONST_VECTOR): Likewise.
6412 * fold-const.c (test_vec_duplicate_folding): Likewise.
6413 * genrecog.c (validate_pattern): Likewise.
6414 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
6415 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6416 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
6417 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
6418 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
6419 * rtlanal.c (subreg_get_info): Likewise.
6420 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
6421 (vect_grouped_load_supported): Likewise.
6422 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
6423 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
6424 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
6425 (simplify_const_unary_operation, simplify_binary_operation_1)
6426 (simplify_const_binary_operation, simplify_ternary_operation)
6427 (test_vector_ops_duplicate, test_vector_ops): Likewise.
6428 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
6429 instead of CONST_VECTOR_NUNITS.
6430 * varasm.c (output_constant_pool_2): Likewise.
6431 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
6432 explicit-encoded elements in the XVEC for variable-length vectors.
6434 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6436 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
6438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6439 Alan Hayward <alan.hayward@arm.com>
6440 David Sherwood <david.sherwood@arm.com>
6442 * coretypes.h (fixed_size_mode): Declare.
6443 (fixed_size_mode_pod): New typedef.
6444 * builtins.h (target_builtins::x_apply_args_mode)
6445 (target_builtins::x_apply_result_mode): Change type to
6446 fixed_size_mode_pod.
6447 * builtins.c (apply_args_size, apply_result_size, result_vector)
6448 (expand_builtin_apply_args_1, expand_builtin_apply)
6449 (expand_builtin_return): Update accordingly.
6451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6453 * cse.c (hash_rtx_cb): Hash only the encoded elements.
6454 * cselib.c (cselib_hash_rtx): Likewise.
6455 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
6456 CONST_VECTOR encoding.
6458 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6459 Jeff Law <law@redhat.com>
6462 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
6463 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
6464 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
6465 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
6468 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
6469 explicitly probe *sp in a noreturn function if there were any callee
6470 register saves or frame pointer is needed.
6472 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6475 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
6476 BLKmode for ternary, binary or unary expressions.
6479 * var-tracking.c (delete_vta_debug_insn): New inline function.
6480 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
6481 insns from get_insns () to NULL instead of each bb separately.
6482 Use delete_vta_debug_insn. No longer static.
6483 (vt_debug_insns_local, variable_tracking_main_1): Adjust
6484 delete_vta_debug_insns callers.
6485 * rtl.h (delete_vta_debug_insns): Declare.
6486 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
6487 instead of variable_tracking_main.
6489 2018-01-03 Martin Sebor <msebor@redhat.com>
6491 PR tree-optimization/83603
6492 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
6493 arguments past the endof the argument list in functions declared
6494 without a prototype.
6495 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
6496 Avoid checking when arguments are null.
6498 2018-01-03 Martin Sebor <msebor@redhat.com>
6501 * doc/extend.texi (attribute const): Fix a typo.
6502 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
6503 issuing -Wsuggest-attribute for void functions.
6505 2018-01-03 Martin Sebor <msebor@redhat.com>
6507 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
6508 offset_int::from instead of wide_int::to_shwi.
6509 (maybe_diag_overlap): Remove assertion.
6510 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
6511 * gimple-ssa-sprintf.c (format_directive): Same.
6512 (parse_directive): Same.
6513 (sprintf_dom_walker::compute_format_length): Same.
6514 (try_substitute_return_value): Same.
6516 2018-01-03 Jeff Law <law@redhat.com>
6519 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
6520 non-constant residual for zero at runtime and avoid probing in
6521 that case. Reorganize code for trailing problem to mirror handling
6524 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6526 PR tree-optimization/83501
6527 * tree-ssa-strlen.c (get_string_cst): New.
6528 (handle_char_store): Call get_string_cst.
6530 2018-01-03 Martin Liska <mliska@suse.cz>
6532 PR tree-optimization/83593
6533 * tree-ssa-strlen.c: Include tree-cfg.h.
6534 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
6535 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
6536 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
6538 (strlen_dom_walker::before_dom_children): Call
6539 gimple_purge_dead_eh_edges. Dump tranformation with details
6541 (strlen_dom_walker::before_dom_children): Update call by adding
6542 new argument cleanup_eh.
6543 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
6545 2018-01-03 Martin Liska <mliska@suse.cz>
6548 * cif-code.def (VARIADIC_THUNK): New enum value.
6549 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
6552 2018-01-03 Jan Beulich <jbeulich@suse.com>
6554 * sse.md (mov<mode>_internal): Tighten condition for when to use
6555 vmovdqu<ssescalarsize> for TI and OI modes.
6557 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6559 Update copyright years.
6561 2018-01-03 Martin Liska <mliska@suse.cz>
6564 * ipa-visibility.c (function_and_variable_visibility): Skip
6565 functions with noipa attribure.
6567 2018-01-03 Jakub Jelinek <jakub@redhat.com>
6569 * gcc.c (process_command): Update copyright notice dates.
6570 * gcov-dump.c (print_version): Ditto.
6571 * gcov.c (print_version): Ditto.
6572 * gcov-tool.c (print_version): Ditto.
6573 * gengtype.c (create_file): Ditto.
6574 * doc/cpp.texi: Bump @copying's copyright year.
6575 * doc/cppinternals.texi: Ditto.
6576 * doc/gcc.texi: Ditto.
6577 * doc/gccint.texi: Ditto.
6578 * doc/gcov.texi: Ditto.
6579 * doc/install.texi: Ditto.
6580 * doc/invoke.texi: Ditto.
6582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6584 * vector-builder.h (vector_builder::m_full_nelts): Change from
6585 unsigned int to poly_uint64.
6586 (vector_builder::full_nelts): Update prototype accordingly.
6587 (vector_builder::new_vector): Likewise.
6588 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
6589 (vector_builder::operator ==): Likewise.
6590 (vector_builder::finalize): Likewise.
6591 * int-vector-builder.h (int_vector_builder::int_vector_builder):
6592 Take the number of elements as a poly_uint64 rather than an
6594 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
6595 from unsigned int to poly_uint64.
6596 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
6597 (vec_perm_indices::new_vector): Likewise.
6598 (vec_perm_indices::length): Likewise.
6599 (vec_perm_indices::nelts_per_input): Likewise.
6600 (vec_perm_indices::input_nelts): Likewise.
6601 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
6602 number of elements per input as a poly_uint64 rather than an
6603 unsigned int. Use the original encoding for variable-length
6604 vectors, rather than clamping each individual element.
6605 For the second and subsequent elements in each pattern,
6606 clamp the step and base before clamping their sum.
6607 (vec_perm_indices::series_p): Handle polynomial element counts.
6608 (vec_perm_indices::all_in_range_p): Likewise.
6609 (vec_perm_indices_to_tree): Likewise.
6610 (vec_perm_indices_to_rtx): Likewise.
6611 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
6612 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
6613 (tree_vector_builder::new_binary_operation): Handle polynomial
6614 element counts. Return false if we need to know the number
6615 of elements at compile time.
6616 * fold-const.c (fold_vec_perm): Punt if the number of elements
6617 isn't known at compile time.
6619 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6621 * vec-perm-indices.h (vec_perm_builder): Change element type
6622 from HOST_WIDE_INT to poly_int64.
6623 (vec_perm_indices::element_type): Update accordingly.
6624 (vec_perm_indices::clamp): Handle polynomial element_types.
6625 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
6626 (vec_perm_indices::all_in_range_p): Likewise.
6627 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
6629 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
6630 polynomial vec_perm_indices element types.
6631 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
6632 * fold-const.c (fold_vec_perm): Likewise.
6633 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
6634 * tree-vect-generic.c (lower_vec_perm): Likewise.
6635 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
6636 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
6637 element type to HOST_WIDE_INT.
6639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6640 Alan Hayward <alan.hayward@arm.com>
6641 David Sherwood <david.sherwood@arm.com>
6643 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
6644 rather than an int. Use plus_constant.
6645 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
6646 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
6648 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6649 Alan Hayward <alan.hayward@arm.com>
6650 David Sherwood <david.sherwood@arm.com>
6652 * calls.c (emit_call_1, expand_call): Change struct_value_size from
6653 a HOST_WIDE_INT to a poly_int64.
6655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6656 Alan Hayward <alan.hayward@arm.com>
6657 David Sherwood <david.sherwood@arm.com>
6659 * calls.c (load_register_parameters): Cope with polynomial
6660 mode sizes. Require a constant size for BLKmode parameters
6661 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
6662 forces a parameter to be padded at the lsb end in order to
6663 fill a complete number of words, require the parameter size
6664 to be ordered wrt UNITS_PER_WORD.
6666 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6667 Alan Hayward <alan.hayward@arm.com>
6668 David Sherwood <david.sherwood@arm.com>
6670 * reload1.c (spill_stack_slot_width): Change element type
6671 from unsigned int to poly_uint64_pod.
6672 (alter_reg): Treat mode sizes as polynomial.
6674 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6675 Alan Hayward <alan.hayward@arm.com>
6676 David Sherwood <david.sherwood@arm.com>
6678 * reload.c (complex_word_subreg_p): New function.
6679 (reload_inner_reg_of_subreg, push_reload): Use it.
6681 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6682 Alan Hayward <alan.hayward@arm.com>
6683 David Sherwood <david.sherwood@arm.com>
6685 * lra-constraints.c (process_alt_operands): Reject matched
6686 operands whose sizes aren't ordered.
6687 (match_reload): Refer to this check here.
6689 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6690 Alan Hayward <alan.hayward@arm.com>
6691 David Sherwood <david.sherwood@arm.com>
6693 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
6694 that the mode size is in the set {1, 2, 4, 8, 16}.
6696 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6697 Alan Hayward <alan.hayward@arm.com>
6698 David Sherwood <david.sherwood@arm.com>
6700 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
6701 Use plus_constant instead of gen_rtx_PLUS.
6703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6704 Alan Hayward <alan.hayward@arm.com>
6705 David Sherwood <david.sherwood@arm.com>
6707 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
6708 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
6709 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
6710 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
6711 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
6712 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
6713 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
6714 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
6715 * config/i386/i386.c (ix86_push_rounding): ...this new function.
6716 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
6718 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
6719 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
6720 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
6721 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
6722 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
6723 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
6724 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
6725 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
6726 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
6727 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
6729 * expr.c (emit_move_resolve_push): Treat the input and result
6730 of PUSH_ROUNDING as a poly_int64.
6731 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
6732 (emit_push_insn): Likewise.
6733 * lra-eliminations.c (mark_not_eliminable): Likewise.
6734 * recog.c (push_operand): Likewise.
6735 * reload1.c (elimination_effects): Likewise.
6736 * rtlanal.c (nonzero_bits1): Likewise.
6737 * calls.c (store_one_arg): Likewise. Require the padding to be
6738 known at compile time.
6740 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6741 Alan Hayward <alan.hayward@arm.com>
6742 David Sherwood <david.sherwood@arm.com>
6744 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
6745 Use plus_constant instead of gen_rtx_PLUS.
6747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6748 Alan Hayward <alan.hayward@arm.com>
6749 David Sherwood <david.sherwood@arm.com>
6751 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
6754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6755 Alan Hayward <alan.hayward@arm.com>
6756 David Sherwood <david.sherwood@arm.com>
6758 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
6759 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
6760 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
6767 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
6768 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
6769 as a poly_uint64 too.
6771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6772 Alan Hayward <alan.hayward@arm.com>
6773 David Sherwood <david.sherwood@arm.com>
6775 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
6777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6778 Alan Hayward <alan.hayward@arm.com>
6779 David Sherwood <david.sherwood@arm.com>
6781 * combine.c (can_change_dest_mode): Handle polynomial
6782 REGMODE_NATURAL_SIZE.
6783 * expmed.c (store_bit_field_1): Likewise.
6784 * expr.c (store_constructor): Likewise.
6785 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
6786 and polynomial REGMODE_NATURAL_SIZE.
6787 (gen_lowpart_common): Likewise.
6788 * reginfo.c (record_subregs_of_mode): Likewise.
6789 * rtlanal.c (read_modify_subreg_p): Likewise.
6791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6792 Alan Hayward <alan.hayward@arm.com>
6793 David Sherwood <david.sherwood@arm.com>
6795 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
6796 numbers of elements.
6798 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6799 Alan Hayward <alan.hayward@arm.com>
6800 David Sherwood <david.sherwood@arm.com>
6802 * match.pd: Cope with polynomial numbers of vector elements.
6804 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6805 Alan Hayward <alan.hayward@arm.com>
6806 David Sherwood <david.sherwood@arm.com>
6808 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
6809 in a POINTER_PLUS_EXPR.
6811 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6812 Alan Hayward <alan.hayward@arm.com>
6813 David Sherwood <david.sherwood@arm.com>
6815 * omp-simd-clone.c (simd_clone_subparts): New function.
6816 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
6817 (ipa_simd_modify_function_body): Likewise.
6819 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6820 Alan Hayward <alan.hayward@arm.com>
6821 David Sherwood <david.sherwood@arm.com>
6823 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
6824 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
6825 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
6826 (expand_vector_condition, vector_element): Likewise.
6827 (subparts_gt): New function.
6828 (get_compute_type): Use subparts_gt.
6829 (count_type_subparts): Delete.
6830 (expand_vector_operations_1): Use subparts_gt instead of
6831 count_type_subparts.
6833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6834 Alan Hayward <alan.hayward@arm.com>
6835 David Sherwood <david.sherwood@arm.com>
6837 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
6838 (vect_compile_time_alias): ...this new function. Do the calculation
6839 on poly_ints rather than trees.
6840 (vect_prune_runtime_alias_test_list): Update call accordingly.
6842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6843 Alan Hayward <alan.hayward@arm.com>
6844 David Sherwood <david.sherwood@arm.com>
6846 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
6848 (vect_schedule_slp_instance): Likewise.
6850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6851 Alan Hayward <alan.hayward@arm.com>
6852 David Sherwood <david.sherwood@arm.com>
6854 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
6855 constant and extern definitions for variable-length vectors.
6856 (vect_get_constant_vectors): Note that the number of units
6857 is known to be constant.
6859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6860 Alan Hayward <alan.hayward@arm.com>
6861 David Sherwood <david.sherwood@arm.com>
6863 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
6864 of units as polynomial. Choose between WIDE and NARROW based
6867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6868 Alan Hayward <alan.hayward@arm.com>
6869 David Sherwood <david.sherwood@arm.com>
6871 * tree-vect-stmts.c (simd_clone_subparts): New function.
6872 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
6874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6875 Alan Hayward <alan.hayward@arm.com>
6876 David Sherwood <david.sherwood@arm.com>
6878 * tree-vect-stmts.c (vectorizable_call): Treat the number of
6879 vectors as polynomial. Use build_index_vector for
6882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6883 Alan Hayward <alan.hayward@arm.com>
6884 David Sherwood <david.sherwood@arm.com>
6886 * tree-vect-stmts.c (get_load_store_type): Treat the number of
6887 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
6888 for variable-length vectors.
6889 (vectorizable_mask_load_store): Treat the number of units as
6890 polynomial, asserting that it is constant if the condition has
6891 already been enforced.
6892 (vectorizable_store, vectorizable_load): Likewise.
6894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6895 Alan Hayward <alan.hayward@arm.com>
6896 David Sherwood <david.sherwood@arm.com>
6898 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
6899 of units as polynomial. Punt if we can't tell at compile time
6900 which vector contains the final result.
6902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6903 Alan Hayward <alan.hayward@arm.com>
6904 David Sherwood <david.sherwood@arm.com>
6906 * tree-vect-loop.c (vectorizable_induction): Treat the number
6907 of units as polynomial. Punt on SLP inductions. Use an integer
6908 VEC_SERIES_EXPR for variable-length integer reductions. Use a
6909 cast of such a series for variable-length floating-point
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 * tree.h (build_index_vector): Declare.
6917 * tree.c (build_index_vector): New function.
6918 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
6919 of units as polynomial, forcibly converting it to a constant if
6920 vectorizable_reduction has already enforced the condition.
6921 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
6922 to create a {1,2,3,...} vector.
6923 (vectorizable_reduction): Treat the number of units as polynomial.
6924 Choose vectype_in based on the largest scalar element size rather
6925 than the smallest number of units. Enforce the restrictions
6928 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6929 Alan Hayward <alan.hayward@arm.com>
6930 David Sherwood <david.sherwood@arm.com>
6932 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
6933 number of units as polynomial.
6935 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6936 Alan Hayward <alan.hayward@arm.com>
6937 David Sherwood <david.sherwood@arm.com>
6939 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
6940 * target.def (autovectorize_vector_sizes): Return the vector sizes
6941 by pointer, using vector_sizes rather than a bitmask.
6942 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
6943 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
6944 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
6946 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
6947 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
6948 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
6949 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
6950 * omp-general.c (omp_max_vf): Likewise.
6951 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6952 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
6953 * tree-vect-loop.c (vect_analyze_loop): Likewise.
6954 * tree-vect-slp.c (vect_slp_bb): Likewise.
6955 * doc/tm.texi: Regenerate.
6956 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
6958 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
6959 the vector size as a poly_uint64 rather than an unsigned int.
6960 (current_vector_size): Change from an unsigned int to a poly_uint64.
6961 (get_vectype_for_scalar_type): Update accordingly.
6962 * tree.h (build_truth_vector_type): Take the size and number of
6963 units as a poly_uint64 rather than an unsigned int.
6964 (build_vector_type): Add a temporary overload that takes
6965 the number of units as a poly_uint64 rather than an unsigned int.
6966 * tree.c (make_vector_type): Likewise.
6967 (build_truth_vector_type): Take the number of units as a poly_uint64
6968 rather than an unsigned int.
6970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6971 Alan Hayward <alan.hayward@arm.com>
6972 David Sherwood <david.sherwood@arm.com>
6974 * target.def (get_mask_mode): Take the number of units and length
6975 as poly_uint64s rather than unsigned ints.
6976 * targhooks.h (default_get_mask_mode): Update accordingly.
6977 * targhooks.c (default_get_mask_mode): Likewise.
6978 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
6979 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
6986 * omp-general.c (omp_max_vf): Likewise.
6987 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
6988 (expand_omp_simd): Handle polynomial safelen.
6989 * omp-low.c (omplow_simd_context): Add a default constructor.
6990 (omplow_simd_context::max_vf): Change from int to poly_uint64.
6991 (lower_rec_simd_input_clauses): Update accordingly.
6992 (lower_rec_input_clauses): Likewise.
6994 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6995 Alan Hayward <alan.hayward@arm.com>
6996 David Sherwood <david.sherwood@arm.com>
6998 * tree-vectorizer.h (vect_nunits_for_cost): New function.
6999 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7000 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7001 (vect_analyze_slp_cost): Likewise.
7002 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7003 (vect_model_load_cost): Likewise.
7005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7006 Alan Hayward <alan.hayward@arm.com>
7007 David Sherwood <david.sherwood@arm.com>
7009 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7010 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7011 from an unsigned int * to a poly_uint64_pod *.
7012 (calculate_unrolling_factor): New function.
7013 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7015 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7016 Alan Hayward <alan.hayward@arm.com>
7017 David Sherwood <david.sherwood@arm.com>
7019 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7020 from an unsigned int to a poly_uint64.
7021 (_loop_vec_info::slp_unrolling_factor): Likewise.
7022 (_loop_vec_info::vectorization_factor): Change from an int
7024 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7025 (vect_get_num_vectors): New function.
7026 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7027 (vect_get_num_copies): Use vect_get_num_vectors.
7028 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7029 to an unsigned int *.
7030 (vect_analyze_data_refs): Change min_vf from an int * to a
7032 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7033 than an unsigned HOST_WIDE_INT.
7034 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7035 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7036 to an unsigned int *.
7037 (vect_analyze_data_ref_dependences): Likewise.
7038 (vect_compute_data_ref_alignment): Handle polynomial vf.
7039 (vect_enhance_data_refs_alignment): Likewise.
7040 (vect_prune_runtime_alias_test_list): Likewise.
7041 (vect_shift_permute_load_chain): Likewise.
7042 (vect_supportable_dr_alignment): Likewise.
7043 (dependence_distance_ge_vf): Take the vectorization factor as a
7044 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7045 (vect_analyze_data_refs): Change min_vf from an int * to a
7047 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7048 vfm1 as a poly_uint64 rather than an int. Make the same change
7049 for the returned bound_scalar.
7050 (vect_gen_vector_loop_niters): Handle polynomial vf.
7051 (vect_do_peeling): Likewise. Update call to
7052 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7053 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7055 * tree-vect-loop.c (vect_determine_vectorization_factor)
7056 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7057 (vect_get_known_peeling_cost): Likewise.
7058 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7059 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7060 (vect_transform_loop): Likewise. Use the lowest possible VF when
7061 updating the upper bounds of the loop.
7062 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7064 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7065 polynomial unroll factors.
7066 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7067 (vect_make_slp_decision): Likewise.
7068 (vect_supported_load_permutation_p): Likewise, and polynomial
7070 (vect_analyze_slp_cost): Handle polynomial vf.
7071 (vect_slp_analyze_node_operations): Likewise.
7072 (vect_slp_analyze_bb_1): Likewise.
7073 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7074 than an unsigned HOST_WIDE_INT.
7075 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7076 (vectorizable_load): Handle polynomial vf.
7077 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7079 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7082 Alan Hayward <alan.hayward@arm.com>
7083 David Sherwood <david.sherwood@arm.com>
7085 * match.pd: Handle bit operations involving three constants
7086 and try to fold one pair.
7088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7090 * tree-vect-loop-manip.c: Include gimple-fold.h.
7091 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7092 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7093 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7094 Add a path that uses a step of VF instead of 1, but disable it
7096 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7097 and niters_no_overflow parameters. Update calls to
7098 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7099 Create a new SSA name if the latter choses to use a ste other
7100 than zero, and return it via niters_vector_mult_vf_var.
7101 * tree-vect-loop.c (vect_transform_loop): Update calls to
7102 vect_do_peeling, vect_gen_vector_loop_niters and
7103 slpeel_make_loop_iterate_ntimes.
7104 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7105 (vect_gen_vector_loop_niters): Update declarations after above changes.
7107 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7109 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7110 128-bit round to integer instructions.
7111 (ceil<mode>2): Likewise.
7112 (btrunc<mode>2): Likewise.
7113 (round<mode>2): Likewise.
7115 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7117 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7118 unaligned VSX load/store on P8/P9.
7119 (expand_block_clear): Allow the use of unaligned VSX
7120 load/store on P8/P9.
7122 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7124 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7126 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7127 swap associated with both a load and a store.
7129 2018-01-02 Andrew Waterman <andrew@sifive.com>
7131 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7132 * config/riscv/riscv.md (clear_cache): Use it.
7134 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7136 * web.c: Remove out-of-date comment.
7138 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7140 * expr.c (fixup_args_size_notes): Check that any existing
7141 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7142 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7143 (emit_single_push_insn): ...here.
7145 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7147 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7148 (const_vector_encoded_nelts): New function.
7149 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7150 (const_vector_int_elt, const_vector_elt): Declare.
7151 * emit-rtl.c (const_vector_int_elt_1): New function.
7152 (const_vector_elt): Likewise.
7153 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7154 of CONST_VECTOR_ELT.
7156 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7158 * expr.c: Include rtx-vector-builder.h.
7159 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7160 directly on the tree encoding.
7161 (const_vector_from_tree): Likewise.
7162 * optabs.c: Include rtx-vector-builder.h.
7163 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7164 sequence of "u" values.
7165 * vec-perm-indices.c: Include rtx-vector-builder.h.
7166 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7167 directly on the vec_perm_indices encoding.
7169 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7171 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7172 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7173 * rtx-vector-builder.h: New file.
7174 * rtx-vector-builder.c: Likewise.
7175 * rtl.h (rtx_def::u2): Add a const_vector field.
7176 (CONST_VECTOR_NPATTERNS): New macro.
7177 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7178 (CONST_VECTOR_DUPLICATE_P): Likewise.
7179 (CONST_VECTOR_STEPPED_P): Likewise.
7180 (CONST_VECTOR_ENCODED_ELT): Likewise.
7181 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7182 (unwrap_const_vec_duplicate): Likewise.
7183 (const_vec_series_p): Check for a non-duplicated vector encoding.
7184 Say that the function only returns true for integer vectors.
7185 * emit-rtl.c: Include rtx-vector-builder.h.
7186 (gen_const_vec_duplicate_1): Delete.
7187 (gen_const_vector): Call gen_const_vec_duplicate instead of
7188 gen_const_vec_duplicate_1.
7189 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7190 (gen_const_vec_duplicate): Use rtx_vector_builder.
7191 (gen_const_vec_series): Likewise.
7192 (gen_rtx_CONST_VECTOR): Likewise.
7193 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7194 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7195 Build a new vector rather than modifying a CONST_VECTOR in-place.
7196 (handle_special_swappables): Update call accordingly.
7197 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7198 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7199 Build a new vector rather than modifying a CONST_VECTOR in-place.
7200 (handle_special_swappables): Update call accordingly.
7202 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7204 * simplify-rtx.c (simplify_const_binary_operation): Use
7205 CONST_VECTOR_ELT instead of XVECEXP.
7207 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7209 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7210 the selector elements to be different from the data elements
7211 if the selector is a VECTOR_CST.
7212 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7213 ssizetype for the selector.
7215 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7217 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7218 before testing each element individually.
7219 * tree-vect-generic.c (lower_vec_perm): Likewise.
7221 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7223 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7224 * selftest-run-tests.c (selftest::run_tests): Call it.
7225 * vector-builder.h (vector_builder::operator ==): New function.
7226 (vector_builder::operator !=): Likewise.
7227 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7228 (vec_perm_indices::all_from_input_p): New function.
7229 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7230 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7231 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7232 instead of reading the VECTOR_CST directly. Detect whether both
7233 vector inputs are the same before constructing the vec_perm_indices,
7234 and update the number of inputs argument accordingly. Use the
7235 utility functions added above. Only construct sel2 if we need to.
7237 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7239 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7240 the broadcast of the low byte.
7241 (expand_mult_highpart): Use an explicit encoding for the permutes.
7242 * optabs-query.c (can_mult_highpart_p): Likewise.
7243 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7244 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7245 (vectorizable_bswap): Likewise.
7246 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7247 explicit encoding for the power-of-2 permutes.
7248 (vect_permute_store_chain): Likewise.
7249 (vect_grouped_load_supported): Likewise.
7250 (vect_permute_load_chain): Likewise.
7252 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7254 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7255 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7256 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7257 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7258 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7259 (vect_gen_perm_mask_any): Likewise.
7261 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7263 * int-vector-builder.h: New file.
7264 * vec-perm-indices.h: Include int-vector-builder.h.
7265 (vec_perm_indices): Redefine as an int_vector_builder.
7266 (auto_vec_perm_indices): Delete.
7267 (vec_perm_builder): Redefine as a stand-alone class.
7268 (vec_perm_indices::vec_perm_indices): New function.
7269 (vec_perm_indices::clamp): Likewise.
7270 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7271 (vec_perm_indices::new_vector): New function.
7272 (vec_perm_indices::new_expanded_vector): Update for new
7273 vec_perm_indices class.
7274 (vec_perm_indices::rotate_inputs): New function.
7275 (vec_perm_indices::all_in_range_p): Operate directly on the
7276 encoded form, without computing elided elements.
7277 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7278 encoding. Update for new vec_perm_indices class.
7279 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7280 the given vec_perm_builder.
7281 (expand_vec_perm_var): Update vec_perm_builder constructor.
7282 (expand_mult_highpart): Use vec_perm_builder instead of
7283 auto_vec_perm_indices.
7284 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7285 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7286 or double series encoding as appropriate.
7287 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7288 vec_perm_indices instead of auto_vec_perm_indices.
7289 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7290 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7291 (vect_permute_store_chain): Likewise.
7292 (vect_grouped_load_supported): Likewise.
7293 (vect_permute_load_chain): Likewise.
7294 (vect_shift_permute_load_chain): Likewise.
7295 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7296 (vect_transform_slp_perm_load): Likewise.
7297 (vect_schedule_slp_instance): Likewise.
7298 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7299 (vectorizable_mask_load_store): Likewise.
7300 (vectorizable_bswap): Likewise.
7301 (vectorizable_store): Likewise.
7302 (vectorizable_load): Likewise.
7303 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7304 vec_perm_indices instead of auto_vec_perm_indices. Use
7305 tree_to_vec_perm_builder to read the vector from a tree.
7306 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7307 vec_perm_builder instead of a vec_perm_indices.
7308 (have_whole_vector_shift): Use vec_perm_builder and
7309 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7310 truncation to calc_vec_perm_mask_for_shift.
7311 (vect_create_epilog_for_reduction): Likewise.
7312 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7313 from auto_vec_perm_indices to vec_perm_indices.
7314 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7315 instead of changing individual elements.
7316 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7317 the vector in d.perm.
7318 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7319 from auto_vec_perm_indices to vec_perm_indices.
7320 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7321 instead of changing individual elements.
7322 (arm_vectorize_vec_perm_const): Use new_vector to install
7323 the vector in d.perm.
7324 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7325 Update vec_perm_builder constructor.
7326 (rs6000_expand_interleave): Likewise.
7327 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7328 (rs6000_expand_interleave): Likewise.
7330 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7332 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7333 to qimode could truncate the indices.
7334 * optabs.c (expand_vec_perm_var): Likewise.
7336 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7338 * Makefile.in (OBJS): Add vec-perm-indices.o.
7339 * vec-perm-indices.h: New file.
7340 * vec-perm-indices.c: Likewise.
7341 * target.h (vec_perm_indices): Replace with a forward class
7343 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7344 * optabs.h: Include vec-perm-indices.h.
7345 (expand_vec_perm): Delete.
7346 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7347 (expand_vec_perm_const): Declare.
7348 * target.def (vec_perm_const_ok): Replace with...
7349 (vec_perm_const): ...this new hook.
7350 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7351 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7352 * doc/tm.texi: Regenerate.
7353 * optabs.def (vec_perm_const): Delete.
7354 * doc/md.texi (vec_perm_const): Likewise.
7355 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7356 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7357 expand_vec_perm for constant permutation vectors. Assert that
7358 the mode of variable permutation vectors is the integer equivalent
7359 of the mode that is being permuted.
7360 * optabs-query.h (selector_fits_mode_p): Declare.
7361 * optabs-query.c: Include vec-perm-indices.h.
7362 (selector_fits_mode_p): New function.
7363 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7364 is defined, instead of checking whether the vec_perm_const_optab
7365 exists. Use targetm.vectorize.vec_perm_const instead of
7366 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7367 fit in the vector mode before using a variable permute.
7368 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7369 vec_perm_indices instead of an rtx.
7370 (expand_vec_perm): Replace with...
7371 (expand_vec_perm_const): ...this new function. Take the selector
7372 as a vec_perm_indices rather than an rtx. Also take the mode of
7373 the selector. Update call to shift_amt_for_vec_perm_mask.
7374 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7375 Use vec_perm_indices::new_expanded_vector to expand the original
7376 selector into bytes. Check whether the indices fit in the vector
7377 mode before using a variable permute.
7378 (expand_vec_perm_var): Make global.
7379 (expand_mult_highpart): Use expand_vec_perm_const.
7380 * fold-const.c: Includes vec-perm-indices.h.
7381 * tree-ssa-forwprop.c: Likewise.
7382 * tree-vect-data-refs.c: Likewise.
7383 * tree-vect-generic.c: Likewise.
7384 * tree-vect-loop.c: Likewise.
7385 * tree-vect-slp.c: Likewise.
7386 * tree-vect-stmts.c: Likewise.
7387 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
7389 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
7390 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
7391 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
7392 (aarch64_vectorize_vec_perm_const): ...this new function.
7393 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7394 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7395 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
7396 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
7397 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7398 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7399 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
7401 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
7402 check for NEON modes.
7403 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
7404 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
7405 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
7406 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
7408 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
7409 the old VEC_PERM_CONST conditions.
7410 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
7411 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
7412 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
7413 (ia64_vectorize_vec_perm_const_ok): Merge into...
7414 (ia64_vectorize_vec_perm_const): ...this new function.
7415 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
7416 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
7417 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
7418 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
7419 * config/mips/mips.c (mips_expand_vec_perm_const)
7420 (mips_vectorize_vec_perm_const_ok): Merge into...
7421 (mips_vectorize_vec_perm_const): ...this new function.
7422 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
7423 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
7424 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
7425 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
7426 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
7427 (rs6000_expand_vec_perm_const): Delete.
7428 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
7430 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7431 (altivec_expand_vec_perm_const_le): Take each operand individually.
7432 Operate on constant selectors rather than rtxes.
7433 (altivec_expand_vec_perm_const): Likewise. Update call to
7434 altivec_expand_vec_perm_const_le.
7435 (rs6000_expand_vec_perm_const): Delete.
7436 (rs6000_vectorize_vec_perm_const_ok): Delete.
7437 (rs6000_vectorize_vec_perm_const): New function.
7438 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7439 an element count and rtx array.
7440 (rs6000_expand_extract_even): Update call accordingly.
7441 (rs6000_expand_interleave): Likewise.
7442 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
7443 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
7444 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
7445 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
7446 (rs6000_expand_vec_perm_const): Delete.
7447 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
7448 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7449 (altivec_expand_vec_perm_const_le): Take each operand individually.
7450 Operate on constant selectors rather than rtxes.
7451 (altivec_expand_vec_perm_const): Likewise. Update call to
7452 altivec_expand_vec_perm_const_le.
7453 (rs6000_expand_vec_perm_const): Delete.
7454 (rs6000_vectorize_vec_perm_const_ok): Delete.
7455 (rs6000_vectorize_vec_perm_const): New function. Remove stray
7456 reference to the SPE evmerge intructions.
7457 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
7458 an element count and rtx array.
7459 (rs6000_expand_extract_even): Update call accordingly.
7460 (rs6000_expand_interleave): Likewise.
7461 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
7462 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
7464 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
7466 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7468 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
7469 vector mode and that that mode matches the mode of the data
7471 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
7472 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
7473 directly using expand_vec_perm_1 when forcing selectors into
7475 (expand_vec_perm_var): New function, split out from expand_vec_perm.
7477 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7479 * optabs-query.h (can_vec_perm_p): Delete.
7480 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
7481 * optabs-query.c (can_vec_perm_p): Split into...
7482 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
7483 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
7484 particular selector is valid.
7485 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7486 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7487 (vect_grouped_load_supported): Likewise.
7488 (vect_shift_permute_load_chain): Likewise.
7489 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7490 (vect_transform_slp_perm_load): Likewise.
7491 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7492 (vectorizable_bswap): Likewise.
7493 (vect_gen_perm_mask_checked): Likewise.
7494 * fold-const.c (fold_ternary_loc): Likewise. Don't take
7495 implementations of variable permutation vectors into account
7496 when deciding which selector to use.
7497 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
7498 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
7499 with a false third argument.
7500 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
7501 to test whether the constant selector is valid and can_vec_perm_var_p
7502 to test whether a variable selector is valid.
7504 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7506 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
7507 * optabs-query.c (can_vec_perm_p): Likewise.
7508 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
7509 instead of vec_perm_indices.
7510 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
7511 (vect_gen_perm_mask_checked): Likewise,
7512 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
7513 (vect_gen_perm_mask_checked): Likewise,
7515 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7517 * optabs-query.h (qimode_for_vec_perm): Declare.
7518 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
7519 (qimode_for_vec_perm): ...this new function.
7520 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
7522 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7524 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
7525 does not have a conditional at the top.
7527 2018-01-02 Richard Biener <rguenther@suse.de>
7529 * ipa-inline.c (big_speedup_p): Fix expression.
7531 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7534 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
7537 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
7541 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
7542 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
7543 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
7544 cond_taken_branch_cost 3->4.
7546 2018-01-01 Jakub Jelinek <jakub@redhat.com>
7548 PR tree-optimization/83581
7549 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
7550 TODO_cleanup_cfg if any changes have been made.
7553 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
7554 convert_modes if target mode has the right side, but different mode
7558 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
7559 last argument when extracting from CONCAT. If either from_real or
7560 from_imag is NULL, use expansion through memory. If result is not
7561 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
7562 the parts directly to inner mode, if even that fails, use expansion
7566 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
7567 check for bswap in mode rather than HImode and use that in expand_unop
7570 Copyright (C) 2018 Free Software Foundation, Inc.
7572 Copying and distribution of this file, with or without modification,
7573 are permitted in any medium without royalty provided the copyright
7574 notice and this notice are preserved.