1 2018-04-17 Martin Liska <mliska@suse.cz>
4 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5 in message, remote space in between '_G' and '('.
7 2018-04-17 Jakub Jelinek <jakub@redhat.com>
10 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
11 avx512f_vmcmp<mode>3<round_saeonly_name>,
12 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
13 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
14 avx512f_rndscale<mode><round_saeonly_name>,
15 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
16 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
17 Use %<iptr>2 instead of %2 for -masm=intel.
18 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
19 avx512f_vcvttss2usi<round_saeonly_name>,
20 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
22 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
23 avx512f_vcvttsd2usi<round_saeonly_name>,
24 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
25 Use %q1 instead of %1 for -masm=intel.
26 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
27 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
28 of %3 for -masm=intel.
29 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
31 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
33 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
35 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
36 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
38 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
40 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
41 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
42 %g1 and one with %0 and %1.
43 (avx512er_vmrcp28<mode><round_saeonly_name>,
44 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
46 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
47 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
48 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
49 of %0 and %{%4%} for -masm=intel.
50 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
51 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
52 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
53 order of %0 and %{%5%}%{z%} for -masm=intel.
55 2018-04-17 Jan Hubicka <jh@suse.cz>
58 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
60 2018-04-17 Martin Liska <mliska@suse.cz>
63 * multiple_target.c (create_dispatcher_calls): Set apostrophes
64 for target_clone error message. Make default implementation
65 clone to be a local declaration.
66 (separate_attrs): Add new argument and check for an empty
68 (expand_target_clones): Handle it.
69 (ipa_target_clone): Make redirection just for target_clones
72 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
73 Tom de Vries <tom@codesourcery.com>
76 * omp-expand.c (expand_oacc_for): Add dummy false branch for
77 tiled basic blocks without omp continue statements.
79 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
82 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
83 vec_extract expression as having side effects to make sure it gets
86 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
89 * config/i386/i386.c (get_builtin_code_for_version): Check
92 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
95 * gcc/config.gcc: Support "skylake".
96 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
98 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
99 (processor_target_table): Add "skylake".
100 (ix86_option_override_internal): Add "skylake".
101 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
102 PROCESSOR_CANNONLAKE.
103 (get_builtin_code_for_version): Fix priority for
104 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
105 PROCESSOR_SKYLAKE-AVX512.
106 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
107 (processor_type): Add PROCESSOR_SKYLAKE.
109 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
110 Jason Merrill <jason@redhat.com>
113 * convert.c (convert_to_integer_1): Use direct recursion for
114 enumeral types and types with a precision less than the number
115 of bits in their mode.
117 2018-04-16 Julia Koval <julia.koval@intel.com>
120 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
121 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
123 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
126 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
127 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
128 and -mno-direct-move.
130 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
133 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
134 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
135 Ensure negative shifts result in {0}.
137 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
139 PR rtl-optimization/79916
140 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
141 regs (if any) to define how to gnerate SD moves when LRA is in
144 2018-04-13 Jakub Jelinek <jakub@redhat.com>
146 PR rtl-optimization/85393
147 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
148 * except.c (expand_dw2_landing_pad_for_region): Make static.
149 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
150 a label and unconditional jump to old_bb, rather than
151 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
154 PR rtl-optimization/85376
155 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
156 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
157 instead of a specific value.
159 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
160 Bin Cheng <bin.cheng@arm.com>
162 PR tree-optimization/82965
163 PR tree-optimization/83991
164 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
165 by_profile_only parameter.
166 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
167 information if the loop was predicted to iterate too many times.
168 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
170 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
173 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
176 2018-04-13 Martin Liska <mliska@suse.cz>
177 Jakub Jelinek <jakub@redhat.com>
180 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
181 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
182 * builtins.c (expand_builtin_memory_copy_args): Use
183 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
184 handle dest_addr == pc_rtx.
186 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
189 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
190 asked to not generate direct moves.
191 (fix_trunc<mode>si2_stfiwx): Similar.
192 (fix_trunc<mode>si2_internal): Similar.
194 2018-04-12 Jakub Jelinek <jakub@redhat.com>
197 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
198 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
199 lookup if dest in some wider mode is known to be const0_rtx and
200 if so, record permanent equivalence for it to be ZERO_EXTEND of
201 the narrower mode destination.
203 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
205 * lto-streamer-out.c (output_function): Revert 259346.
206 * omp-expand.c (expand_oacc_for): Likewise.
208 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
210 PR rtl-optimization/85354
211 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
212 * sel-sched.c (sel_global_init): ... here.
214 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
217 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
218 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
219 mode for PE-COFF targets.
220 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
221 (i386_pe_asm_lto_end): Likewise.
222 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
223 (TARGET_ASM_LTO_END): Likewise.
224 * config/i386/winnt.c (saved_debug_info_level): New static variable.
225 (i386_pe_asm_lto_start): New function.
226 (i386_pe_asm_lto_end): Likewise.
228 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
229 Richard Biener <rguenther@suse.de>
232 * lto-streamer-out.c (output_function): Fix CFG loop state before
234 * omp-expand.c (expand_oacc_for): Handle calls to internal
235 functions like regular functions.
237 2018-04-12 Richard Biener <rguenther@suse.de>
240 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
241 for the early LTO debug to properly generate references to it
242 during DIE emission. Do not re-use that for the skeleton for
244 (dwarf2out_early_finish): Likewise.
246 2018-04-12 Jakub Jelinek <jakub@redhat.com>
250 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
251 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
252 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
253 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
254 and output is a reg, avoid creating invalid lowpart subreg, but
255 instead split into a 512-bit move. Don't split if not AVX512VL,
256 input is xmm16+ reg and output is a mem.
257 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
258 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
259 xmm16+ reg and output is a mem.
261 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
263 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
264 also for flag_dwarf2_cfi_asm.
266 2018-04-12 Jakub Jelinek <jakub@redhat.com>
268 PR rtl-optimization/85342
269 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
270 a bool scalar var inside of the loop instead. Don't try to update
271 recog_data.operand after failed apply_change_group.
273 2018-04-12 Tom de Vries <tom@codesourcery.com>
276 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
277 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
278 array with flexible array member as array without given dimension.
279 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
280 argument for undefined param to true.
282 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
285 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
286 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
287 from PowerPC section.
288 * config/rs6000/sysv4.opt (mcall-): Improve help text.
289 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
290 help text that is too long.
291 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
292 help text that is too long.
293 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
294 help text that is too long.
296 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
298 * config/alpha/alpha.md (stack_probe_internal): Rename
299 from "probe_stack". Update all callers.
301 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
303 PR rtl-optimization/84566
304 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
305 sched_macro_fuse_insns.
307 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
310 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
311 (compute_block_dependences): ... from here.
313 2018-04-11 Jakub Jelinek <jakub@redhat.com>
315 PR tree-optimization/85331
316 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
317 from int to HOST_WIDE_INT.
319 2018-04-11 Martin Jambor <mjambor@suse.cz>
322 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
323 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
324 not the same as the source val.
325 (cgraph_edge_brings_value_p): New parameter.
326 (gather_edges_for_value): Pass destination value to
327 cgraph_edge_brings_value_p.
328 (perhaps_add_new_callers): Likewise.
329 (get_info_about_necessary_edges): Likewise and exclude values brought
330 only by self-recursive edges.
331 (create_specialized_node): Redirect only clones of self-calling edges.
332 (+self_recursive_pass_through_p): New function.
333 (find_more_scalar_values_for_callers_subset): Use it.
334 (find_aggregate_values_for_callers_subset): Likewise.
335 (known_aggs_to_agg_replacement_list): Removed.
336 (decide_whether_version_node): Re-calculate known constants for all
337 remaining context clones.
339 2018-04-11 Richard Biener <rguenther@suse.de>
342 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
343 from early DWARF output.
344 (dwarf2out_early_finish): Output line info unconditionally into
345 early DWARF and add reference to it.
347 2018-04-11 Jakub Jelinek <jakub@redhat.com>
350 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
351 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
352 other than V2DFmode using iptr mode attribute.
353 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
355 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
357 PR rtl-optimization/84659
358 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
360 2018-04-11 Jakub Jelinek <jakub@redhat.com>
363 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
365 (output_loc_list): Pass address of a dummy size variable even in the
366 locview handling loop.
367 (index_location_lists): Add comment on why skip_loc_list_entry can't
370 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
373 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
376 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
379 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
380 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
381 and -mstring-compare-inline-limit.
383 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
386 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
387 for stack clash protection in a register whenever we need it to be in
390 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
392 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
393 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
395 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
398 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
400 (mlong-double-): Ditto.
401 * config/rs6000/sysv4.opt (msdata=): Ditto.
404 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
406 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
407 erroneous entries for
408 "vector int vec_ldl (int, long int *)", and
409 "vector unsigned int vec_ldl (int, unsigned long int *)".
410 Add comments and entries for
411 "vector bool char vec_ldl (int, bool char *)",
412 "vector bool short vec_ldl (int, bool short *)",
413 "vector bool int vec_ldl (int, bool int *)",
414 "vector bool long long vec_ldl (int, bool long long *)",
415 "vector pixel vec_ldl (int, pixel *)",
416 "vector long long vec_ldl (int, long long *)",
417 "vector unsigned long long vec_ldl (int, unsigned long long *)".
418 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
419 type tree bool_long_long_type_node and correct definition of
420 bool_V2DI_type_node to make reference to this new type tree.
421 (rs6000_mangle_type): Replace erroneous reference to
422 bool_long_type_node with bool_long_long_type_node.
423 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
424 comments to emphasize sign distinctions for char and int types and
425 replace RS6000_BTI_bool_long constant with
426 RS6000_BTI_bool_long_long constant. Also add comment to restrict
427 use of RS6000_BTI_pixel.
428 (bool_long_type_node): Remove this macro definition.
429 (bool_long_long_type_node): New macro definition
431 2018-04-10 Jakub Jelinek <jakub@redhat.com>
433 PR rtl-optimization/85300
434 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
435 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
436 simplify_unary_operation fails.
438 2018-04-10 Martin Liska <mliska@suse.cz>
440 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
441 cgraph_edge and ipa_ref.
443 2018-04-10 Jakub Jelinek <jakub@redhat.com>
448 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
449 computation of the VEC_MERGE selector from mask.
450 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
451 Fix decoding of the VEC_MERGE selector into mask.
453 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
455 PR tree-optimization/85286
456 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
458 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
460 * final.c (final_1): Set insn_last_address as well as
461 insn_current_address.
463 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
466 * explow.c (emit_stack_probe): Call validize_mem on memory location
467 before passing it to gen_probe_stack. Create address operand and
468 legitimize it for the probe_stack_address case.
470 2018-04-09 Jan Hubicka <jh@suse.cz>
473 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
474 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
475 * tree.c (free_lang_data_in_type): Fix handling of binfos;
477 (free_lang_data): Rebuild type inheritance graph.
479 2018-04-09 Martin Sebor <msebor@redhat.com>
481 * invoke.texi (-finline-small-functions): Mention other optimization
483 (-findirect-inlining, -fpartial-inlining): Same.
484 (-finline-functions-called-once): Same.
485 (-freorder-blocks-and-partition): Same.
487 2018-04-09 Jan Hubicka <jh@suse.cz>
490 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
491 jumps; choose last target that matches the criteria (i.e.
492 no partition changes for non-crossing jumps).
493 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
494 support for redirecting crossing jumps to non-crossing.
496 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
498 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
499 also for naked functions.
501 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
503 * config/arc/arc.md (add_shift): New pattern.
504 (add_shift2): Likewise.
505 (sub_shift): Likewise.
506 (sub_shift_cmp0_noout): Likewise.
507 (compare_si_ashiftsi): Likewise.
508 (xbfu_cmp0_noout): New combine pattern.
509 (xbfu_cmp0"): Likewise.
510 (movsi_set_cc_insn): Place the predicable variant first.
511 (commutative_binary_cmp0_noout): Remove clobber.
512 (commutative_binary_cmp0): New pattern.
513 (noncommutative_binary_cmp0): Likewise.
514 (noncommutative_binary_cmp0_noout): Likewise.
515 (noncommutative_binary_comparison_result_used): Removed.
516 (rsub_cmp0): New pattern.
517 (rsub_cmp0_noout): Likewise.
518 (extzvsi): Changed, keep only meaningful variants.
519 (SQH, SEZ): New iterators.
520 (SQH_postfix): New mode attribute.
521 (SEZ_prefix): New code attribute.
522 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
523 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
524 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
526 (noncommutative_operator): Check the availability of barrel
529 2018-04-09 Richard Biener <rguenther@suse.de>
531 PR tree-optimization/85284
532 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
533 Only use the niter constraining form of simple_iv when the exit
536 2018-04-09 Tom de Vries <tom@codesourcery.com>
539 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
540 (define_expand "*memory_barrier"): New define_expand.
541 (define_insn "memory_barrier"): New insn.
543 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
545 PR rtl-optimization/80463
546 PR rtl-optimization/83972
547 PR rtl-optimization/83480
549 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
550 correct producer for the insn.
551 (tidy_control_flow): Fixup seqnos in case of debug insns.
553 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
555 PR rtl-optimization/83913
557 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
558 different sched-times when merging exprs.
560 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
562 PR rtl-optimization/83962
564 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
565 tidy_fallthru_edge and tidy_control_flow.
567 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
569 PR rtl-optimization/83530
571 * sel-sched.c (force_next_insn): New global variable.
572 (remove_insn_for_debug): When force_next_insn is true, also leave only
573 next insn in the ready list.
574 (sel_sched_region): When the region wasn't scheduled, make another pass
575 over it with force_next_insn set to 1.
577 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
579 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
581 * config/nds32/constants.md (unspec_volatile_element): Add enum values
582 for interrupt control.
583 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
584 functions for interrupt control.
585 * config/nds32/nds32-intrinsic.md: Likewise.
586 * config/nds32/nds32_intrinsic.h: Likewise.
587 * config/nds32/nds32.h (nds32_builtins): Likewise.
589 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
591 * config/nds32/nds32.c (nds32_init_machine_status,
592 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
593 strict_aligned_p field.
594 (nds32_expand_to_rtl_hook): New function.
595 (TARGET_EXPAND_TO_RTL_HOOK): Define.
596 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
598 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
599 Chung-Ju Wu <jasonwucj@gmail.com>
601 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
602 * config/nds32/nds32-n7.md: New file.
603 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
604 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
606 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
607 * config/nds32/nds32.md (pipeline_model): Add n7.
608 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
609 * config/nds32/pipelines.md: Include n7 settings.
611 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
612 Chung-Ju Wu <jasonwucj@gmail.com>
614 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
615 * config/nds32/nds32-e8.md: New file.
616 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
617 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
619 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
620 * config/nds32/nds32.md (pipeline_model): Add e8.
621 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
622 * config/nds32/pipelines.md: Include e8 settings.
624 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
625 Chung-Ju Wu <jasonwucj@gmail.com>
627 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
628 * config/nds32/nds32-n8.md: New file.
629 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
630 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
632 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
633 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
634 * config/nds32/nds32.md (pipeline_model): Add n8.
635 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
636 * config/nds32/pipelines.md: Include n8 settings.
638 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
639 Chung-Ju Wu <jasonwucj@gmail.com>
641 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
642 * config/nds32/nds32-n9-2r1w.md: New file.
643 * config/nds32/nds32-n9-3r2w.md: New file.
644 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
645 nds32_register_ports): New or modify for cpu n9.
646 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
648 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
649 * config/nds32/nds32-utils.c: New file.
650 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
651 TARGET_MUL_SLOW): Define.
652 * config/nds32/nds32.md (pipeline_model): New attribute.
653 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
654 New options that support cpu n9.
655 * config/nds32/pipelines.md: Include n9 settings.
656 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
658 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
660 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
661 information if necessary.
662 (output_cond_branch_compare_zero): Likewise.
663 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
664 (nds32_target_alignment): Refine for alignment.
665 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
666 (FUNCTION_BOUNDARY): Modify.
667 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
669 * config/nds32/nds32.opt (malways-align, malign-functions): New.
671 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
673 * config/nds32/constants.md (unspec_volatile_element): Add values for
674 TLB operation and data prefetch.
675 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
676 functions for TLB operation and data prefetch.
677 * config/nds32/nds32-intrinsic.md: Likewise.
678 * config/nds32/nds32_intrinsic.h: Likewise.
679 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
680 (nds32_print_operand): Likewise.
681 * config/nds32/nds32.h (nds32_builtins): Likewise.
683 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
684 Andrew Pinski <pinsika@gcc.gnu.org>
687 * match.pd: Use constant_boolean_node of correct type instead of
688 boolean_true_node or boolean_false_node for simplifying
689 pointer comparisons to zero.
691 2018-04-07 Jakub Jelinek <jakub@redhat.com>
693 PR tree-optimization/80021
694 * tree.c (verify_type_variant): Make error call in verify_variant_match
695 translatable and remove final full stop.
697 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
699 * config/nds32/constants.md (unspec_volatile_element): Add
700 UNSPEC_VOLATILE_EH_RETURN.
701 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
702 nds32_output_stack_pop): Support dwarf exception handling process.
703 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
704 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
705 exception handling process.
706 (nds32_compute_stack_frame): Likewise.
707 (nds32_return_addr_rtx): Likewise.
708 (nds32_initial_elimination_offset): Likewise.
709 (nds32_expand_prologue): Likewise.
710 (nds32_expand_epilogue): Likewise.
711 (nds32_dynamic_chain_address): New function.
712 * config/nds32/nds32.h (machine_function): Add fields for dwarf
714 (DYNAMIC_CHAIN_ADDRESS): Define.
715 (EH_RETURN_DATA_REGNO): Define.
716 (EH_RETURN_STACKADJ_RTX): Define.
717 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
718 patterns for dwarf exception handling.
720 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
722 * config/nds32/nds32.h: Clean up obsolete macros.
724 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
726 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
727 Add enum values for particular instructions.
728 * config/nds32/nds32-intrinsic.c: Implementation of expanding
729 particular intrinsic functions.
730 * config/nds32/nds32-intrinsic.md: Likewise.
731 * config/nds32/nds32_intrinsic.h: Likewise.
732 * config/nds32/nds32.h (nds32_builtins): Likewise.
733 * config/nds32/nds32.md (type): Add pbsad and pbsada.
734 (btst, ave): New patterns for particular instructions.
736 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
738 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
739 Add enum values for atomic load/store and memory sync.
740 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
742 * config/nds32/nds32-intrinsic.md: Likewise.
743 * config/nds32/nds32_intrinsic.h: Likewise.
744 * config/nds32/nds32.h (nds32_builtins): Likewise.
746 2018-04-07 Jakub Jelinek <jakub@redhat.com>
748 PR tree-optimization/85257
749 * fold-const.c (native_encode_vector): If not all elts could fit
750 and off is -1, return 0 rather than offset.
751 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
752 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
753 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
754 adjust buffer in native_interpret_expr call.
756 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
758 * config/nds32/constants.md (unspec_volatile_element): Add cache
760 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
761 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
762 * config/nds32/nds32.c (nds32_cctl_names): New.
763 (nds32_print_operand): Handle cache control register names.
764 * config/nds32/nds32.h (nds32_builtins): New enum values.
765 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
767 * config/nds32/nds32.md (type): Add mmu.
768 * config/nds32/pipelines.md (simple_insn): Add mmu.
770 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
772 * config/nds32/nds32.md (type): Remove call.
773 * config/nds32/pipelines.md (simple_insn): Likewise.
775 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
777 * config/nds32/constants.md (unspec_volatile_element): Add
778 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
779 UNSPEC_VOLATILE_FMFCFG.
780 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
781 description for fmfcfg and fmfcsr.
782 (bdesc_1arg): Add fmtcsr.
783 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
784 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
785 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
786 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
787 unspec_fmfcfg): New patterns.
788 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
789 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
790 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
791 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
792 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
793 __nds32__fmfcfg): Define.
795 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
797 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
798 intrinsic register names.
799 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
800 intrinsic register enum values and macros.
802 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
804 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
805 for load/store addressing form.
806 (nds32_print_operand_address): Likewise.
808 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
811 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
812 based on LABEL_REF. Remove useless assertion.
813 (pic_address_needs_scratch): Fix formatting.
814 (sparc_legitimize_pic_address): Minor tweaks.
815 (sparc_delegitimize_address): Adjust assertion accordingly.
816 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
817 into symbolic_operand.
818 (movsi_high_pic_label_ref): Likewise.
819 (movsi_lo_sum_pic_label_ref): Likewise.
820 (movdi_pic_label_ref): Likewise.
821 (movdi_high_pic_label_ref): Likewise.
822 (movdi_lo_sum_pic_label_ref): Likewise.
824 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
826 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
827 custom LIB_SPEC setup.
829 2018-04-06 Ruslan Bukin <br@bsdpad.com>
830 Kito Cheng <kito.cheng@gmail.com>
832 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
833 * config/riscv/freebsd.h: New.
835 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
837 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
838 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
841 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
842 Kito Cheng <kito.cheng@gmail.com>
844 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
845 nds32_output_call, nds32_symbol_binds_local_p): New functions.
846 * config/nds32/nds32-protos.h (nds32_output_call,
847 nds32_output_return): Declare.
848 * config/nds32/nds32.md: Refine all the call and return patterns.
850 2018-04-06 Jakub Jelinek <jakub@redhat.com>
853 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
854 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
856 PR rtl-optimization/84872
857 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
858 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
861 2018-04-06 Tamar Christina <tamar.christina@arm.com>
863 * expr.c (copy_blkmode_to_reg): Revert 254862.
864 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
866 2018-04-06 Richard Biener <rguenther@suse.de>
869 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
870 after seeing a component reference with an adjacent field. Treat
871 refs to arrays at struct end of external decls similar to
872 refs to unconstrained commons.
874 2018-04-06 Jakub Jelinek <jakub@redhat.com>
877 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
878 look through SAVE_EXPRs with non-side-effects argument. Adjust
880 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
883 2018-04-06 Richard Biener <rguenther@suse.de>
886 * alias.c (find_base_term): New wrapper around find_base_term
887 unwinding CSELIB_VAL_PTR changes.
888 (find_base_term): Do not restore CSELIB_VAL_PTR during the
891 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
893 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
895 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
896 constant definitions.
897 ("nop"): lr 0,0 -> nopr r0
898 ("nop_lr0", "nop_lr1"): New insn definitions.
900 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
902 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
903 NDS32_V3PUSH_AVAILABLE_P macro.
905 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
906 Chung-Ju Wu <jasonwucj@gmail.com>
908 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
909 (nds32*-*-*): Add float and fpu_config into supported_defaults.
910 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
911 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
912 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
913 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
914 * config/nds32/constraints.md: New constraints and checking for hard
916 * config/nds32/iterators.md: New mode iterator and attribute for hard
918 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
920 * config/nds32/nds32-fpu.md: New file.
921 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
922 deal with hard float code generation.
923 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
925 (abi_type, float_reg_number): New enum type.
926 * config/nds32/nds32-predicates.c: New predicates for hard float.
927 * config/nds32/nds32-protos.h: Declare functions for hard float.
928 * config/nds32/nds32.c: Implementation for hard float configuration.
929 * config/nds32/nds32.h: Definitions for hard float configuration.
930 * config/nds32/nds32.md: Include hard float machine description and
931 modify patterns for hard float configuration.
932 * config/nds32/nds32.opt: New options for hard float configuration.
933 * config/nds32/predicates.md: New predicates for hard float
936 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
938 * common/config/nds32/nds32-common.c
939 (nds32_option_optimization_table): Enable -mreleax-hint by default.
941 2018-04-05 Jakub Jelinek <jakub@redhat.com>
944 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
945 CONSTRUCTOR_ELT (ctor, ...)->value.
947 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
950 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
952 2018-04-05 Tom de Vries <tom@codesourcery.com>
955 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
958 2018-04-05 Shiva Chen <shiva0217@gmail.com>
959 Kito Cheng <kito.cheng@gmail.com>
961 * config/nds32/constraints.md (U33): Fine-tune checking condition.
962 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
963 * config/nds32/nds32.h (nds32_16bit_address_type): Add
964 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
966 2018-04-05 Shiva Chen <shiva0217@gmail.com>
967 Kito Cheng <kito.cheng@gmail.com>
969 * config/nds32/constraints.md (Ufe): New memory constraint.
970 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
971 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
972 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
974 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
975 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
977 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
979 * config/nds32/nds32.md: Use optimize_size in the condition for
980 alu-shift instructions.
982 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
984 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
986 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
988 * config/nds32/nds32.md (negsi2): Refine pattern.
990 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
991 Chung-Ju Wu <jasonwucj@gmail.com>
993 * config/nds32/iterators.md (shift_rotate): New code iterator.
994 (shift): New code attribute.
995 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
996 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
997 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
998 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
1000 (andsi3, *andsi3): Ditto.
1001 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
1002 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
1003 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
1004 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
1005 nds32_ior_operand, nds32_xor_operand): New predicates.
1007 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1009 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
1010 (addsi3, subsi3): ... this.
1012 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1014 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
1016 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1018 * config/nds32/nds32.md: Adjust indention.
1020 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
1022 * config/nds32/nds32.md (feature): New attribute.
1024 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
1026 * config/nds32/nds32.md (subtype): New attribute.
1028 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
1031 * config/arm/arm-builtins.c (arm_expand_builtin): Change
1032 expansion to perform a bitwise AND of the argument followed by a
1033 boolean negation of the result.
1035 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
1037 PR rtl-optimization/84878
1038 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
1039 the basic block. Assert the use reference is not artificial and that
1040 it has an associated insn.
1042 2018-04-04 Michael Matz <matz@suse.de>
1044 * builtins.c (compute_objsize): Pass correct operand
1045 to array_at_struct_end_p.
1047 2018-04-04 Richard Biener <rguenther@suse.de>
1050 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
1051 from contexts for DINFO_LEVEL_TERSE and below.
1053 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1055 * config/nds32/nds32-doubleword.md (move_<mode>): Require
1056 resiter_operand condition.
1057 * config/nds32/nds32.md (*move<mode>): Ditto.
1059 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1060 Monk Chiang <sh.chiang04@gmail.com>
1062 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
1064 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1066 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
1068 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1069 Kito Cheng <kito.cheng@gmail.com>
1071 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
1072 nds32_cond_code_str, output_cond_branch,
1073 output_cond_branch_compare_zero, nds32_expand_cbranch,
1074 nds32_expand_cstore, nds32_expand_movcc,
1075 nds32_output_cbranchsi4_equality_zero,
1076 nds32_output_cbranchsi4_equality_reg,
1077 nds32_output_cbranchsi4_equality_reg_or_const_int,
1078 nds32_output_cbranchsi4_greater_less_zero: New functions.
1079 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
1080 nds32_expand_cstore, nds32_expand_movcc,
1081 nds32_output_cbranchsi4_equality_zero,
1082 nds32_output_cbranchsi4_equality_reg,
1083 nds32_output_cbranchsi4_equality_reg_or_const_int,
1084 nds32_output_cbranchsi4_greater_less_zero): Declare.
1085 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
1086 nds32_rimm11s_operand): New predicates.
1087 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
1088 * config/nds32/nds32.md: Rewrite all the branch and conditional move
1091 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
1093 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
1094 * config/nds32/nds32.md: Ditto.
1095 * config/nds32/pipelines.md: Ditto.
1097 2018-04-04 Richard Biener <rguenther@suse.de>
1099 PR tree-optimization/85168
1100 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
1101 propagating abnormals.
1103 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1105 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
1107 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
1108 Kito Cheng <kito.cheng@gmail.com>
1110 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
1111 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
1112 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
1113 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
1114 * config/nds32/nds32.md (sibcall_internal): New.
1115 (sibcall_register): Remove.
1116 (sibcall_immediate): Remove.
1117 (sibcall_value_internal): New.
1118 (sibcall_value_register): Remove.
1119 (sibcall_value_immediate): Remove.
1120 * config/nds32/predicates.md (nds32_general_register_operand): New.
1121 (nds32_call_address_operand): New.
1123 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1125 PR rtl-optimization/85167
1126 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1127 bb_defs if *split_p, instead preinitialize it to NULL.
1129 PR tree-optimization/85156
1130 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1131 evaluating the argument multiple times.
1133 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1135 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1137 (_mm_cvtpd_ps): Likewise.
1138 (_mm_cvttpd_epi32): Likewise.
1139 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1140 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1141 vector, pixel, and bool following altivec.h include.
1143 2018-04-03 Martin Sebor <msebor@redhat.com>
1145 * doc/extend.texi (Common Function Attributes): Clarify.
1146 (const attribute): Likewise.
1147 (pure attribute): Likewise.
1149 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1152 * config/i386/i386.c (ix86_expand_vector_set): Use
1153 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1155 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1157 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1158 instructions when changing rounding bits to preserve precision bits
1159 in the x87 control word.
1161 2018-04-03 Martin Liska <mliska@suse.cz>
1163 PR tree-optimization/82491
1164 * rtl.h (strip_offset_and_add): Replace += suboffset with
1165 poly_uint64 () + suboffset.
1167 2018-03-29 Martin Liska <mliska@suse.cz>
1168 Martin Jambor <mjambor@suse.cz>
1171 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1172 param_type is not an integral or pointer type.
1174 2018-04-03 Richard Biener <rguenther@suse.de>
1176 * sese.h (recompute_all_dominators): Remove.
1178 2018-04-02 Martin Sebor <msebor@redhat.com>
1180 * doc/invoke.texi (-Wrestrict): Fix typos.
1182 2018-04-02 Jim Wilson <jimw@sifive.com>
1184 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1185 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1186 (<optab>di3, <optab>si3_extend): Likewise.
1187 (<optab>si3_mask, <optab>si3_mask_1): New.
1188 (<optab>di3_mask, <optab>di3_mask_1): New.
1189 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1190 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1191 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1193 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1195 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1198 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1200 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1201 (nds32_canonicalize_comparison): New function.
1203 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1204 Kito Cheng <kito.cheng@gmail.com>
1205 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1207 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1208 * config/nds32/constants.md (unspec_volatile_element): Add
1209 UNSPEC_VOLATILE_RELAX_GROUP.
1210 * config/nds32/nds32-relax-opt.c: New file.
1211 * config/nds32/nds32-predicates.c
1212 (nds32_symbol_load_store_p): New function.
1213 * config/nds32/nds32-protos.h
1214 (nds32_symbol_load_store_p): Declare function.
1215 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1216 * config/nds32/nds32.c
1217 (nds32_register_pass): New function to register pass.
1218 (nds32_register_passes): New function to register passes.
1219 * config/nds32/nds32.md (relax_group): New pattern.
1220 * config/nds32/nds32.opt (mrelax-hint): New option.
1221 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1223 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1225 * config/nds32/t-nds32: Modify files dependency.
1227 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1229 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1230 (PROFILE_HOOK): Define its implementation.
1232 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1234 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1235 type and 32-bit size.
1237 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1240 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1241 (V_128_256): New mode iterator.
1242 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1243 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1244 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1246 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1247 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1249 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1252 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1253 NaN inputs correctly.
1255 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1258 * config/rs6000/vsx.md (??r): New mode attribute.
1259 (*vsx_mov<mode>_64bit): Use it.
1260 (*vsx_mov<mode>_32bit): Likewise.
1262 2018-03-30 Martin Sebor <msebor@redhat.com>
1264 PR tree-optimization/84818
1265 * builtins.c (check_access): Use warning_n.
1267 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1270 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1272 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1275 2018-03-30 Julia Koval <julia.koval@intel.com>
1278 * x86-tune.def (movx, partial_reg_dependency): Enable for
1281 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1284 * lra-constraints.c (process_alt_operands): Move setting
1285 this_alternative_matches below.
1287 2018-03-29 Martin Liska <mliska@suse.cz>
1290 * doc/invoke.texi: Document how LTO works with debug info.
1291 Describe auto-load support of binutils. Mention 'x86-64'
1292 as valid option value of -march option.
1294 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1296 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1299 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1300 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1301 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1304 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1307 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1308 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1309 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1310 for RS6000_BTM_POWERPC64.
1311 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1312 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1313 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1318 2018-03-28 Carl Love <cel@us.ibm.com>
1321 2017-09-27 Carl Love <cel@us.ibm.com>
1323 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1324 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1325 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1328 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1330 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1331 instead of __vector bool.
1332 (_mm_max_pu8): Likewise.
1333 (_mm_min_pi16): Likewise.
1335 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1338 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1339 (DIVWEUO): Likewise.
1341 (DIVDEUO): Likewise.
1342 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1343 DIVWEUO and DIVDEUO.
1344 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1345 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1346 (div_extend): Likewise.
1347 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1349 (__builtin_divweuo): Likewise.
1350 (__builtin_divdeo): Likewise.
1351 (__builtin_divdeuo): Likewise.
1353 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1356 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1357 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1359 PR tree-optimization/82004
1360 * gimple-match-head.c (optimize_pow_to_exp): New function.
1361 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1362 Don't fold to exp if optimize_pow_to_exp is false.
1364 2018-03-28 Martin Liska <mliska@suse.cz>
1367 * calls.c (initialize_argument_information): Fix trailing space.
1368 * common.opt: Fix typo and provide better explanation for
1369 -fsanitize-coverage option.
1370 * config/i386/i386.opt: Fix typo.
1372 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1373 Martin Liska <mliska@suse.cz>
1376 * gimplify.c (asan_poison_variable): Don't do the check for
1377 gimplify_omp_ctxp here.
1378 (gimplify_decl_expr): Do it here.
1379 (gimplify_target_expr): Likewise.
1381 2018-03-28 Martin Liska <mliska@suse.cz>
1384 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1385 chkp_type_bounds_count if MPX is not enabled.
1387 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1389 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1391 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1394 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1395 function to create the function decl for complex long double
1396 multiply and divide for -mabi=ieeelongdouble.
1397 (init_float128_ieee): Call it.
1399 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1402 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1403 -fcf-protection=branch -mibt.
1404 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1406 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1409 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1411 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1414 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1415 extern array declarations.
1417 2018-03-27 Richard Biener <rguenther@suse.de>
1420 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1421 explicit single_use checks.
1423 2018-03-27 Richard Biener <rguenther@suse.de>
1425 PR tree-optimization/85082
1426 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1429 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1431 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1432 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1433 Turn on fasynchronous-unwind-tables and funwind-tables.
1435 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1438 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1439 (*bmi_blsr_<mode>_ccz): Ditto.
1441 2018-03-26 Tom de Vries <tom@codesourcery.com>
1443 PR tree-optimization/85063
1444 * omp-general.c (offloading_function_p): New function. Factor out
1446 * omp-offload.c (pass_omp_target_link::gate): ... here.
1447 * omp-general.h (offloading_function_p): Declare.
1448 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1449 with attribute omp declare target for offloading functions.
1451 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1453 PR tree-optimization/84005
1454 * tree-data-ref.h (get_base_for_alignment): Declare.
1455 * tree-data-ref.c (get_base_for_alignment_1): New function.
1456 (get_base_for_alignment): Likewise.
1457 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1458 get_base_for_alignment to find a suitable base object, instead
1459 of always using drb->base_address.
1461 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1464 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1465 known size by default.
1467 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1470 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1471 and non BLKmode operands.
1473 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1476 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1477 Clean up attributes.
1479 2018-03-23 Richard Biener <rguenther@suse.de>
1482 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1483 we are going to emit early debug for LTO.
1485 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1488 * function.c (match_asm_constraints_1): Don't optimize if input
1489 doesn't satisfy general_operand predicate for output's mode.
1492 * alias.c (write_dependence_p): Don't require for x_canonicalized
1493 non-VOIDmode if x has VOIDmode.
1496 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1497 just don't try to optimize it rather than assert it never happens.
1499 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1501 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1502 macro expansions for definition of ST_INTERNAL_<mode> and
1503 LD_INTERNAL_<mode> builtins.
1504 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1506 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1508 (altivec_expand_st_builtin): Likewise.
1509 (altivec_expand_builtin): Remove calls to deleted functions.
1510 (rs6000_address_for_altivec): Delete this function.
1511 * config/rs6000/vector.md: Remove expands for
1512 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1514 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1517 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1518 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1519 re-computing once computed.
1520 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1521 (arm_init_machine_status): Initialize
1522 machine->static_chain_stack_bytes.
1524 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1527 * doc/extend.texi: Add four new prototypes for vec_ld.
1528 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1529 definitions for more logical presentation.
1530 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1531 entries for V1TI variants of __builtin_altivec_ld builtin.
1532 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1533 handling of V1TI variant of LVX icode pattern.
1534 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1535 (rs6000_gimple_fold_builtin): Likewise.
1536 (altivec_init_builtins): Add code to define
1537 __builtin_altivec_lvx_v1ti function.
1539 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1542 * function.c (match_asm_constraints_1): Don't do the optimization
1543 if input isn't a REG, SUBREG, MEM or constant.
1545 2018-03-22 Tom de Vries <tom@codesourcery.com>
1547 PR tree-optimization/84956
1548 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1549 bb_has_abnormal_pred.
1551 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1554 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1555 DECL_INITIAL (decl) to decl at the end.
1556 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1559 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1561 * doc/extend.texi (__builtin_tgmath): Document when complex
1562 integer types are treated as _Complex _Float64.
1564 2018-03-21 Tom de Vries <tom@codesourcery.com>
1566 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1568 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1570 PR tree-optimization/84960
1571 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1572 if it is ENTRY block, move them into single succ of ENTRY in that case.
1574 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1576 PR tree-optimization/84811
1577 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1578 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1579 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1580 (known_subrange_p): Update accordingly. Cast each value involved
1581 in the size comparison, rather than casting the result of the
1584 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1586 PR tree-optimization/84982
1587 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1588 by flipping the least significant bit rather than all bits from
1589 bitpos to bitpos + bitsize - 1.
1591 2018-03-21 Nathan Sidwell <nathan@acm.org>
1593 * doc/extend.texi (Deprecated Features): Remove mention of
1594 long-deleted deprecations.
1596 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1599 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1600 * configure: Regenerate.
1602 2018-03-21 Tom de Vries <tom@codesourcery.com>
1604 PR tree-optimization/83126
1605 * tree-parloops.c (num_phis): New function.
1606 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1608 2018-03-21 Nathan Sidwell <nathan@acm.org>
1610 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1611 mention anon-struct/union members and trailing attributes.
1613 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1615 PR tree-optimization/84969
1616 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1617 builtin memset partitions if they set different rhs values.
1619 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1621 PR rtl-optimization/84989
1622 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1623 VEC_DUPLICATE with scalar result mode.
1625 2018-03-21 Martin Liska <mliska@suse.cz>
1628 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1629 not intended return statement.
1631 2018-03-21 Martin Liska <mliska@suse.cz>
1634 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1635 (chkp_find_bound_slots_1): Limit number of iterations.
1637 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1640 * Minor grammar fixes for x86 options.
1642 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1645 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1646 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1648 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1651 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1652 (altivec_lvx_<mode>_1op): Likewise.
1653 (altivec_stvx_<mode>_2op): Likewise.
1654 (altivec_stvx_<mode>_1op): Likewise.
1655 (altivec_lvx_<VM2:mode>): New define_expand.
1656 (altivec_stvx_<VM2:mode>): Likewise.
1657 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1658 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1659 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1660 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1661 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1662 (rs6000_gen_lvx): Likewise.
1663 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1664 (altivec_expand_stv_builtin): Likewise.
1665 (altivec_expand_builtin): Likewise.
1666 * config/rs6000/vector.md: Likewise.
1668 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1671 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1674 2018-03-20 Richard Biener <rguenther@suse.de>
1677 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1678 sign-conversions as zero, fall back to standard scalar_stmt
1681 2018-03-20 Martin Liska <mliska@suse.cz>
1684 * predict.c (rebuild_frequencies): Handle case when we have
1685 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1687 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1690 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1691 flag_section_anchors.
1692 * varasm.c (use_blocks_for_decl_p): Remove hack for
1693 dw2_force_const_mem.
1696 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1698 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1699 be created, use lowpart_subreg of operands[0] rather than operands[0]
1701 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1702 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1703 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1704 and n constraint instead of aarch64_shift_imm_di and Usd.
1705 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1706 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1708 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1711 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1712 to favor GPR over NEON registers.
1713 (<shift>di3_neon): Likewise.
1715 2018-03-20 Tom de Vries <tom@codesourcery.com>
1718 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1719 (nvptx_process_pars): Emit bar.sync asap and alap.
1721 2018-03-20 Tom de Vries <tom@codesourcery.com>
1724 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1725 seen_label if seen_label is already set.
1727 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1730 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1731 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1732 Use 1U instead of 1. Formatting fixes.
1735 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1736 instead of TREE_TYPE (s1) for the return value.
1738 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1740 PR tree-optimization/84946
1741 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1742 bitsize + bitsize in poly_uint64 rather than poly_int64.
1745 * dwarf2asm.c: Include fold-const.c.
1746 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1747 of decl rather than decl itself.
1749 PR rtl-optimization/84643
1750 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1752 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1755 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1756 calling assemble_variable.
1758 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1761 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1762 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1764 2018-03-19 Jim Wilson <jimw@sifive.com>
1767 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1768 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1769 (riscv_first_stack_step): Likewise.
1770 (riscv_option_override): Use STACK_BOUNDARY instead of
1772 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1774 (BIGGEST_ALIGNMENT): Set to 128.
1775 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1776 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1779 2018-03-19 Richard Biener <rguenther@suse.de>
1781 PR tree-optimization/84933
1782 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1783 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1785 2018-03-19 Richard Biener <rguenther@suse.de>
1787 PR tree-optimization/84859
1788 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1789 (cond_if_else_store_replacement): Perform sinking operation on
1790 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1791 Generalize what a BB with a single eligible store is.
1793 2018-03-19 Richard Biener <rguenther@suse.de>
1795 PR tree-optimization/84929
1796 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1797 chrec_is_positive against non-chrec arg.
1799 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1802 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1804 2018-03-18 Martin Liska <mliska@suse.cz>
1806 PR rtl-optimization/84635
1807 * regrename.c (build_def_use): Use matches_mode only when
1810 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1812 PR tree-optimization/84913
1813 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1814 vectorize chains of COND_EXPRs.
1816 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1818 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1820 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1822 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1824 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1826 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1828 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1829 Kito Cheng <kito.cheng@gmail.com>
1831 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1832 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1833 (nds32_adjust_reg_alloc_order): New function.
1834 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1836 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1838 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1839 nds32_print_operand, nds32_print_operand_address): Use
1840 HOST_WIDE_INT_PRINT_DEC instead.
1842 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1844 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1846 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1849 * config/i386/i386.c (initial_ix86_tune_features,
1850 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1852 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1853 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1854 rather than 1u << ix86_tune. Formatting fix.
1855 (ix86_option_override_internal): Change ix86_arch_mask from
1856 unsigned int to unsigned HOST_WIDE_INT, initialize to
1857 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1858 (ix86_function_specific_restore): Likewise.
1860 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1863 * postreload.c (reload_combine_recognize_pattern): Perform
1864 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1865 truncate_int_for_mode the result for the destination's mode.
1868 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1869 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1871 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1874 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1877 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1878 regno_allocno_class_array and sorted_pseudos.
1879 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1880 insns where regno is used.
1882 2018-03-16 Martin Liska <mliska@suse.cz>
1885 * multiple_target.c (create_dispatcher_calls): Redirect
1886 reference in the symbol table.
1888 2018-03-16 Martin Liska <mliska@suse.cz>
1891 * multiple_target.c (create_dispatcher_calls): Redirect also
1894 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1898 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1899 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1900 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1902 2018-03-16 Julia Koval <julia.koval@intel.com>
1904 * doc/invoke.texi (Skylake Server): Add CLWB.
1905 Cannonlake): Remove CLWB.
1907 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1909 PR tree-optimization/84841
1910 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1912 (FLOAT_ONE_CONST_TYPE): Define.
1913 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1914 (sort_by_operand_rank): Put entries with higher constant_type last
1915 rather than first to match comments.
1917 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1919 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1922 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1925 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1926 check and use address of target always.
1928 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1931 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1932 (indirect_thunk_bnd_needed): Likewise.
1933 (indirect_thunks_used): Likewise.
1934 (indirect_thunks_bnd_used): Likewise.
1935 (indirect_return_needed): New.
1936 (indirect_return_bnd_needed): Likewise.
1937 (output_indirect_thunk_function): Add a bool argument for
1939 (output_indirect_thunk_function): Don't generate alias for
1940 function return thunk.
1941 (ix86_code_end): Call output_indirect_thunk_function to generate
1942 function return thunks.
1943 (ix86_output_function_return): Set indirect_return_bnd_needed
1944 and indirect_return_needed instead of indirect_thunk_bnd_needed
1945 and indirect_thunk_needed.
1947 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1949 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1950 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1951 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1953 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1954 Paul Hua <paul.hua.gm@gmail.com>
1957 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1959 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1961 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1962 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1965 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1968 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1969 instead of GET_MODE_SIZE when comparing Units.
1971 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1974 * varasm.c (hash_section): Return an unchangeble hash value
1975 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1976 Return !aarch64_can_use_per_function_literal_pools_p ().
1978 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1981 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1982 rather than address of cmode as last argument to prepare_cmp_insn.
1984 2018-03-15 Julia Koval <julia.koval@intel.com>
1986 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1987 F_AVX512VNNI, F_AVX512BITALG): New.
1989 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1992 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1993 insn for floating-point loads and stores.
1995 2018-03-14 Carl Love <cel@us.ibm.com>
1997 * config/rs6000/rs6000-c.c: Add macro definitions for
1998 ALTIVEC_BUILTIN_VEC_PERMXOR.
1999 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
2000 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
2001 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
2002 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
2004 * config/doc/extend.texi: Add prototypes for vec_permxor.
2006 2018-03-14 David Malcolm <dmalcolm@redhat.com>
2009 * diagnostic-show-locus.c (class layout_point): Convert m_line
2010 from int to linenum_type.
2011 (line_span::comparator): Use linenum "compare" function when
2012 comparing line numbers.
2013 (test_line_span): New function.
2014 (layout_range::contains_point): Convert param "row" from int to
2016 (layout_range::intersects_line_p): Likewise.
2017 (layout::will_show_line_p): Likewise.
2018 (layout::print_source_line): Likewise.
2019 (layout::should_print_annotation_line_p): Likewise.
2020 (layout::print_annotation_line): Likewise.
2021 (layout::print_leading_fixits): Likewise.
2022 (layout::annotation_line_showed_range_p): Likewise.
2023 (struct line_corrections): Likewise for field m_row.
2024 (line_corrections::line_corrections): Likewise for param "row".
2025 (layout::print_trailing_fixits): Likewise.
2026 (layout::get_state_at_point): Likewise.
2027 (layout::get_x_bound_for_row): Likewise.
2028 (layout::print_line): Likewise.
2029 (diagnostic_show_locus): Likewise for locals "last_line" and
2031 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
2032 * input.c (selftest::test_linenum_comparisons): New function.
2033 (selftest::input_c_tests): Call it.
2034 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
2035 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
2036 * selftest.h (ASSERT_GT): New macro.
2037 (ASSERT_GT_AT): New macro.
2038 (ASSERT_LT): New macro.
2039 (ASSERT_LT_AT): New macro.
2041 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
2043 PR rtl-optimization/84780
2044 * combine.c (distribute_links): Don't make a link based on pc_rtx.
2046 2018-03-14 Martin Liska <mliska@suse.cz>
2048 * tree.c (record_node_allocation_statistics): Use
2049 get_stats_node_kind.
2050 (get_stats_node_kind): New function extracted from
2051 record_node_allocation_statistics.
2052 (free_node): Use get_stats_node_kind.
2054 2018-03-14 Richard Biener <rguenther@suse.de>
2056 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
2057 that the value-set of ANTIC_IN doesn't grow.
2060 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2062 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2063 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2065 2018-03-14 Julia Koval <julia.koval@intel.com>
2067 * config.gcc (icelake-client, icelake-server): New.
2069 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
2070 (initial_ix86_arch_features): Ditto.
2071 (PTA_SKYLAKE): Add SGX.
2072 (PTA_ICELAKE): Remove.
2073 (PTA_ICELAKE_CLIENT): New.
2074 (PTA_ICELAKE_SERVER): New.
2075 (ix86_option_override_internal): Split up icelake on icelake client and
2077 (get_builtin_code_for_version): Ditto.
2078 (fold_builtin_cpu): Ditto.
2079 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
2080 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
2081 * config/i386/i386.h (processor_type): Ditto.
2082 * doc/invoke.texi: Ditto.
2084 2018-03-14 Jakub Jelinek <jakub@redhat.com>
2087 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
2088 INTEGER_CST offset, add it together with bitpos / 8 and
2089 sign extend based on POINTER_SIZE.
2093 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
2096 * config/i386/constraints.md (Yc): New register constraint.
2097 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
2098 Use Yc constraint for alternative 2 of operand 0. Remove
2099 preferred_for_speed attribute.
2101 2018-03-14 Richard Biener <rguenther@suse.de>
2103 PR tree-optimization/84830
2104 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
2105 with the old one to avoid oscillations.
2107 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
2110 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
2112 (assign_by_spills): Return a flag of reload assignment failure.
2113 Do not process the reload assignment failures. Do not spill other
2114 reload pseudos if they has the same reg class. Update n if
2116 (lra_assign): Add a return arg. Set up from the result of
2117 assign_by_spills call.
2118 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2119 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2120 usage_insns if it is not NULL.
2121 (spill_hard_reg_in_range): New function.
2122 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2123 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2124 function prototypes.
2125 (lra_assign): Change prototype.
2126 * lra.c (lra): Add code to deal with fails by splitting hard reg
2129 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2131 * config/riscv/riscv.opt (mrelax): New option.
2132 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2133 "norelax" when riscv_mrelax is disabled.
2134 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2136 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2139 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2140 reassociation for int modes.
2142 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2144 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2145 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2147 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2148 * config/aarch64/aarch64-sve.md
2149 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2150 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2151 (*extend<mode><Vwide>2): Rename to...
2152 (aarch64_sve_extend<mode><Vwide>2): ...this.
2153 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2154 renaming the old pattern to...
2155 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2157 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2158 define_expand, renaming the old pattern to...
2159 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2160 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2161 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2162 account when deciding which SVE instruction the optab should use.
2163 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2165 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2167 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2168 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2169 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2170 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2171 (tlsdesc_small_<mode>): Turn a define_expand and use
2172 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2173 (tlsdesc_small_advsimd_<mode>): ...this.
2174 (tlsdesc_small_sve_<mode>): New pattern.
2176 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2178 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2179 (UNSPEC_UMUL_HIGHPART): New constants.
2180 (MUL_HIGHPART): New int iteraor.
2181 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2182 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2184 (*<su>mul<mode>3_highpart): New define_insn.
2186 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2189 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2192 2018-03-13 Martin Liska <mliska@suse.cz>
2195 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2197 (sem_item_optimizer::~sem_item_optimizer): Release it.
2198 (sem_item_optimizer::merge_classes): Register variable aliases.
2199 (sem_item_optimizer::fixup_pt_set): New function.
2200 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2201 * ipa-icf.h: Declare new variables and functions.
2203 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2206 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2207 integer_pow2p@2 and test integer_pow2p in condition.
2208 (A < 0 ? C : 0): Similarly for @1.
2211 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2212 characters starting at p contain '\0' character, don't look beyond
2216 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2217 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2220 * reg-stack.c (change_stack): Change update_end var from int to
2221 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2222 also call set_block_for_insn on the newly added insns and rescan.
2225 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2226 on the last operand.
2229 * tree.c (stabilize_reference_1): Return save_expr (e) for
2230 STATEMENT_LIST even if it doesn't have side-effects.
2232 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2234 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2236 2018-03-12 Renlin Li <renlin.li@arm.com>
2238 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2239 aarch64_output_scalar_simd_mov_immediate.
2241 2018-03-12 Martin Sebor <msebor@redhat.com>
2243 PR tree-optimization/83456
2244 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2245 for perfectly overlapping calls to memcpy.
2246 (gimple_fold_builtin_memory_chk): Same.
2247 (gimple_fold_builtin_strcpy): Handle no-warning.
2248 (gimple_fold_builtin_stxcpy_chk): Same.
2249 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2251 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2253 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2254 parameter. Use it for SFmode.
2255 (rs6000_function_arg_advance_1): Adjust.
2256 (rs6000_function_arg): Adjust.
2257 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2259 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2261 PR rtl-optimization/84169
2262 PR rtl-optimization/84780
2263 * combine.c (can_combine_p): Check for a 2-insn combination whether
2264 the destination register is used between the two insns, too.
2266 2018-03-12 Richard Biener <rguenther@suse.de>
2268 PR tree-optimization/84803
2269 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2270 for refs DR analysis didn't process.
2272 2018-03-12 Richard Biener <rguenther@suse.de>
2274 PR tree-optimization/84777
2275 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2276 force-vectorize loops ignore whether we are optimizing for size.
2278 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2280 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2281 (TARGET_MD_ASM_ADJUST): Define.
2283 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2284 Kito Cheng <kito.cheng@gmail.com>
2285 Chung-Ju Wu <jasonwucj@gmail.com>
2287 * config/nds32/nds32.c (nds32_compute_stack_frame,
2288 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2289 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2290 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2291 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2292 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2293 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2294 * config/nds32/nds32.md (prologue, epilogue): Use macro
2295 NDS32_V3PUSH_AVAILABLE_P to do checking.
2297 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2300 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2301 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2302 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2303 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2304 addition of most attributes on !orig_type_die or the attribute not
2305 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2307 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2308 Chung-Ju Wu <jasonwucj@gmail.com>
2310 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2312 * config/nds32/nds32.opt (mvh): New option.
2314 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2315 Chung-Ju Wu <jasonwucj@gmail.com>
2317 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2319 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2320 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2323 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2324 Chung-Ju Wu <jasonwucj@gmail.com>
2326 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2328 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2329 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2331 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2332 Kito Cheng <kito.cheng@gmail.com>
2333 Chung-Ju Wu <jasonwucj@gmail.com>
2335 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2336 UNSPEC_FFMISM and UNSPEC_FLMISM.
2337 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2338 for ffb, ffmism and flmism.
2339 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2340 (unspec_ffmism): Ditto.
2341 (unspec_flmism): Ditto.
2342 (nds32_expand_builtin_impl): Check if string extension is available.
2343 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2344 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2346 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2349 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2352 * lra-assigns.c (assign_by_spills): Return a flag of reload
2353 assignment failure. Do not process the reload assignment
2354 failures. Do not spill other reload pseudos if they has the same
2356 (lra_assign): Add a return arg. Set up from the result of
2357 assign_by_spills call.
2358 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2359 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2360 usage_insns if it is not NULL.
2361 (spill_hard_reg_in_range): New function.
2362 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2363 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2364 function prototypes.
2365 (lra_assign): Change prototype.
2366 * lra.c (lra): Add code to deal with fails by splitting hard reg
2369 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2372 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2374 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2377 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2378 (dw_val_node): Add val_symbolic_view.
2379 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2380 (symview_upper_bound): New.
2381 (new_line_info_table): Initialize symviews_since_reset.
2382 (dwarf2out_source_line): Count symviews_since_reset and set
2383 symview_upper_bound.
2384 (dw_val_equal_p): Handle symview.
2385 (add_AT_symview): New.
2386 (print_dw_val): Handle symview.
2387 (attr_checksum, attr_checksum_ordered): Likewise.
2388 (same_dw_val_p, size_of_die): Likewise.
2389 (value_format, output_die): Likewise.
2390 (add_high_low_attributes): Use add_AT_symview for entry_view.
2391 (dwarf2out_finish): Reset symview_upper_bound, clear
2394 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2397 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2398 Add strict argument and use it.
2399 (rs6000_split_multireg_move): Update for new strict argument.
2400 (mem_operand_gpr): Disallow all non-offsettable addresses.
2401 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2403 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2406 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2407 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2408 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2411 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2412 decl, use remap_type if we want to use the type.
2414 2018-03-09 Martin Sebor <msebor@redhat.com>
2416 PR tree-optimization/84526
2417 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2419 (builtin_access::generic_overlap): Be prepared to handle non-array
2422 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2424 PR rtl-optimization/84682
2425 * lra-constraints.c (process_address_1): Check is_address flag
2426 for address constraints.
2427 (process_alt_operands): Likewise.
2428 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2429 preprocess_constraints.
2430 * recog.h (preprocess_constraints): Add oploc parameter.
2432 * recog.c (preprocess_constraints): Test address_operand for
2433 CT_ADDRESS constraints.
2435 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2438 * lra-assigns.c (assign_by_spills): Return a flag of reload
2439 assignment failure. Do not process the reload assignment
2440 failures. Do not spill other reload pseudos if they has the same
2442 (lra_assign): Add a return arg. Set up from the result of
2443 assign_by_spills call.
2444 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2445 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2446 usage_insns if it is not NULL.
2447 (spill_hard_reg_in_range): New function.
2448 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2449 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2450 function prototypes.
2451 (lra_assign): Change prototype.
2452 * lra.c (lra): Add code to deal with fails by splitting hard reg
2455 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2458 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2459 Accept complain bool parameter. Only emit errors if it is true.
2460 (arm_parse_cpu_option_name): Likewise.
2461 (arm_target_thumb_only): Adjust callers of the above.
2462 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2463 prototype to take a default true bool parameter.
2464 (arm_parse_arch_option_name): Likewise.
2466 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2467 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2471 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2472 * configure: Regenerate.
2473 * configure.ac ("linker --version-script option"): New.
2474 ("linker soname option"): New.
2476 2018-03-09 Richard Biener <rguenther@suse.de>
2478 PR tree-optimization/84775
2479 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2480 immediate uses of predicate stmts and mark them modified.
2483 PR tree-optimization/84178
2484 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2486 (version_loop_for_if_conversion): Delay update_ssa call.
2487 (tree_if_conversion): Delay update_ssa until after predicate
2490 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2493 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2494 when the function accesses prior frames.
2496 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2499 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2500 gen_llsym, otherwise call maybe_gen_llsym.
2503 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2504 has ',' character inside of it.
2506 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2509 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2510 as clobbering CC_REGNUM.
2512 2018-03-08 Richard Biener <rguenther@suse.de>
2515 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2516 (follow_copies_to_constant): Do not follow SSA names registered
2519 2018-03-08 Richard Biener <rguenther@suse.de>
2521 PR tree-optimization/84178
2522 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2524 (version_loop_for_if_conversion): Delay update_ssa call.
2525 (tree_if_conversion): Delay update_ssa until after predicate
2528 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2530 PR tree-optimization/84178
2531 * tree-if-conv.c (release_bb_predicate): Remove the
2532 the assertion that the stmts have NULL use_ops.
2533 Discard the statements, asserting that they haven't
2534 yet been added to a BB.
2536 2018-03-08 Richard Biener <rguenther@suse.de>
2538 PR tree-optimization/84746
2539 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2540 (phi_translate): Pass in destination ANTIC_OUT set.
2541 (phi_translate_1): Likewise. For a simplified result lookup
2542 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2543 (phi_translate_set): Adjust.
2544 (do_pre_regular_insertion): Likewise.
2545 (do_pre_partial_partial_insertion): Likewise.
2547 2018-03-08 Martin Liska <mliska@suse.cz>
2549 PR gcov-profile/84735
2550 * doc/gcov.texi: Document usage of profile files.
2551 * gcov-io.h: Document changes in the format.
2553 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2557 * dwarf2out.c (struct dw_line_info_table): Update comments for
2559 (FORCE_RESET_NEXT_VIEW): New.
2560 (FORCE_RESETTING_VIEW_P): New.
2561 (RESETTING_VIEW_P): Check for -1 too.
2562 (ZERO_VIEW_P): Likewise.
2563 (new_line_info_table): Force-reset next view.
2564 (dwarf2out_begin_function): Likewise.
2565 (dwarf2out_source_line): Simplify zero_view_p initialization.
2566 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2567 view directly. Omit view when omitting .loc at line 0.
2569 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2571 PR tree-optimization/84740
2572 * tree-switch-conversion.c (process_switch): Call build_constructors
2573 only if info.phi_count is non-zero.
2575 PR tree-optimization/84739
2576 * tree-tailcall.c (find_tail_calls): Check call arguments against
2577 DECL_ARGUMENTS (current_function_decl) rather than
2578 DECL_ARGUMENTS (func) when checking for tail recursion.
2580 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2582 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2583 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2584 Volker Reichelt's entry and add entries for people that perform
2585 GCC fuzzy testing and report numerous bugs.
2587 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2590 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2591 readonly data in sdata, if that is disabled.
2592 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2593 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2594 -mreadonly-in-sdata option.
2596 2018-03-07 Martin Sebor <msebor@redhat.com>
2598 PR tree-optimization/84468
2599 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2600 basic block when looking for nul assignment.
2602 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2605 * except.h (output_function_exception_table): Adjust prototype.
2606 * except.c (output_function_exception_table): Remove FNNAME parameter
2607 and add SECTION parameter. Ouput one part of the table at a time.
2608 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2609 the first part of the exception table and emit unwind directives.
2610 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2611 (i386_pe_seh_cold_init): Likewise.
2612 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2613 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2614 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2615 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2616 * config/i386/winnt.c: Include except.h.
2617 (struct seh_frame_state): Add reg_offset, after_prologue and
2618 in_cold_section fields.
2619 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2620 (i386_pe_seh_cold_init): New function.
2621 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2622 to seh->in_cold_section.
2623 (seh_emit_push): Record the offset of the push.
2624 (seh_emit_save): Record the offet of the save.
2625 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2626 Test seh->after_prologue to disregard the epilogue.
2627 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2628 (i386_pe_end_cold_function): New function.
2630 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2633 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2634 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2637 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2639 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2640 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2643 * multiple_target.c: Include tree-inline.h and intl.h.
2644 (expand_target_clones): Diagnose and fail if node->definition and
2645 !tree_versionable_function_p (node->decl).
2647 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2649 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2651 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2652 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2653 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2655 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2658 * combine.c (try_combine): Use reg_or_subregno instead of handling
2659 just paradoxical SUBREGs and REGs.
2661 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2663 * config/arc/arc.c (arc_finalize_pic): Remove function.
2664 (arc_must_save_register): We use single base PIC register, remove
2665 checks to save/restore the PIC register.
2666 (arc_expand_prologue): Likewise.
2667 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2669 (arc_verify_short): Likewise.
2670 (arc_attr_type): Likewise.
2671 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2672 (walk_stores): Likewise.
2673 (arc_address_cost): Make it static.
2674 (arc_verify_short): Likewise.
2675 (branch_dest): Likewise.
2676 (arc_attr_type): Likewise.
2677 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2678 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2679 (arc_final_prescan_insn): Remove inserting the nops due to
2680 hardware hazards. It is done in reorg step.
2681 (insn_length_variant_t): Remove.
2682 (insn_length_parameters_t): Likewise.
2683 (arc_insn_length_parameters): Likewise.
2684 (arc_get_insn_variants): Likewise.
2685 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2687 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2690 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2693 PR tree-optimization/84687
2694 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2696 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2698 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2700 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2701 Rename to ppc_speculation_barrier.
2702 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2703 __builtin_ppc_speculation_barrier.
2705 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2708 * combine.c (combine_simplify_rtx): Don't try to simplify if
2709 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2712 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2714 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2715 to 32 bytes when compiling for POWER9.
2717 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2720 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2721 regparm >= 3 with no arg reg available also for calls with
2722 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2725 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2727 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2729 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2732 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2734 2018-03-05 Richard Biener <rguenther@suse.de>
2736 PR tree-optimization/84486
2737 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2738 When inserting a __builtin_assume_aligned call set the LHS
2739 SSA name alignment info accordingly.
2741 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2743 PR tree-optimization/84114
2744 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2745 Avoid reassociation of FLOAT_MODE addition.
2747 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2749 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2750 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2751 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2752 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2753 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2754 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2755 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2757 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2758 __builtin_ia32_wbinvd): New builtins.
2759 (SPECIAL_ARGS2): New.
2760 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2761 (SPECIAL_ARGS2): New.
2762 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2763 (ix86_valid_target_attribute_inner_p): Ditto.
2764 (ix86_init_mmx_sse_builtins): Add special_args2.
2765 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2766 TARGET_WBNOINVD_P): New.
2767 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2768 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2769 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2770 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2771 * config/i386/pconfigintrin.h: New file.
2772 * config/i386/wbnoinvdintrin.h: Ditto.
2773 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2774 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2776 2018-03-05 Richard Biener <rguenther@suse.de>
2778 PR tree-optimization/84670
2779 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2781 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2782 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2783 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2784 doesn't grow if all successors (recursively) were visited at least
2787 2018-03-05 Richard Biener <rguenther@suse.de>
2789 PR tree-optimization/84650
2790 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2791 if executed in the loop pipeline.
2793 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2795 * doc/configfiles.texi (Configuration Files): Move info about
2796 conditionalizing $target-protos.h to...
2797 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2798 differs from $target-protos.h.
2800 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2801 Chung-Ju Wu <jasonwucj@gmail.com>
2803 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2804 * config/nds32/nds32-multiple.md (setmemsi): Define.
2805 * config/nds32/nds32-memory-manipulation.c
2806 (nds32_gen_dup_4_byte_to_word_value): New.
2807 (emit_setmem_word_loop): New.
2808 (emit_setmem_byte_loop): New.
2809 (nds32_expand_setmem_loop): New.
2810 (nds32_expand_setmem_loop_v3m): New.
2811 (nds32_expand_setmem_unroll): New.
2812 (nds32_expand_setmem): New.
2814 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2815 Chung-Ju Wu <jasonwucj@gmail.com>
2817 * config/nds32/nds32-memory-manipulation.c
2818 (nds32_emit_load_store): New.
2819 (nds32_emit_post_inc_load_store): New.
2820 (nds32_emit_mem_move): New.
2821 (nds32_emit_mem_move_block): New.
2822 (nds32_expand_movmemsi_loop_unknown_size): New.
2823 (nds32_expand_movmemsi_loop_known_size): New.
2824 (nds32_expand_movmemsi_loop): New.
2825 (nds32_expand_movmemsi_unroll): New.
2826 (nds32_expand_movmemqi): Rename ...
2827 (nds32_expand_movmemsi): ... to this.
2828 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2829 (movmemsi): ... to this.
2830 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2831 (nds32_expand_movmemsi): ... to this.
2833 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2834 Monk Chiang <sh.chiang04@gmail.com>
2835 Chung-Ju Wu <jasonwucj@gmail.com>
2837 * config/nds32/nds32-protos.h
2838 (nds32_expand_load_multiple): New arguments.
2839 (nds32_expand_store_multiple): Ditto.
2840 (nds32_valid_multiple_load_store): Rename ...
2841 (nds32_valid_multiple_load_store_p): ... to this.
2842 * config/nds32/nds32-memory-manipulation.c
2843 (nds32_expand_load_multiple): Refine implementation.
2844 (nds32_expand_store_multiple): Ditto.
2845 * config/nds32/nds32-multiple.md
2846 (load_multiple): Update nds32_expand_load_multiple interface.
2847 (store_multiple): Update nds32_expand_store_multiple interface.
2848 * config/nds32/nds32-predicates.c
2849 (nds32_valid_multiple_load_store): Rename ...
2850 (nds32_valid_multiple_load_store_p): ... to this and refine
2852 * config/nds32/predicates.md
2853 (nds32_load_multiple_and_update_address_operation): New predicate.
2854 (nds32_store_multiple_and_update_address_operation): New predicate.
2856 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2857 Chung-Ju Wu <jasonwucj@gmail.com>
2859 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2860 (combo): New attribute.
2861 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2863 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2865 * config/nds32/nds32.opt: Change -mcmodel= default value.
2867 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2868 Monk Chiang <sh.chiang04@gmail.com>
2869 Chung-Ju Wu <jasonwucj@gmail.com>
2871 * config/nds32/constants.md (unspec_element): New enum.
2872 * config/nds32/constraints.md (Umw): New constraint.
2873 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2874 * config/nds32/nds32-intrinsic.md: Likewise.
2875 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2876 (nds32_valid_smw_lwm_base_p): New.
2877 (nds32_output_smw_single_word): New.
2878 (nds32_output_lmw_single_word): New.
2879 (nds32_expand_unaligned_load): New.
2880 (nds32_expand_unaligned_store): New.
2881 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2882 (nds32_output_smw_single_word): Declare.
2883 (nds32_output_lmw_single_word): Declare.
2884 (nds32_expand_unaligned_load): Declare.
2885 (nds32_expand_unaligned_store): Declare.
2886 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2887 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2888 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2889 NDS32_BUILTIN_UASTORE_DW.
2890 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2893 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2894 Kito Cheng <kito.cheng@gmail.com>
2895 Chung-Ju Wu <jasonwucj@gmail.com>
2897 * config/nds32/nds32-intrinsic.c
2898 (nds32_expand_builtin_null_ftype_reg): Delete.
2899 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2900 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2901 (nds32_read_argument): New.
2902 (nds32_legitimize_target): Ditto.
2903 (nds32_legitimize_argument): Ditto.
2904 (nds32_check_constant_argument): Ditto.
2905 (nds32_expand_unop_builtin): Ditto.
2906 (nds32_expand_unopimm_builtin): Ditto.
2907 (nds32_expand_binop_builtin): Ditto.
2908 (nds32_builtin_decl_impl): Ditto.
2909 (builtin_description): Ditto.
2910 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2911 (nds32_init_builtins_impl): Ditto.
2912 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2913 (nds32_builtin_decl): New.
2914 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2915 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2917 2018-03-02 Jeff Law <law@redhat.com>
2919 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2920 (redundant_insn, fill_simple_delay_slots): Likewise.
2921 (fill_slots_from_thread): Likewise.
2922 * resource.c (mark_referenced_resources): Likewise.
2923 (mark_set_resources, find_dead_or_set_registers): Likewise.
2925 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2927 * substring-locations.h (format_warning_va): Formatting fix for
2929 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2931 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2932 * substring-locations.c: Include intl.h.
2933 (format_warning_va): Turned into small wrapper around
2934 format_warning_n_va, renamed to ...
2935 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2936 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2938 (format_warning_at_substring_n): New function.
2939 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2940 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2941 format_warning_at_substring with just a shorter name instead of
2942 const function pointer.
2943 (fmtwarn_n): New function.
2944 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2945 appropriate, get rid of all the fmtstr temporaries, move conditionals
2946 with G_() wrapped string literals directly into fmtwarn arguments,
2947 cast dir.len to (int), formatting fixes.
2949 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2951 * doc/invoke.texi: Remove "Cilk Plus" references.
2953 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2954 Richard Biener <rguenther@suse.de>
2957 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2958 for error or warning attributes if CALL_FROM_THUNK_P is set.
2961 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2964 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2965 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2968 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2969 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2971 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2972 instead of -1U in last predictors element's probability member.
2974 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2977 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2978 arguments if they are comparable.
2980 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2982 PR tree-optimization/84634
2983 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2984 masks and masked_loop_p with a single loop_masks, making sure it's
2985 null for bb vectorization.
2987 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2989 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2990 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2991 loop->force_vectorize to check whether there is no alias.
2993 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2996 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2998 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
3000 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
3001 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
3002 instead of a loop around prev_real_insn.
3003 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
3007 * config/i386/i386.c (ix86_print_operand): Use conditional
3008 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
3011 2018-03-02 Richard Biener <rguenther@suse.de>
3013 PR tree-optimization/84427
3014 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
3015 (bitmap_set_subtract_values): Rewrite to handle multiple
3018 (prune_clobbered_mems): Likewise.
3019 (phi_translate): Take edge instead of pred/phiblock.
3020 (phi_translate_1): Likewise.
3021 (phi_translate_set): Likewise. Insert all translated
3022 exprs for a value into the set, keeping possibly multiple
3023 expressions per value.
3024 (compute_antic_aux): Adjust for phi_translate changes.
3025 When intersecting union the expressions and prune those
3026 not in the final value set, keeping possibly multiple
3027 expressions per value. Do not use value-insertion
3028 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
3029 all expressions. Add verification that the value-sets
3030 only shrink during iteration.
3031 (compute_partial_antic_aux): Adjust for the phi_translate changes.
3032 (do_pre_regular_insertion): Likewise.
3033 (do_pre_partial_partial_insertion): Likewise.
3035 2018-03-02 Richard Biener <rguenther@suse.de>
3038 * config/darwin.c (saved_debug_info_level): New static global.
3039 (darwin_asm_lto_start): Disable debug info generation for LTO out.
3040 (darwin_asm_lto_end): Restore debug info generation settings.
3042 2018-03-01 Martin Liska <mliska@suse.cz>
3045 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
3048 2018-03-01 Richard Biener <rguenther@suse.de>
3051 * dwarf2out.c (gen_variable_die): Properly handle late VLA
3052 type annotation with LTO when debug was disabled at compile-time.
3054 2018-03-01 Matthew Fortune <mfortune@gmail.com>
3056 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
3058 (mips_final_postscan_insn): Likewise.
3060 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
3062 PR rtl-optimization/84528
3063 * alias.c (init_alias_target): Add commentary.
3064 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
3065 a unique base value if the frame pointer is not eliminated
3066 to the stack pointer.
3068 2018-03-01 Tom de Vries <tom@codesourcery.com>
3070 PR rtl-optimization/83327
3071 * lra-int.h (hard_regs_spilled_into): Declare.
3072 * lra.c (hard_regs_spilled_into): Define.
3073 (init_reg_info): Init hard_regs_spilled_into.
3074 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
3075 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
3076 (process_bb_lives): Handle hard_regs_spilled_into.
3077 (lra_create_live_ranges_1): Before doing liveness propagation, clear
3078 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
3080 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
3082 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
3083 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
3084 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
3085 * config/rs6000/aix72.h: New file.
3087 2018-02-28 Jakub Jelinek <jakub@redhat.com>
3089 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
3090 instead of warning_at with conditional singular and plural messages
3094 * stor-layout.c (update_alignment_for_field): For
3095 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
3096 && !DECL_PACKED (field), do the alignment update, just use
3097 only desired_align instead of MAX (type_align, desired_align)
3099 (place_field): Don't do known_align < desired_align handling
3100 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
3101 is non-NULL, instead do it after rli->prev_field handling and
3102 only if not within a bitfield word. For DECL_PACKED (field)
3103 use type_align of BITS_PER_UNIT.
3105 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
3107 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
3108 superfluous parentheses and trailing spaces.
3110 2018-02-28 Richard Biener <rguenther@suse.de>
3112 PR tree-optimization/84584
3113 * graphite-scop-detection.c (scop_detection::add_scop): Discard
3114 SCoPs with fake exit edge.
3116 2018-02-28 Martin Liska <mliska@suse.cz>
3119 * timevar.c (timer::print): Fix format to properly print 100%
3122 2018-02-28 Richard Biener <rguenther@suse.de>
3125 * genmatch.c (capture_info::walk_match): Do not mark
3126 captured expressions without operands as expr_p given
3127 they act more like predicates and should be subject to
3128 "lost tail" side-effect preserving.
3130 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3132 PR rtl-optimization/81611
3133 * auto-inc-dec.c (attempt_change): Move dead note from
3134 mem_insn if it's the next use of regno
3135 (find_address): Take address use of reg holding
3136 non-incremented value. Add parm to limit search to the named
3138 (merge_in_block): Attempt to use a mem insn that is the next
3139 use of the original regno.
3141 2018-02-27 Martin Sebor <msebor@redhat.com>
3144 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3145 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3147 2018-02-27 Martin Sebor <msebor@redhat.com>
3149 PR translation/84207
3150 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3151 n argument to unsigned HOST_WIDE_INT.
3152 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3153 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3154 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3155 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3157 2018-02-27 Richard Biener <rguenther@suse.de>
3159 PR tree-optimization/84512
3160 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3161 Do not use the estimate returned from record_stmt_cost for
3162 the scalar iteration cost but sum properly using add_stmt_cost.
3164 2018-02-27 Richard Biener <rguenther@suse.de>
3166 PR tree-optimization/84466
3167 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3168 Adjust last change to less strictly validate use operands.
3170 2018-02-27 Martin Liska <mliska@suse.cz>
3172 PR gcov-profile/84548
3173 * gcov.c (process_file): Allow partial overlap and consider it
3174 also as group functions.
3175 (output_lines): Properly calculate range of lines for a group.
3177 2018-02-27 Martin Liska <mliska@suse.cz>
3179 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3180 'ggc' suffixes. Change first column width.
3181 (timer::print): Fix formatting of the column.
3183 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3185 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3186 preserve inline entry blocks for the sake of debug inline
3187 entry point markers alone.
3188 (remove_unused_locals): Suggest in comments a better place to
3189 force the preservation of inline entry blocks that are
3190 otherwise unused, but do not preserve them.
3192 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3194 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3196 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3199 * config/i386/constraints.md (Bs): Replace
3200 ix86_indirect_branch_register with
3201 TARGET_INDIRECT_BRANCH_REGISTER.
3203 * config/i386/i386.md (indirect_jump): Likewise.
3204 (tablejump): Likewise.
3205 (*sibcall_memory): Likewise.
3206 (*sibcall_value_memory): Likewise.
3207 Peepholes of indirect call and jump via memory: Likewise.
3208 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3209 (*sibcall_value_GOT_32): Likewise.
3210 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3211 (GOT_memory_operand): Likewise.
3212 (call_insn_operand): Likewise.
3213 (sibcall_insn_operand): Likewise.
3214 (GOT32_symbol_operand): Likewise.
3215 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3217 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3219 PR rtl-optimization/83496
3220 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3221 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3222 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3223 redundant insn, if any.
3224 (relax_delay_slots): Likewise.
3225 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3227 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3229 PR tree-optimization/83965
3230 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3231 that grouped statements are part of a reduction chain. Return
3232 true if the statement is not marked as a reduction itself but
3234 (vect_recog_dot_prod_pattern): Don't check whether the statement
3235 is part of a group here.
3236 (vect_recog_sad_pattern): Likewise.
3237 (vect_recog_widen_sum_pattern): Likewise.
3239 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3242 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3244 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3247 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3249 (ix86_output_indirect_function_return): New prototype.
3250 (ix86_split_simple_return_pop_internal): Likewise.
3251 * config/i386/i386.c (indirect_return_via_cx): New.
3252 (indirect_return_via_cx_bnd): Likewise.
3253 (indirect_thunk_name): Handle return va CX_REG.
3254 (output_indirect_thunk_function): Create alias for
3255 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3256 (ix86_output_indirect_jmp): Remove the bool argument.
3257 (ix86_output_indirect_function_return): New function.
3258 (ix86_split_simple_return_pop_internal): Likewise.
3259 * config/i386/i386.md (*indirect_jump): Don't pass false
3260 to ix86_output_indirect_jmp.
3261 (*tablejump_1): Likewise.
3262 (simple_return_pop_internal): Change it to define_insn_and_split.
3263 Call ix86_split_simple_return_pop_internal to split it for
3265 (simple_return_indirect_internal): Call
3266 ix86_output_indirect_function_return instead of
3267 ix86_output_indirect_jmp.
3269 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3272 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3273 memset and value initialization afterwards.
3275 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3277 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3279 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3282 * common/config/aarch64/aarch64-common.c
3283 (aarch_option_optimization_table[]): Switch
3284 off fomit-frame-pointer
3286 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3287 Chung-Ju Wu <jasonwucj@gmail.com>
3289 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3291 (store_multiple): Ditto.
3293 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3295 * config.gcc: Add --with-cpu support for nds32 target.
3296 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3297 * config/nds32/nds32.opt: Add -mcpu= option.
3299 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3301 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3302 isel=yes): Warn for these deprecated options.
3304 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3306 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3307 ISA_2_5_MASKS_EMBEDDED.
3309 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3311 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3312 p->max as pointers rather than using iterative_hash_expr.
3314 2018-02-23 Carl Love <cel@us.ibm.com>
3316 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3317 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3319 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3320 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3321 P8V_BUILTIN_VEC_VUNSIGNED2.
3323 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3326 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3327 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3329 * lra-constraints.c (curr_insn_transform): Set up
3330 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3332 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3333 * lra-eliminations.c (spill_pseudos): Ditto.
3334 (process_insn_for_elimination): Ditto.
3335 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3336 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3337 LRA_NON_CLOBBERED_ALT.
3339 2018-02-22 Martin Sebor <msebor@redhat.com>
3341 PR tree-optimization/84480
3342 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3343 to maybe_diag_stxncpy_trunc. Call it.
3344 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3345 from gimple_fold_builtin_strcpy. Print inlining stack.
3346 (handle_builtin_stxncpy): Print inlining stack.
3347 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3349 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3352 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3353 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3354 and -fcheck-pointer-bounds are used together.
3355 (indirect_thunk_prefix): New enum.
3356 (indirect_thunk_need_prefix): New function.
3357 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3358 "_nt" instead of "_bnd" for NOTRACK prefix.
3359 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3360 (output_indirect_thunk_function): Likewise.
3362 (ix86_code_end): Update output_indirect_thunk_function calls.
3363 (ix86_output_indirect_branch_via_reg): Replace
3364 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3365 (ix86_output_indirect_branch_via_push): Likewise.
3366 (ix86_output_function_return): Likewise.
3367 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3368 incompatible with -fcf-protection=branch and
3369 -fcheck-pointer-bounds.
3371 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3374 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3375 Change gcc_assert call to output_operand_lossage.
3377 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3379 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3381 2018-02-22 DJ Delorie <dj@redhat.com>
3382 Sebastian Perta <sebastian.perta@renesas.com>
3383 Oleg Endo <olegendo@gcc.gnu.org>
3385 * config/rx/rx.c (rx_rtx_costs): New function.
3386 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3388 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3390 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3392 2018-02-22 Martin Liska <mliska@suse.cz>
3395 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3396 Add "native" as a possible value.
3398 2018-02-22 Martin Liska <mliska@suse.cz>
3401 * config/i386/i386.c (ix86_option_override_internal):
3402 Add "native" as a possible value for -march and -mtune.
3404 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3407 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3408 to all type variants.
3410 PR tree-optimization/84503
3411 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3412 width as info->bitpos + info->bitsize - start.
3413 (merged_store_group::merge_overlapping): Simplify width computation.
3414 (check_no_overlap): New function.
3415 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3416 start + width and last_order of the group, fail if check_no_overlap
3418 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3419 to group if check_no_overlap fails.
3421 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3423 * config/rs6000/altivec.md: Delete contraint arguments to
3424 define_expand, define_split, and define_peephole2, and in
3425 define_insn_and_split if always unused.
3426 * config/rs6000/darwin.md: Ditto.
3427 * config/rs6000/dfp.md: Ditto.
3428 * config/rs6000/rs6000.md: Ditto.
3429 * config/rs6000/sync.md: Ditto.
3430 * config/rs6000/vector.md: Ditto.
3431 * config/rs6000/vsx.md: Ditto.
3433 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3435 * config/rs6000/altivec.md: Write output control strings as braced
3436 blocks instead of double-quoted strings.
3437 * config/rs6000/darwin.md: Ditto.
3438 * config/rs6000/rs6000.md: Ditto.
3439 * config/rs6000/vector.md: Ditto.
3440 * config/rs6000/vsx.md: Ditto.
3442 2018-02-21 Jason Merrill <jason@redhat.com>
3444 PR c++/84314 - ICE with templates and fastcall attribute.
3445 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3447 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3449 * ipa-cp.c (determine_versionability): Fix comment typos.
3451 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3454 * ipa-cp.c (determine_versionability): Do not version functions caling
3457 2018-02-21 Martin Liska <mliska@suse.cz>
3460 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3461 Add "native" as a possible value.
3462 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3463 the macro when native cpu detection is available.
3465 2018-02-21 Martin Liska <mliska@suse.cz>
3468 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3469 Add "native" as a possible value.
3470 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3471 when native cpu detection is available.
3473 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3474 Martin Sebor <msebor@redhat.com>
3476 PR tree-optimization/84478
3477 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3479 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3480 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3481 support which is conservatively correct, for 2 only stay conservative
3482 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3483 argument to the 2 argument get_range_strlen, adjust 6 arg
3484 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3486 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3487 (gimple_fold_builtin_strlen): Pass true as last argument to
3490 2018-02-20 Martin Sebor <msebor@redhat.com>
3493 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3494 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3495 (builtin_memref::builtin_memref): Factor out parts into
3496 set_base_and_offset and call it.
3498 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3501 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3502 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3503 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3504 search at the associated MODE_INT.
3506 2018-02-20 Jeff Law <law@redhat.com>
3509 PR tree-optimization/81592
3511 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3512 for range data rather than using global data.
3513 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3514 range data rather than using global data.
3515 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3516 pass it to children as needed.
3517 (struct directive::fmtresult): Similarly.
3518 (struct directive::set_width): Similarly.
3519 (struct directive::set_precision): Similarly.
3520 (format_integer, format_directive, parse_directive): Similarly.
3521 (format_none): Accept unnamed vr_values parameter.
3522 (format_percent, format_floating, format_character): Similarly.
3523 (format_string, format_plain): Similarly.
3524 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3525 the EVRP range analyzer for range data rather than using global data.
3526 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3527 gimple-ssa-evrp-analyze.h
3528 (class sprintf_dom_walker): Add after_dom_children member function.
3529 Add evrp_range_analyzer member.
3530 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3531 range analyzer as needed.
3532 (sprintf_dom_walker::after_dom_children): New member function.
3533 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3535 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3536 (evrp_range_analyzer::pop_to_marker): Likewise.
3538 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3540 PR tree-optimization/84419
3541 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3542 with the required type if its current type is compatible but
3545 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3548 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3549 after vectorization.
3551 2018-02-20 Martin Liska <mliska@suse.cz>
3554 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3555 possible values if we don't have a hint.
3557 2018-02-20 Martin Liska <mliska@suse.cz>
3561 * final.c (shorten_branches): Build align_tab array with one
3563 * opts.c (finish_options): Add alignment option limit check.
3564 (MAX_CODE_ALIGN): Likewise.
3565 (MAX_CODE_ALIGN_VALUE): Likewise.
3566 * doc/invoke.texi: Document maximum allowed option value for
3567 all -falign-* options.
3569 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3572 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3573 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3574 * var-tracking.c (emit_note_insn_var_location): Remove all references
3575 to NOTE_INSN_CALL_ARG_LOCATION.
3576 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3577 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3578 Use copy_rtx_if_shared.
3579 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3580 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3581 (dwarf2out_var_location): Remove handling of
3582 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3584 * final.c (final_scan_insn): Remove all references to
3585 NOTE_INSN_CALL_ARG_LOCATION.
3586 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3587 before dumping final insns.
3588 * except.c (emit_note_eh_region_end): Remove all references to
3589 NOTE_INSN_CALL_ARG_LOCATION.
3590 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3591 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3592 * config/arc/arc.c (hwloop_optimize): Likewise.
3593 * config/arm/arm.c (create_fix_barrier): Likewise.
3594 * config/s390/s390.c (s390_chunkify_start): Likewise.
3595 * config/sh/sh.c (find_barrier): Likewise.
3596 * config/i386/i386.c (rest_of_insert_endbranch,
3597 ix86_seh_fixup_eh_fallthru): Likewise.
3598 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3599 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3600 * config/frv/frv.c (frv_function_prologue): Likewise.
3601 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3603 (note_outside_basic_block_p): Remove all references to
3604 NOTE_INSN_CALL_ARG_LOCATION.
3605 * gengtype.c (adjust_field_rtx_def): Likewise.
3606 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3608 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3609 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3612 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3615 PR tree-optimization/84452
3616 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3617 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3620 2018-02-19 Martin Liska <mliska@suse.cz>
3623 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3625 2018-02-19 Martin Liska <mliska@suse.cz>
3626 Richard Sandiford <richard.sandiford@linaro.org>
3628 PR tree-optimization/82491
3629 * gimple-fold.c (get_base_constructor): Make earlier bail out
3632 2018-02-19 Carl Love <cel@us.ibm.com>
3634 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3635 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3637 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3638 P8V_BUILTIN_VEC_NEG.
3640 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3642 * config/rl78/rl78.md (movdf): New define expand.
3644 2018-02-19 Martin Liska <mliska@suse.cz>
3647 * doc/invoke.texi: Fix typo.
3648 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3650 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3652 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3653 handle rs6000_single_float and rs6000_double_float specially for
3656 2018-02-16 Jeff Law <law@redhat.com>
3658 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3661 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3664 * ipa-inline.c (inline_small_functions): Fix a typo.
3666 2018-02-16 Nathan Sidwell <nathan@acm.org>
3668 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3670 2018-02-16 Carl Love <cel@us.ibm.com>
3672 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3673 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3674 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3675 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3676 expansion to P8V_BUILTIN_VEC_FLOAT2.
3678 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3680 PR rtl-optimization/70023
3681 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3682 src_regno into account.
3684 2018-02-16 Carl Love <cel@us.ibm.com>
3686 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3687 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3688 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3689 * config/rs6000/rs6000.c: Remove case statements for
3690 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3691 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3692 and P9V_BUILTIN_VEC_VINSERT4B.
3693 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3694 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3695 * config/rs6000/vsx.md:
3696 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3699 2018-02-16 Carl Love <cel@us.ibm.com>
3701 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3703 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3705 * config/rs6000/rs6000-c.c: Add the definitions for
3706 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3707 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3708 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3709 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3710 definition for insert4b and define insn *insert3b_internal.
3711 * doc/extend.texi: Add documentation for vec_extract4b.
3713 2018-02-16 Nathan Sidwell <nathan@acm.org>
3715 * doc/extend.texi (Backwards Compatibility): Mention friend
3716 injection. Note for-scope is deprecated.
3717 * doc/invoke.texi (-ffriend-injection): Deprecate.
3719 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3721 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3722 that moved to I2, also allow destinations that are a paradoxical
3723 subreg (instead of a normal reg).
3725 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3728 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3731 2018-02-16 Richard Biener <rguenther@suse.de>
3733 PR tree-optimization/84037
3734 PR tree-optimization/84016
3736 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3737 Adjust vec_construct for the fact we need additional higher latency
3738 128bit inserts for AVX256 and AVX512 vector builds.
3739 (ix86_add_stmt_cost): Scale vector construction cost for
3742 2018-02-16 Richard Biener <rguenther@suse.de>
3744 PR tree-optimization/84417
3745 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3746 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3747 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3749 2018-02-16 Martin Liska <mliska@suse.cz>
3752 * internal-fn.def (ASAN_CHECK): Set proper flags.
3753 (ASAN_MARK): Likewise.
3755 2018-02-16 Julia Koval <julia.koval@intel.com>
3757 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3758 from PTA_CANNONLAKE.
3760 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3763 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3764 Use ++iter rather than iter++ for std::list iterators.
3765 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3766 defer deleting them until all nodes in the forest are processed. Do
3767 free even leaf nodes. Change to_process into auto_vec.
3770 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3771 * vec.h (vec_default_construct): Use memset instead of placement new
3772 if BROKEN_VALUE_INITIALIZATION is defined.
3773 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3774 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3777 PR rtl-optimization/83723
3778 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3779 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3780 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3782 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3784 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3786 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3788 PR rtl-optimization/81443
3789 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3790 from inner REGs to paradoxical SUBREGs.
3792 2018-02-16 Richard Biener <rguenther@suse.de>
3794 PR tree-optimization/84399
3795 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3796 For operands we can analyze at their definition make sure we can
3797 analyze them at each use as well.
3799 2018-02-16 Richard Biener <rguenther@suse.de>
3801 PR tree-optimization/84190
3802 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3803 volatile accesses if the decl isn't volatile.
3805 2018-02-15 Jason Merrill <jason@redhat.com>
3807 PR c++/84314 - ICE with templates and fastcall attribute.
3808 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3809 TYPE_CANONICAL on an existing type.
3811 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3813 PR tree-optimization/84383
3814 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3815 dstoff nor call operand_equal_p if dstbase is NULL.
3817 PR tree-optimization/84334
3818 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3819 also a CONSTANT_CLASS_P, punt.
3821 2018-02-14 Jim Wilson <jimw@sifive.com>
3823 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3824 first SMALL_OPERAND check. New local min_second_step. Move assert
3825 to where locals are set. Add TARGET_RVC support.
3826 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3828 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3830 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3832 2018-02-14 Martin Sebor <msebor@redhat.com>
3834 PR tree-optimization/83698
3835 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3836 arrays constrain the offset range to their bounds.
3837 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3838 (builtin_access::overlap): Avoid setting the size of overlap if it's
3840 (maybe_diag_overlap): Also consider arrays when deciding what values
3841 of offsets to include in diagnostics.
3843 2018-02-14 Martin Sebor <msebor@redhat.com>
3846 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3847 that correspond to the kind of a declaration.
3849 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3852 * config/pa/pa.md: Load address of PIC label using the linkage table
3853 if the label is nonlocal.
3855 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3857 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3858 warning message if user requests -maltivec=be.
3859 * doc/invoke.texi: Document deprecation of -maltivec=be.
3861 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3864 * config/rs6000/rs6000-c.c: Update definitions for
3865 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3866 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3868 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3871 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3872 add _get_ssp intrinsics. Remove argument from
3873 __builtin_ia32_rdssp[d|q].
3874 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3875 * config/i386/i386-builtin.def: Remove argument from
3876 __builtin_ia32_rdssp[d|q].
3877 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3878 ix86_expand_special_args_builtin for _rdssp[d|q].
3879 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3880 Clear register before usage.
3881 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3882 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3884 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3886 PR tree-optimization/84357
3887 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3888 operand 1 of an ARRAY_REF too.
3890 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3893 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3894 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3896 (set_of_reg): New struct.
3897 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3898 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3899 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3901 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3902 Split into bitclr, bitset, bitinvert patterns if appropriate.
3903 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3904 use rx_fuse_in_memory_bitop.
3905 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3906 to named insn, correct maximum insn length.
3908 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3911 * machmode.def: Define a complex mode for PARTIAL_INT.
3912 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3914 * doc/rtl.texi: Document CSPImode.
3915 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3917 (msp430_hard_regno_nregs_with_padding): Likewise.
3919 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3922 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3924 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3926 PR rtl-optimization/84169
3927 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3928 we generated a parallel as new i3 and we split that to new i2 and i3
3929 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3930 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3931 those to i2, not i1. Partially rewrite this scan code.
3933 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3936 * stor-layout.c (place_field): For variable length fields, adjust
3937 offset_align afterwards not just based on the field's alignment,
3938 but also on the size.
3941 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3942 of exps and logs in the use_exp2 case.
3944 2018-02-13 Jeff Law <law@redhat.com>
3946 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3949 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3952 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3956 * common.opt (gas-loc-support, gas-locview-support): New.
3957 (ginline-points, ginternal-reset-location-views): New.
3958 * doc/invoke.texi: Document them. Use @itemx where intended.
3959 (gvariable-location-views): Adjust.
3960 * target.def (reset_location_view): New.
3961 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3962 (TARGET_RESET_LOCATION_VIEW): New.
3963 * doc/tm.texi: Rebuilt.
3964 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3965 (dwarf2out_default_as_locview_support): New.
3966 (output_asm_line_debug_info): Use option variables.
3967 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3968 (output_loc_list): Likewise.
3969 (add_high_low_attributes): Check option variables.
3970 Don't output entry view attribute in strict mode.
3971 (gen_inlined_subroutine_die): Check option variables.
3972 (dwarf2out_inline_entry): Likewise.
3973 (init_sections_and_labels): Likewise.
3974 (dwarf2out_early_finish): Likewise.
3975 (maybe_reset_location_view): New, from...
3976 (dwarf2out_var_location): ... here. Call it.
3977 * debug.h (dwarf2out_default_as_loc_support): Declare.
3978 (dwarf2out_default_as_locview_support): Declare.
3979 * hooks.c (hook_int_rtx_insn_0): New.
3980 * hooks.h (hook_int_rtx_insn_0): Declare.
3981 * toplev.c (process_options): Take -gas-loc-support and
3982 -gas-locview-support from dwarf2out. Enable
3983 -gvariable-location-views by default only with locview
3984 assembler support. Enable -ginternal-reset-location-views by
3985 default only if the target defines the corresponding hook.
3986 Enable -ginline-points by default if location views are
3987 enabled; force it disabled if statement frontiers are
3989 * tree-inline.c (expand_call_inline): Check option variables.
3990 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3992 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3994 PR tree-optimization/84321
3995 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3996 handling. Also check whether the anti-range contains any values
3997 that satisfy the mask; switch to a VR_RANGE if not.
3999 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
4002 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
4004 2018-02-13 Martin Jambor <mjambor@suse.cz>
4007 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
4008 of call statements, also set location of a load to a temporary.
4010 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
4012 * config/rl78/rl78.c (add_vector_labels): New function.
4013 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
4014 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
4015 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
4016 which checks that no arguments are passed.
4017 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
4018 * doc/extend.texi: Documentation for the new attribute.
4020 2018-02-13 Andreas Schwab <schwab@suse.de>
4022 * config/riscv/linux.h (CPP_SPEC): Define.
4024 2018-02-13 Jakub Jelinek <jakub@redhat.com>
4027 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
4028 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
4029 OPTION_MASK_ISA_AES as first argument to def_builtin_const
4030 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
4031 instead of OPTION_MASK_ISA_PCLMUL as first argument to
4032 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
4033 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
4034 temporarily for AES and PCLMUL builtins.
4036 PR tree-optimization/84339
4037 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
4038 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
4042 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
4043 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
4044 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
4046 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
4048 * omp-simd-clone.h: New file.
4049 * omp-simd-clone.c: Include omp-simd-clone.h.
4050 (expand_simd_clones): No longer static.
4051 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
4052 cgraph.h and omp-simd-clone.h.
4053 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
4054 (vect_recog_widen_shift_pattern): Formatting fix.
4055 (vect_pattern_recog_1): Don't check optab for calls.
4058 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
4059 operands[2] into a REG before using gen_lowpart on it.
4061 2018-02-12 Jeff Law <law@redhat.com>
4064 * config/sh/sh.c (find_barrier): Consider a sibling call
4067 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
4068 successfully back substituting a reg.
4070 2018-02-12 Richard Biener <rguenther@suse.de>
4072 PR tree-optimization/84037
4073 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
4074 parameter, move visited init to caller.
4075 (vect_slp_analyze_operations): Separate cost from validity
4076 check, initialize visited once for all instances.
4077 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
4079 * tree-vect-stmts.c (vect_model_simple_cost): Make early
4081 (vect_model_promotion_demotion_cost): Likewise.
4082 (vectorizable_bswap): Guard cost modeling with !slp_node
4083 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
4085 (vectorizable_call): Likewise.
4086 (vectorizable_conversion): Likewise.
4087 (vectorizable_assignment): Likewise.
4088 (vectorizable_shift): Likewise.
4089 (vectorizable_operation): Likewise.
4090 (vectorizable_store): Likewise.
4091 (vectorizable_load): Likewise.
4092 (vectorizable_condition): Likewise.
4093 (vectorizable_comparison): Likewise.
4095 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
4098 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
4099 (ASAN_MARK): Fix fnspec to account for return value, change pointer
4100 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
4102 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4105 * params.def (inline-min-speedup): Increase from 8 to 15.
4106 (max-inline-insns-auto): Decrease from 40 to 30.
4107 * ipa-split.c (consider_split): Add some buffer for function to
4108 be considered inlining candidate.
4109 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
4112 2018-02-12 Richard Biener <rguenther@suse.de>
4114 PR tree-optimization/84037
4115 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
4116 matched stmts if we cannot swap the non-matched ones.
4118 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4120 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
4121 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4122 _mm_maskz_scalef_round_ss): New intrinsics.
4123 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4124 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4125 __builtin_ia32_scalefss_round): Remove.
4126 (__builtin_ia32_scalefsd_mask_round,
4127 __builtin_ia32_scalefss_mask_round): New intrinsics.
4128 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4129 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4130 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4131 "<round_constraint>")): Changed to ...
4132 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4133 "<round_scalar_constraint>")): ... this.
4134 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4135 %0, %1, %2<round_op3>}"): Changed to ...
4136 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4137 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4138 %2<round_scalar_mask_op3>}"): ... this.
4139 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4141 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4143 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4144 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4145 (_mm_maskz_sqrt_round_ss): New intrinsics.
4146 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4147 (__builtin_ia32_sqrtsd_mask_round)
4148 (__builtin_ia32_sqrtss_mask_round): New builtins.
4149 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4150 (__builtin_ia32_sqrtss_round): Remove.
4151 (__builtin_ia32_sqrtsd_mask_round)
4152 (__builtin_ia32_sqrtss_mask_round): New builtins.
4153 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4154 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4155 ((match_operand:VF_128 1 "vector_operand"
4156 "xBm,<round_constraint>")): Changed to ...
4157 ((match_operand:VF_128 1 "vector_operand"
4158 "xBm,<round_scalar_constraint>")): ... this.
4159 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4160 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4161 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4162 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4163 %<iptr>1<round_scalar_mask_op3>}): ... this.
4164 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4165 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4167 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4170 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4171 Cast vec_cmpeq result to correct type.
4172 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4173 Cast vec_cmpgt result to correct type.
4175 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4177 * final.c (final_scan_insn_1): Renamed from...
4178 (final_scan_insn): ... this. New wrapper, to recover
4179 seen from the outermost call in recursive ones.
4180 * config/sparc/sparc.c (output_return): Drop seen from call.
4181 (output_sibcall): Likewise.
4182 * config/visium/visium.c (output_branch): Likewise.
4184 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4186 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4189 2018-02-10 Alan Modra <amodra@gmail.com>
4192 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4193 Specify LR as an input.
4195 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4198 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4199 remove_member_access_dummy_vars): New functions.
4200 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4201 lower_omp_1, execute_lower_omp): Use them.
4203 PR rtl-optimization/84308
4204 * shrink-wrap.c (spread_components): Release todo vector.
4206 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4208 PR rtl-optimization/57193
4209 * ira-color.c (struct allocno_color_data): Add member
4210 conflict_allocno_hard_prefs.
4211 (update_conflict_allocno_hard_prefs): New.
4212 (bucket_allocno_compare_func): Add a preference based on
4213 conflict_allocno_hard_prefs.
4214 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4215 (color_allocnos): Remove a dead code. Initiate
4216 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4218 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4221 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4222 constraint from =wa to wa. Avoid a subreg on the output operand,
4223 instead use a pseudo and subreg it in a move.
4224 (p9_xxbrd_<mode>): Changed to ...
4225 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4226 (p9_xxbrd_v2df): New expander.
4227 (p9_xxbrw_<mode>): Changed to ...
4228 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4229 (p9_xxbrw_v4sf): New expander.
4231 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4233 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4235 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4238 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4239 multiply in 32-bit mode.
4240 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4241 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4244 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4246 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4247 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4248 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4249 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4251 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4254 * dwarf2out.c (is_trivial_indirect_ref): New function.
4255 (dwarf2out_late_global_decl): Do not generate a location
4256 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4257 and that are not defined in the current unit.
4259 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4261 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4262 instead of a libcall for UNORDERED.
4264 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4267 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4268 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4270 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4273 * config/s390/s390.c (s390_set_current_function): Invoke
4274 s390_indirect_branch_settings also if fndecl didn't change.
4276 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4278 * config/rs6000/rs6000.md (blockage): Set length to zero.
4280 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4282 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4284 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4287 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4288 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4292 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4293 PARALLEL incoming that failed vt_get_decl_and_offset check.
4296 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4297 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4299 (get_variable_section): For decls in named .bss* sections pass true as
4300 second argument to bss_initializer_p.
4302 2018-02-09 Marek Polacek <polacek@redhat.com>
4303 Jakub Jelinek <jakub@redhat.com>
4306 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4307 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4308 Sync some changes from cxx_fold_indirect_ref.
4310 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4312 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4314 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4315 (BLOCK_INLINE_ENTRY_LABEL): New.
4316 (dwarf2out_var_location): Disregard inline entry markers.
4317 (inline_entry_data): New struct.
4318 (inline_entry_data_hasher): New hashtable type.
4319 (inline_entry_data_hasher::hash): New.
4320 (inline_entry_data_hasher::equal): New.
4321 (inline_entry_data_table): New variable.
4322 (add_high_low_attributes): Add DW_AT_entry_pc and
4323 DW_AT_GNU_entry_view attributes if a pending entry is found
4324 in inline_entry_data_table. Add old entry_pc attribute only
4325 if debug nonbinding markers are disabled.
4326 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4327 markers are enabled.
4328 (block_within_block_p, dwarf2out_inline_entry): New.
4329 (dwarf2out_finish): Check that no entries remained in
4330 inline_entry_data_table.
4331 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4332 (final_scan_insn, notice_source_line): Likewise.
4333 (rest_of_clean_state): Skip inline entry markers.
4334 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4336 * gimple.c (gimple_build_debug_inline_entry): New.
4337 * gimple.h (enum gimple_debug_subcode): Add
4338 GIMPLE_DEBUG_INLINE_ENTRY.
4339 (gimple_build_debug_inline_entry): Declare.
4340 (gimple_debug_inline_entry_p): New.
4341 (gimple_debug_nonbind_marker_p): Adjust.
4342 * insn-notes.def (INLINE_ENTRY): New.
4343 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4344 inline entry marker notes.
4345 (print_insn): Likewise.
4346 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4347 (INSN_DEBUG_MARKER_KIND): Likewise.
4348 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4349 * tree-inline.c (expand_call_inline): Build and insert
4350 debug_inline_entry stmt.
4351 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4352 inline entry blocks early, if nonbind markers are enabled.
4353 (dump_scope_block): Dump fragment info.
4354 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4355 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4356 (gimple_build_debug_inline_entry): New.
4357 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4358 Enable/disable inline entry points too.
4359 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4360 (DEBUG_INSN): Describe inline entry markers.
4362 * common.opt (gvariable-location-views): New.
4363 (gvariable-location-views=incompat5): New.
4364 * config.in: Rebuilt.
4365 * configure: Rebuilt.
4366 * configure.ac: Test assembler for view support.
4367 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4368 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4369 * dwarf2out.c (var_loc_view): New typedef.
4370 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4371 (dwarf2out_locviews_in_attribute): New.
4372 (dwarf2out_locviews_in_loclist): New.
4373 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4374 (enum dw_line_info_opcode): Add LI_adv_address.
4375 (struct dw_line_info_table): Add view.
4376 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4377 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4378 (zero_view_p): New variable.
4379 (ZERO_VIEW_P): New macro.
4380 (output_asm_line_debug_info): New.
4381 (struct var_loc_node): Add view.
4382 (add_AT_view_list, AT_loc_list): New.
4383 (add_var_loc_to_decl): Add view param. Test it against last.
4384 (new_loc_list): Add view params. Record them.
4385 (AT_loc_list_ptr): Handle loc and view lists.
4386 (view_list_to_loc_list_val_node): New.
4387 (print_dw_val): Handle dw_val_class_view_list.
4388 (size_of_die): Likewise.
4389 (value_format): Likewise.
4390 (loc_list_has_views): New.
4391 (gen_llsym): Set vl_symbol too.
4392 (maybe_gen_llsym, skip_loc_list_entry): New.
4393 (dwarf2out_maybe_output_loclist_view_pair): New.
4394 (output_loc_list): Output view list or entries too.
4395 (output_view_list_offset): New.
4396 (output_die): Handle dw_val_class_view_list.
4397 (output_dwarf_version): New.
4398 (output_compilation_unit_header): Use it.
4399 (output_skeleton_debug_sections): Likewise.
4400 (output_rnglists, output_line_info): Likewise.
4401 (output_pubnames, output_aranges): Update version comments.
4402 (output_one_line_info_table): Output view numbers in asm comments.
4403 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4404 Call maybe_gen_llsym.
4405 (loc_list_from_tree_1): Adjust.
4406 (add_AT_location_description): Create view list attribute if
4407 needed, check it's absent otherwise.
4408 (convert_cfa_to_fb_loc_list): Adjust.
4409 (maybe_emit_file): Call output_asm_line_debug_info for test.
4410 (dwarf2out_var_location): Reset views as needed. Precompute
4411 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4412 attribute. Set view.
4413 (new_line_info_table): Reset next view.
4414 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4415 (dwarf2out_source_line): Likewise. Output view resets and labels to
4416 the assembler, or select appropriate line info opcodes.
4417 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4418 (optimize_string_length): Catch it. Adjust.
4419 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4420 dw_val_class_view_list, and remove it if no longer needed.
4421 (hash_loc_list): Hash view numbers.
4422 (loc_list_hasher::equal): Compare them.
4423 (optimize_location_lists): Check whether a view list symbol is
4424 needed, and whether the locview attribute is present, and
4425 whether they match. Remove the locview attribute if no longer
4427 (index_location_lists): Call skip_loc_list_entry for test.
4428 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4429 Use output_dwarf_version.
4430 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4431 (struct dw_val_node): Add val_view_list.
4432 * final.c (SEEN_NEXT_VIEW): New.
4433 (set_next_view_needed): New.
4434 (clear_next_view_needed): New.
4435 (maybe_output_next_view): New.
4436 (final_start_function): Rename to...
4437 (final_start_function_1): ... this. Take pointer to FIRST,
4438 add SEEN parameter. Emit param bindings in the initial view.
4439 (final_start_function): Reintroduce SEEN-less interface.
4440 (final): Rename to...
4441 (final_1): ... this. Take SEEN parameter. Output final pending
4442 next view at the end.
4443 (final): Reintroduce seen-less interface.
4444 (final_scan_insn): Output pending next view before switching
4445 sections or ending a block. Mark the next view as needed when
4446 outputting variable locations. Notify debug backend of section
4447 changes, and of location view changes.
4448 (rest_of_handle_final): Adjust.
4449 * toplev.c (process_options): Autodetect value for debug variable
4450 location views option. Warn on incompat5 without -gdwarf-5.
4451 * doc/invoke.texi (gvariable-location-views): New.
4452 (gvariable-location-views=incompat5): New.
4453 (gno-variable-location-views): New.
4455 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4457 PR tree-optimization/84136
4458 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4459 that the result of find_edge is non-NULL.
4461 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4464 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4465 storing integer register in SImode. Fix cost of 256 and 512
4466 byte aligned SSE register store.
4468 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4470 * config/i386/i386.c (ix86_multiplication_cost): Fix
4471 multiplication cost for TARGET_AVX512DQ.
4473 2018-02-08 Marek Polacek <polacek@redhat.com>
4475 PR tree-optimization/84238
4476 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4479 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4481 PR tree-optimization/84265
4482 * tree-vect-stmts.c (vectorizable_store): Don't treat
4483 VMAT_CONTIGUOUS accesses as grouped.
4484 (vectorizable_load): Likewise.
4486 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4488 PR tree-optimization/81635
4489 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4490 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4491 (test_round_for_mask): New functions.
4492 (wide_int_cc_tests): Call test_round_for_mask.
4493 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4494 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4495 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4496 range returned by get_range_info.
4498 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4501 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4502 * symtab.c: Include builtins.h
4503 (symtab_node::output_to_lto_symbol_table_p): Move here
4504 from lto-streamer-out.c:output_symbol_p.
4505 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4506 (output_symbol_p): Move all logic to symtab.c
4507 (produce_symtab): Update.
4509 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4511 * config/s390/s390-opts.h (enum indirect_branch): Define.
4512 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4513 (s390_indirect_branch_via_thunk)
4514 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4515 (enum s390_indirect_branch_type): Define.
4516 * config/s390/s390.c (struct s390_frame_layout, struct
4517 machine_function): Remove.
4518 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4519 (indirect_branch_table_label_no, indirect_branch_table_name):
4521 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4522 (enum s390_indirect_branch_option): Define.
4523 (s390_return_addr_from_memory): New function.
4524 (s390_handle_string_attribute): New function.
4525 (s390_attribute_table): Add new attribute handler.
4526 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4527 (s390_indirect_branch_via_thunk): New function.
4528 (s390_indirect_branch_via_inline_thunk): New function.
4529 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4530 sibling call optimization for non z10 compiles.
4531 (s390_emit_call): Force indirect branch target to be a single
4532 register. Add r1 clobber for non-z10 compiles.
4533 (s390_emit_epilogue): Emit return jump via return_use expander.
4534 (s390_reorg): Handle JUMP_INSNs as execute targets.
4535 (s390_option_override_internal): Perform validity checks for the
4536 new command line options.
4537 (s390_indirect_branch_attrvalue): New function.
4538 (s390_indirect_branch_settings): New function.
4539 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4540 (s390_output_indirect_thunk_function): New function.
4541 (s390_code_end): Implement target hook.
4542 (s390_case_values_threshold): Implement target hook.
4543 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4545 * config/s390/s390.h (struct s390_frame_layout)
4546 (struct machine_function): Move here from s390.c.
4547 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4548 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4549 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4550 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4551 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4552 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4553 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4554 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4555 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4556 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4557 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4558 (mnemonic attribute): Add values which aren't recognized
4560 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4561 pattern for branch conversion. Fix mnemonic attribute.
4562 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4563 indirect branch via thunk if requested.
4564 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4565 ("*indirect_jump"): Disable for branch conversion using out of
4567 ("indirect_jump_via_thunk<mode>_z10")
4568 ("indirect_jump_via_thunk<mode>")
4569 ("indirect_jump_via_inlinethunk<mode>_z10")
4570 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4571 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4572 ("casesi_jump_via_inlinethunk<mode>_z10")
4573 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4574 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4575 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4576 ("*indirect2_jump"): Disable for branch conversion.
4577 ("casesi_jump"): Turn into expander and expand patterns for branch
4579 ("return_use"): New expander.
4580 ("*return"): Emit return via thunk and rename it to ...
4581 ("*return<mode>"): ... this one.
4582 * config/s390/s390.opt: Add new options and and enum for the
4585 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4587 * lra-constraints.c (match_reload): Unconditionally use
4588 gen_lowpart_SUBREG, rather than selecting between that
4589 and equivalent gen_rtx_SUBREG code.
4591 2018-02-08 Richard Biener <rguenther@suse.de>
4593 PR tree-optimization/84233
4594 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4595 changed flag instead of boguously re-using phi_inserted.
4597 2018-02-08 Martin Jambor <mjambor@suse.cz>
4599 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4600 static local variables.
4602 2018-02-08 Richard Biener <rguenther@suse.de>
4604 PR tree-optimization/84278
4605 * tree-vect-stmts.c (vectorizable_store): When looking for
4606 smaller vector types to perform grouped strided loads/stores
4607 make sure the mode is supported by the target.
4608 (vectorizable_load): Likewise.
4610 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4612 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4613 Increase LDP/STP opportunities by adding adjacent callee-saves.
4615 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4617 PR rtl-optimization/84068
4618 PR rtl-optimization/83459
4619 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4621 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4623 PR tree-optimization/84224
4624 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4625 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4628 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4631 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4632 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4633 count, remove one USE.
4635 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4637 * doc/install.texi (Configuration): Document the
4638 --with-long-double-format={ibm,ieee} PowerPC configuration
4642 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4643 Convert from define_expand to be define_insn_and_split. Rework
4644 float/double/_Float128 conversions to QI/HI/SImode to work with
4645 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4646 conversions to QI/HImode types did a store and then a load to
4647 truncate the value. For conversions to VSX registers, don't split
4648 the insn, instead emit the code directly. Use the code iterator
4649 any_fix to combine signed and unsigned conversions.
4650 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4651 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4652 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4653 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4654 (fix_<mode>di2_hw): Likewise.
4655 (fixuns_<mode>di2_hw): Likewise.
4656 (fix_<mode>si2_hw): Likewise.
4657 (fixuns_<mode>si2_hw): Likewise.
4658 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4659 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4660 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4661 fix<uns>_trunc<SFDF:mode>si2_p8.
4662 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4664 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4665 (fix<uns>_<mode>_mem): Likewise.
4666 (fctiw<u>z_<mode>_mem): Likewise.
4667 (fix<uns>_<mode>_mem): Likewise.
4668 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4669 the register allocator from doing a direct move to the GPRs to do
4670 a store, and instead use the ISA 3.0 store byte/half-word from
4671 vector register instruction. For IEEE 128-bit floating point,
4672 also optimize stores of 32-bit ints.
4673 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4675 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4677 * genextract.c (push_pathstr_operand): New function to support
4679 (walk_rtx): Call push_pathstr_operand.
4680 (print_path): Support [a-zA-Z].
4682 2018-02-07 Richard Biener <rguenther@suse.de>
4684 PR tree-optimization/84037
4685 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4686 (cse_and_gimplify_to_preheader): Declare.
4687 (vect_get_place_in_interleaving_chain): Likewise.
4688 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4690 (_loop_vec_info::~_loop_vec_info): Delete it.
4691 (cse_and_gimplify_to_preheader): New function.
4692 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4693 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4694 (vectorizable_load): Likewise. For grouped stores always base
4695 the IV on the first element.
4696 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4697 condition before gimplifying.
4699 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4701 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4702 *DIV_EXPR and *MOD_EXPR.
4704 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4707 * config/i386/i386.c (ix86_option_override_internal): Mask out
4708 the CF_SET bit when checking -fcf-protection.
4710 2018-02-07 Tom de Vries <tom@codesourcery.com>
4713 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4716 2018-02-07 Richard Biener <rguenther@suse.de>
4718 PR tree-optimization/84204
4719 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4722 PR tree-optimization/84205
4723 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4724 special-case isl_ast_op_zdiv_r.
4726 PR tree-optimization/84223
4727 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4728 Only add conditions from within the region.
4729 (gather_bbs::after_dom_children): Adjust.
4731 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4734 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4735 * config/avr/avr.md: Only post-reload split REG-REG moves if
4736 either register is GENERAL_REG_P.
4738 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4740 PR tree-optimization/84235
4741 * tree-ssa-scopedtables.c
4742 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4743 if the subtraction is performed in floating point type where NaNs are
4744 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4745 build 1. Formatting fix.
4747 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4750 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4751 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4752 and skip it regardless of bb boundaries. Use CALL_P macro,
4753 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4754 unnecessarily, formatting fix.
4756 2018-02-06 Michael Collison <michael.collison@arm.com>
4758 * config/arm/thumb2.md:
4759 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4760 (*thumb_mov_notscc): Ditto.
4762 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4765 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4766 unsigned_fix, not "s".
4768 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4770 * configure.ac (gcc_fn_eh_frame_ro): New function.
4771 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4772 correct .eh_frame permissions.
4773 * configure: Regenerate.
4775 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4777 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4780 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4782 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4783 Display warning message for -mno-speculate-indirect-jumps.
4785 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4787 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4789 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4791 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4793 PR tree-optimization/84225
4794 * tree-eh.c (find_trapping_overflow): Only call
4795 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4797 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4800 * config/i386/i386.c: Reimplement the check of possible options
4801 -mibt/-mshstk conbination. Change error messages.
4802 * doc/invoke.texi: Fix a typo: remove extra '='.
4804 2018-02-06 Marek Polacek <polacek@redhat.com>
4806 PR tree-optimization/84228
4807 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4809 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4812 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4813 emitted arch directives.
4814 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4815 __ARM_FEATURE_COPROC before changing architectures.
4817 2018-02-06 Richard Biener <rguenther@suse.de>
4819 * config/i386/i386.c (print_reg): Fix typo.
4820 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4822 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4824 * configure: Regenerate.
4826 2018-02-05 Martin Sebor <msebor@redhat.com>
4828 PR tree-optimization/83369
4829 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4832 2018-02-05 Martin Liska <mliska@suse.cz>
4834 * doc/invoke.texi: Cherry-pick upstream r323995.
4836 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4838 * ira.c (ira_init_register_move_cost): Adjust comment.
4840 2018-02-05 Martin Liska <mliska@suse.cz>
4842 PR gcov-profile/84137
4843 * doc/gcov.texi: Fix typo in documentation.
4845 2018-02-05 Martin Liska <mliska@suse.cz>
4847 PR gcov-profile/83879
4848 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4849 using dlopen functionality.
4851 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4853 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4854 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4855 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4856 _mm_maskz_range_round_ss): New intrinsics.
4857 (__builtin_ia32_rangesd128_round)
4858 (__builtin_ia32_rangess128_round): Remove.
4859 (__builtin_ia32_rangesd128_mask_round,
4860 __builtin_ia32_rangess128_mask_round): New builtins.
4861 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4862 __builtin_ia32_rangess128_round): Remove.
4863 (__builtin_ia32_rangesd128_mask_round,
4864 __builtin_ia32_rangess128_mask_round): New builtins.
4865 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4866 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4867 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4868 "<round_saeonly_constraint>")): Changed to ...
4869 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4870 "<round_saeonly_scalar_constraint>")): ... this.
4871 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4872 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4873 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4874 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4875 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4877 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4879 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4881 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4882 Remove all values except native, 8540 and 8548.
4884 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4886 * config/i386/i386.c (ix86_output_function_return): Pass
4887 INVALID_REGNUM, instead of -1, as invalid register number to
4888 indirect_thunk_name and output_indirect_thunk.
4890 2018-02-02 Julia Koval <julia.koval@intel.com>
4892 * config.gcc: Add -march=icelake.
4893 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4894 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4895 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4896 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4897 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4898 (processor_target_table): Add icelake.
4899 (ix86_option_override_internal): Handle new PTAs.
4900 (get_builtin_code_for_version): Handle icelake.
4901 (M_INTEL_COREI7_ICELAKE): New.
4902 (fold_builtin_cpu): Handle icelake.
4903 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4904 * doc/invoke.texi: Add -march=icelake.
4906 2018-02-02 Julia Koval <julia.koval@intel.com>
4908 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4909 to wide_int_bitmask.
4910 * wide-int-bitmask.h: New.
4912 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4915 * config/i386/i386.md: Replace Pmode with word_mode in
4916 builtin_setjmp_setup and builtin_longjmp to support x32.
4918 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4922 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4924 (rs6000_supported_cpu_names): New static variable.
4925 (linux_cpu_translation_table): Likewise.
4926 (elf_platform) <cpu>: Define new static variable and use it.
4927 Translate kernel AT_PLATFORM name to canonical name if needed.
4928 Error if platform name is unknown.
4930 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4933 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4935 2018-02-01 Jeff Law <law@redhat.com>
4938 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4939 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4940 the scratch if RELEASE_VIA_POP is false.
4941 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4942 If we have to save a temporary register, decrement SIZE appropriately.
4943 Pass new arguments to release_scratch_register_on_entry.
4944 (ix86_adjust_stack_and_probe): Likewise.
4945 (ix86_emit_probe_stack_range): Pass new arguments to
4946 release_scratch_register_on_entry.
4948 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4950 PR rtl-optimization/84157
4951 * combine.c (change_zero_ext): Use REG_P predicate in
4952 front of HARD_REGISTER_P predicate.
4954 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4956 * config/avr/avr.c (avr_option_override): Move disabling of
4957 -fdelete-null-pointer-checks to...
4958 * common/config/avr/avr-common.c (avr_option_optimization_table):
4961 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4963 PR tree-optimization/81635
4964 * tree-data-ref.c (split_constant_offset_1): For types that
4965 wrap on overflow, try to use range info to prove that wrapping
4968 2018-02-01 Renlin Li <renlin.li@arm.com>
4971 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4973 (aarch64_register_move_cost): Likewise.
4974 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4976 (REG_CLASS_NAMES): Likewise.
4977 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4978 TAILCALL_ADDR_REGS. Remove IP registers.
4979 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4981 2018-02-01 Richard Biener <rguenther@suse.de>
4983 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4984 for specifying RPO order and allow NULL for that.
4985 * domwalk.c (dom_walker::dom_walker): Likewise.
4986 (dom_walker::walk): Handle NULL RPO order.
4987 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4989 (rewrite_update_dom_walker): Likewise.
4990 (mark_def_dom_walker): Likewise.
4992 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4994 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4995 (aarch64_maybe_expand_sve_subreg_move): Declare.
4996 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4997 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4999 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
5000 that are semantically a reverse operation.
5001 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
5002 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
5003 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
5005 (aarch64_can_change_mode_class): For big-endian, forbid changes
5006 between two SVE modes if they have different element sizes.
5008 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5010 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
5011 the TImode handling for big-endian targets.
5013 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5015 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
5016 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
5018 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
5019 Remove BSWAP handing for big-endian targets and use the form of
5020 LD1RQ appropariate for the mode.
5022 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5024 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
5025 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
5028 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
5031 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
5032 check for operands that need to go through aarch64_sve_reload_be.
5034 2018-02-01 Jakub Jelinek <jakub@redhat.com>
5036 PR tree-optimization/81661
5037 PR tree-optimization/84117
5038 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
5039 * tree-eh.c: Include gimplify.h.
5040 (find_trapping_overflow, replace_trapping_overflow,
5041 rewrite_to_non_trapping_overflow): New functions.
5042 * tree-vect-loop.c: Include tree-eh.h.
5043 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
5044 * tree-data-ref.c: Include tree-eh.h.
5045 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
5047 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
5049 PR rtl-optimization/84123
5050 * combine.c (change_zero_ext): Check if hard register satisfies
5051 can_change_dest_mode before calling gen_lowpart_SUBREG.
5053 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
5056 * ira.c (ira_init_register_move_cost): Remove assert.
5058 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5060 PR rtl-optimization/84071
5061 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
5062 * doc/tm.texi: Regenerate.
5064 2018-01-31 Richard Biener <rguenther@suse.de>
5066 PR tree-optimization/84132
5067 * tree-data-ref.c (analyze_miv_subscript): Properly
5068 check whether evolution_function_is_affine_multivariate_p
5069 before calling gcd_of_steps_may_divide_p.
5071 2018-01-31 Julia Koval <julia.koval@intel.com>
5074 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
5075 * config/i386/i386.md (rdpid_rex64) New.
5076 (rdpid): Make 32bit only.
5078 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
5081 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
5082 an IDENTIFIER_NODE for FUNCTION_TYPE's.
5084 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5087 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5089 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5091 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
5093 PR rtl-optimization/84071
5094 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
5095 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
5097 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5099 * config/arc/arc.c (arc_handle_aux_attribute): New function.
5100 (arc_attribute_table): Add 'aux' attribute.
5101 (arc_in_small_data_p): Consider aux like variables.
5102 (arc_is_aux_reg_p): New function.
5103 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
5104 (arc_get_aux_arg): New function.
5105 (prepare_move_operands): Handle aux-register access.
5106 (arc_handle_aux_attribute): New function.
5107 * doc/extend.texi (ARC Variable attributes): Add subsection.
5109 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
5111 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
5112 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
5113 (arc_attribute_table): Add 'uncached' attribute.
5114 (arc_print_operand): Print '.di' flag for uncached memory
5116 (arc_in_small_data_p): Do not consider for small data the uncached
5118 (arc_is_uncached_mem_p): New function.
5119 * config/arc/predicates.md (compact_store_memory_operand): Check
5120 for uncached memory accesses.
5121 (nonvol_nonimm_operand): Likewise.
5122 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5124 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5127 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5128 falign-loops=): Add Optimization flag.
5130 2018-01-30 Jeff Law <law@redhat.com>
5133 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5134 INT_REGISTERS_SAVED. Check it prior to calling
5135 get_scratch_register_on_entry.
5136 (ix86_adjust_stack_and_probe): Similarly.
5137 (ix86_emit_probe_stack_range): Similarly.
5138 (ix86_expand_prologue): Corresponding changes.
5140 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5143 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5144 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5146 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5149 * lra-constraints.c (curr_insn_transform): Process AND in the
5152 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5154 PR rtl-optimization/83986
5155 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5156 dependence against last_pending_memory_flush in addition to
5159 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5161 PR tree-optimization/81611
5162 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5165 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5168 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5171 2018-01-30 Richard Biener <rguenther@suse.de>
5172 Jakub Jelinek <jakub@redhat.com>
5174 PR tree-optimization/84111
5175 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5176 inner loops added during recursion, as they don't have up-to-date
5179 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5182 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5183 (can_inline_edge_by_limits_p): ... here.
5184 (can_early_inline_edge_p, check_callers,
5185 update_caller_keys, update_callee_keys, recursive_inlining,
5186 add_new_edges_to_heap, speculation_useful_p,
5187 inline_small_functions,
5188 inline_small_functions, flatten_function,
5189 inline_to_all_callers_1): Update.
5191 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5193 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5196 2018-01-30 Richard Biener <rguenther@suse.de>
5198 PR tree-optimization/83008
5199 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5200 invariant and constant vector uses in stmts when they need
5203 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5206 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5207 * configure: Regenerate.
5209 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5211 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5213 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5214 Use gen_rtx_REG rather than gen_lowpart.
5216 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5218 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5219 rather than 0 when creating partial subregs.
5221 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5223 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5226 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5229 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5230 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5231 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5232 flags. This restores the settings used before the 2017-07-24.
5233 Turning off pre increment/decrement/modify allows IVOPTS to
5234 optimize DF/SF loops where the index is an int.
5236 2018-01-29 Richard Biener <rguenther@suse.de>
5237 Kelvin Nilsen <kelvin@gcc.gnu.org>
5240 * tree-vect-stmts.c (vectorizable_call): Don't call
5241 targetm.vectorize_builtin_md_vectorized_function if callee is
5244 2018-01-22 Carl Love <cel@us.ibm.com>
5246 * doc/extend.tex: Fix typo in second arg in
5247 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5249 2018-01-29 Richard Biener <rguenther@suse.de>
5251 PR tree-optimization/84086
5252 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5253 (flush_ssaname_freelist): When SSA names were released reset
5254 the SCEV hash table.
5256 2018-01-29 Richard Biener <rguenther@suse.de>
5258 PR tree-optimization/84057
5259 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5260 removed paths when removing edges.
5262 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5264 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5265 -mfunction-return=@var{choice}.
5267 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5270 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5272 (layout::print_source_line): Likewise.
5273 (test_get_line_width_without_trailing_whitespace): Add test cases.
5275 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5278 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5281 2018-01-26 Jim Wilson <jimw@sifive.com>
5283 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5285 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5288 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5290 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5291 and CMP + SUB-immediate -> SUBS.
5293 2018-01-26 Martin Sebor <msebor@redhat.com>
5295 PR tree-optimization/83896
5296 * tree-ssa-strlen.c (get_string_len): Rename...
5297 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5298 Avoid assuming length is constant.
5299 (handle_char_store): Use HOST_WIDE_INT for string length.
5301 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5304 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5305 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5307 2018-01-26 Richard Biener <rguenther@suse.de>
5309 PR rtl-optimization/84003
5310 * dse.c (record_store): Only record redundant stores when
5311 the earlier store aliases at least all accesses the later one does.
5313 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5315 PR rtl-optimization/83985
5316 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5317 REG_CFA_RESTORE insns.
5318 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5319 REG_CFA_RESTORE insns here.
5322 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5323 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5325 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5327 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5328 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5329 (arc_init): Likewise.
5330 (arc_override_options): Likewise.
5331 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5333 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5335 * config/arc/arc.h (TARGET_DBNZ): Define.
5336 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5337 properly set the tune attribute.
5338 (dbnz): Use TARGET_DBNZ guard.
5339 * config/arc/arc.opt (mtune): Add core3 option.
5341 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5343 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5344 recognize new pic like addresses.
5345 (arc_delegitimize_address): Clean up.
5347 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5349 * config/arc/arc-arches.def: Option mrf16 valid for all
5351 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5352 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5353 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5354 * config/arc/arc-tables.opt: Regenerate.
5355 * config/arc/arc.c (arc_conditional_register_usage): Handle
5356 reduced register file case.
5357 (arc_file_start): Set must have build attributes.
5358 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5360 * config/arc/arc.opt (mrf16): Add new option.
5361 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5362 * config/arc/genmultilib.awk: Handle new mrf16 option.
5363 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5364 * config/arc/t-multilib: Regenerate.
5365 * doc/invoke.texi (ARC Options): Document mrf16 option.
5367 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5369 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5370 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5371 (arc_attribute_table): Add 'secure_call' attribute.
5372 (arc_print_operand): Print secure call operand.
5373 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5375 (arc_is_secure_call_p): New function. * config/arc/arc.md
5376 (call_i): Add support for sjli instruction.
5377 (call_value_i): Likewise.
5378 * config/arc/constraints.md (Csc): New constraint.
5380 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5381 John Eric Martin <John.Martin@emmicro-us.com>
5383 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5384 * config/arc/arc.c (_arc_jli_section): New struct.
5385 (arc_jli_section): New type.
5386 (rc_jli_sections): New static variable.
5387 (arc_handle_jli_attribute): New function.
5388 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5389 (arc_file_end): New function.
5390 (TARGET_ASM_FILE_END): Define.
5391 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5392 (arc_add_jli_section): New function.
5393 (jli_call_scan): Likewise.
5394 (arc_reorg): Call jli_call_scan.
5395 (arc_output_addsi): Remove 'S' from printing asm operand.
5396 (arc_is_jli_call_p): New function.
5397 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5399 (movhi_insn): Likewise.
5400 (movsi_insn): Likewise.
5401 (movsi_set_cc_insn): Likewise.
5402 (loadqi_update): Likewise.
5403 (load_zeroextendqisi_update): Likewise.
5404 (load_signextendqisi_update): Likewise.
5405 (loadhi_update): Likewise.
5406 (load_zeroextendhisi_update): Likewise.
5407 (load_signextendhisi_update): Likewise.
5408 (loadsi_update): Likewise.
5409 (loadsf_update): Likewise.
5410 (movsicc_insn): Likewise.
5411 (bset_insn): Likewise.
5412 (bxor_insn): Likewise.
5413 (bclr_insn): Likewise.
5414 (bmsk_insn): Likewise.
5415 (bicsi3_insn): Likewise.
5416 (cmpsi_cc_c_insn): Likewise.
5417 (movsi_ne): Likewise.
5418 (movsi_cond_exec): Likewise.
5419 (clrsbsi2): Likewise.
5430 (call_i): Remove 'S' asm letter, add jli instruction.
5431 (call_value_i): Likewise.
5432 * config/arc/arc.op (mjli-always): New option.
5433 * config/arc/constraints.md (Cji): New constraint.
5434 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5436 (subsf3_fpx): Likewise.
5437 (mulsf3_fpx): Likewise.
5438 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5440 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5442 * doc/invoke.texi (ARC): Document mjli-always option.
5444 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5446 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5447 avoid addition with 0 and use incw and decw where possible.
5449 2018-01-26 Richard Biener <rguenther@suse.de>
5451 PR tree-optimization/81082
5452 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5453 association if it requires casting to unsigned.
5454 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5455 from fold_plusminus_mult_expr to catch important cases late when
5456 range info is available.
5458 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5460 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5461 * configure.ac (hidden_linkonce): New test.
5462 * configure: Regenerate.
5463 * config.in: Regenerate.
5465 2018-01-26 Julia Koval <julia.koval@intel.com>
5467 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5468 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5469 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5470 _mm_mask_bitshuffle_epi64_mask): Fix type.
5471 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5472 USI_FTYPE_V4DI_V4DI_USI): Remove.
5473 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5474 __builtin_ia32_vpshufbitqmb256_mask,
5475 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5476 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5477 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5479 2018-01-26 Alan Modra <amodra@gmail.com>
5482 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5483 UNSPEC_VBPERMQ. Sort other unspecs.
5485 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5487 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5489 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5492 * predict.c (drop_profile): Do not push/pop cfun; update also
5494 (handle_missing_profiles): Fix logic looking for zero profiles.
5496 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5499 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5500 on functions with #pragma omp declare simd or functions with simd
5502 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5503 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5504 Remove trailing \n from warning_at calls.
5506 2018-01-25 Tom de Vries <tom@codesourcery.com>
5509 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5510 for neutered workers.
5512 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5515 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5516 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5518 2018-01-24 Jeff Law <law@redhat.com>
5521 * i386.c (get_probe_interval): Move to earlier point.
5522 (ix86_compute_frame_layout): If -fstack-clash-protection and
5523 the frame is larger than the probe interval, then use pushes
5524 to save registers rather than reg->mem moves.
5525 (ix86_expand_prologue): Remove conditional for int_registers_saved
5528 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5531 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5532 min/max for never referenced object.
5534 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5537 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5539 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5540 attributes from DECL_ATTRIBUTES (decl) without affecting
5541 DECL_ATTRIBUTES (current_function_decl).
5542 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5543 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5545 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5547 PR tree-optimization/83979
5548 * fold-const.c (fold_comparison): Use constant_boolean_node
5549 instead of boolean_{true,false}_node.
5551 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5553 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5556 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5558 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5559 Simplify the clause that sets the length attribute.
5560 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5561 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5562 clause that sets the length attribute.
5563 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5565 2018-01-24 Tom de Vries <tom@codesourcery.com>
5568 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5569 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5570 Add strict parameter.
5571 (prevent_branch_around_nothing): Insert dummy insn between branch to
5572 label and label with no ptx insn inbetween.
5573 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5575 2018-01-24 Tom de Vries <tom@codesourcery.com>
5578 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5579 for neutered threads in warp.
5580 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5582 2018-01-24 Richard Biener <rguenther@suse.de>
5584 PR tree-optimization/83176
5585 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5588 2018-01-24 Richard Biener <rguenther@suse.de>
5590 PR tree-optimization/82819
5591 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5592 code generating pluses that are no-ops in the target precision.
5594 2018-01-24 Richard Biener <rguenther@suse.de>
5597 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5599 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5601 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5602 to merge probabilities.
5603 * predict.c (probably_never_executed): Also mark as cold functions
5604 with global 0 profile and guessed local profile.
5605 * profile-count.c (profile_probability::combine_with_count): New
5607 * profile-count.h (profile_probability::operator*,
5608 profile_probability::operator*=, profile_probability::operator/,
5609 profile_probability::operator/=): Reduce precision to adjusted
5610 and set value to guessed on contradictory divisions.
5611 (profile_probability::combine_with_freq): Remove.
5612 (profile_probability::combine_wiht_count): Declare.
5613 (profile_count::force_nonzero):: Set to adjusted.
5614 (profile_count::probability_in):: Set quality to adjusted.
5615 * tree-ssa-tail-merge.c (replace_block_by): Use
5618 2018-01-23 Andrew Waterman <andrew@sifive.com>
5619 Jim Wilson <jimw@sifive.com>
5621 * config/riscv/riscv.c (riscv_stack_boundary): New.
5622 (riscv_option_override): Set riscv_stack_boundary. Handle
5623 riscv_preferred_stack_boundary_arg.
5624 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5625 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5626 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5627 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5628 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5629 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5631 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5634 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5635 of struct ix86_frame.
5636 (ix86_expand_epilogue): Likewise. Add a local variable for
5637 the reg_save_offset field in struct ix86_frame.
5639 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5641 PR tree-optimization/82604
5642 * tree-loop-distribution.c (enum partition_kind): New enum item
5643 PKIND_PARTIAL_MEMSET.
5644 (partition_builtin_p): Support above new enum item.
5645 (generate_code_for_partition): Ditto.
5646 (compute_access_range): Differentiate cases that equality can be
5647 proven at all loops, the innermost loops or no loops.
5648 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5649 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5650 (finalize_partitions, distribute_loop): Don't fuse partition of
5651 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5652 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5655 2018-01-23 Martin Liska <mliska@suse.cz>
5657 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5658 order to ignore the predictor.
5659 (PRED_POLYMORPHIC_CALL): Likewise.
5660 (PRED_RECURSIVE_CALL): Likewise.
5662 2018-01-23 Martin Liska <mliska@suse.cz>
5664 * tree-profile.c (tree_profiling): Print function header to
5665 aware reader which function we are working on.
5666 * value-prof.c (gimple_find_values_to_profile): Do not print
5667 not interesting value histograms.
5669 2018-01-23 Martin Liska <mliska@suse.cz>
5671 * profile-count.h (enum profile_quality): Add
5672 profile_uninitialized as the first value. Do not number values
5673 as they are zero based.
5674 (profile_count::verify): Update sanity check.
5675 (profile_probability::verify): Likewise.
5677 2018-01-23 Nathan Sidwell <nathan@acm.org>
5679 * doc/invoke.texi (ffor-scope): Deprecate.
5681 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5683 PR tree-optimization/83510
5684 * domwalk.c (set_all_edges_as_executable): New function.
5685 (dom_walker::dom_walker): Convert bool param
5686 "skip_unreachable_blocks" to enum reachability. Move setup of
5687 edge flags to set_all_edges_as_executable and only do it when
5688 reachability is REACHABLE_BLOCKS.
5689 * domwalk.h (enum dom_walker::reachability): New enum.
5690 (dom_walker::dom_walker): Convert bool param
5691 "skip_unreachable_blocks" to enum reachability.
5692 (set_all_edges_as_executable): New decl.
5693 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5694 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5696 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5697 but converting true to REACHABLE_BLOCKS.
5698 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5700 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5701 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5702 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5704 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5705 if check_all_array_refs will be called.
5707 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5709 * tree.c (selftest::test_location_wrappers): Add more test
5712 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5714 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5715 (selftest::test_bit_in_range): Likewise.
5717 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5720 * doc/sourcebuild.texi (vect_float): Say that the selector
5721 only describes the situation when -funsafe-math-optimizations is on.
5722 (vect_float_strict): Document.
5724 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5726 PR tree-optimization/83965
5727 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5728 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5729 instead of checking only for a reduction.
5730 (vect_recog_widen_sum_pattern): Likewise.
5732 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5734 * predict.c (probably_never_executed): Only use precise profile info.
5735 (compute_function_frequency): Skip after inlining hack since we now
5736 have quality checking.
5738 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5740 * profile-count.h (profile_probability::very_unlikely,
5741 profile_probability::unlikely, profile_probability::even): Set
5742 precision to guessed.
5744 2018-01-23 Richard Biener <rguenther@suse.de>
5746 PR tree-optimization/83963
5747 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5748 Properly terminate dominator walk when crossing the exit edge not
5749 when visiting its source block.
5751 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5754 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5755 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5757 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5759 PR tree-optimization/83957
5760 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5761 semicolon after for body surrounded by braces.
5763 PR tree-optimization/83081
5764 * profile-count.h (profile_probability::split): New method.
5765 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5766 Use profile_probability::split.
5767 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5768 when splitting a single conditional jump into 2.
5770 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5772 PR tree-optimization/69452
5773 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5776 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5778 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5779 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5780 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5782 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5784 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5785 * config/rl78/rl78.md (movdi): New define_expand.
5786 * config/rl78/rl78.c (rl78_split_movdi): New function.
5788 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5791 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5793 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5794 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5795 128-bit to produce an UNSPEC move to get the double word with the
5796 signbit and then a shift directly to do signbit.
5797 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5798 implementation with a new version that just does either a direct
5799 move or a regular move. Move memory interface to separate insns.
5800 Move insns so they are next to the expander.
5801 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5802 with signbit move. Split big and little endian case.
5803 (signbit<mode>2_dm_mem_le): Likewise.
5804 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5805 (signbit<mode>2_dm2): Likewise.
5807 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5809 * config/rl78/rl78.md (anddi3): New define_expand.
5811 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5813 * config/rl78/rl78.md (umindi3): New define_expand.
5815 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5817 * config/rl78/rl78.md (smindi3): New define_expand.
5819 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5821 * config/rl78/rl78.md (smaxdi3): New define_expand.
5823 2018-01-22 Carl Love <cel@us.ibm.com>
5825 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5826 LVX_V1TI): Add macro expansion.
5827 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5828 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5829 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5830 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5831 Change check to determine if the instruction is a byte reversing
5832 entry. Fix typo in comment.
5833 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5834 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5835 Add def_builtin calls for new builtins.
5836 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5837 Add define_insn expansion.
5839 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5841 * config/rl78/rl78.md (umaxdi3): New define_expand.
5843 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5845 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5846 for non-QImode registers.
5848 2018-01-22 Richard Biener <rguenther@suse.de>
5850 PR tree-optimization/83963
5851 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5852 including the loop exit block.
5853 (scop_detection::merge_sese): Likewise.
5854 (scop_detection::add_scop): Do it here instead.
5856 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5858 * doc/sourcebuild.texi (arm_softfloat): Document.
5860 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5863 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5864 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5865 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5867 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5868 David Edelsohn <dje.gcc@gmail.com>
5871 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5872 Change "crset eq" to "crset 2".
5873 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5874 (*call_indirect_aix<mode>_nospec): Likewise.
5875 (*call_value_indirect_aix<mode>_nospec): Likewise.
5876 (*call_indirect_elfv2<mode>_nospec): Likewise.
5877 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5878 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5879 change assembly output from . to $.
5880 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5881 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5882 (*tablejump<mode>_internal1_nospec): Likewise.
5884 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5887 * config/sh/sh_optimize_sett_clrt.cc:
5888 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5890 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5892 PR tree-optimization/83940
5893 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5894 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5895 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5896 use it to pass back the definition type.
5897 (vect_check_store_rhs): Likewise rhs_dt_out.
5898 (vect_build_gather_load_calls): Add a mask_dt argument and use
5899 it instead of a call to vect_is_simple_use.
5900 (vectorizable_store): Update calls to vect_check_load_store_mask
5901 and vect_check_store_rhs. Use the dt returned by the latter instead
5902 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5903 instead of calls to vect_is_simple_use. Pass the scalar rather
5904 than the vector operand to vect_is_simple_use when handling
5905 second and subsequent copies of an rhs value.
5906 (vectorizable_load): Update calls to vect_check_load_store_mask
5907 and vect_build_gather_load_calls. Use the cached mask_dt and
5908 gs_info.offset_dt instead of calls to vect_is_simple_use.
5910 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5913 * tree-emutls.c: Include gimplify.h.
5914 (lower_emutls_2): New function.
5915 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5916 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5917 it before further processing.
5920 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5921 UINTVAL (trueop1) instead of INTVAL (op1).
5923 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5927 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5928 INCOMING_FRAME_SP_OFFSET if not defined.
5929 (scan_trace): Add ENTRY argument. If true and
5930 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5931 emit a note to adjust the CFA offset.
5932 (create_cfi_notes): Adjust scan_trace callers.
5933 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5934 INCOMING_FRAME_SP_OFFSET in the CIE.
5935 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5936 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5938 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5939 * doc/tm.texi: Regenerated.
5941 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5943 PR rtl-optimization/83147
5944 * lra-constraints.c (remove_inheritance_pseudos): Use
5945 lra_substitute_pseudo_within_insn.
5947 2018-01-19 Tom de Vries <tom@codesourcery.com>
5948 Cesar Philippidis <cesar@codesourcery.com>
5951 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5953 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5956 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5957 spaces for function labels.
5959 2018-01-19 Martin Liska <mliska@suse.cz>
5961 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5962 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5963 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5964 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5965 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5966 (PRED_CONST_RETURN): Change from 69 to 65.
5967 (PRED_NULL_RETURN): Change from 91 to 71.
5968 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5969 (PRED_LOOP_GUARD): Change from 66 to 73.
5971 2018-01-19 Martin Liska <mliska@suse.cz>
5973 * predict.c (predict_insn_def): Add new assert.
5974 (struct branch_predictor): Change type to signed integer.
5975 (test_prediction_value_range): Amend test to cover
5977 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5978 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5979 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5980 (PRED_LOOP_IV_COMPARE): Likewise.
5981 * predict.h (PROB_UNINITIALIZED): Define new constant.
5983 2018-01-19 Martin Liska <mliska@suse.cz>
5985 * predict.c (dump_prediction): Add new format for
5986 analyze_brprob.py script which is enabled with -details
5988 * profile-count.h (precise_p): New function.
5990 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5992 PR tree-optimization/83922
5993 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5994 there are no statements that need masking.
5995 (vect_active_double_reduction_p): New function.
5996 (vect_analyze_loop_operations): Use it when handling phis that
5997 are not in the loop header.
5999 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
6001 PR tree-optimization/83914
6002 * tree-vect-loop.c (vectorizable_induction): Don't convert
6003 init_expr or apply the peeling adjustment for inductions
6004 that are nested within the vectorized loop.
6006 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6008 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
6011 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6015 * function.h (gimplify_parameters): Add gimple_seq * argument.
6016 * function.c: Include gimple.h and options.h.
6017 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
6018 for the added local temporaries if needed.
6019 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
6020 if there are any parameter cleanups, wrap whole body into a
6021 try/finally with the cleanups.
6023 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
6026 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
6027 Use GET_MODE_CLASS for scalar floating point.
6029 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6033 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
6034 Fix call of call_cgraph_insertion_hooks.
6036 2018-01-18 Martin Sebor <msebor@redhat.com>
6038 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
6040 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
6043 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
6046 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
6049 * common.opt: (-ffile-prefix-map): New option.
6050 * opts.c (common_handle_option): Defer it.
6051 * opts-global.c (handle_common_deferred_options): Handle it.
6052 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
6053 * file-prefix-map.h: New file.
6054 (remap_debug_filename, add_debug_prefix_map): ...here.
6055 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
6056 * final.c (debug_prefix_map, add_debug_prefix_map
6057 remap_debug_filename): Move to...
6058 * file-prefix-map.c: New file.
6059 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
6060 generalize, get rid of alloca(), use strrchr() instead of strchr().
6061 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
6062 Implement in terms of add_prefix_map().
6063 (remap_macro_filename, remap_debug_filename): Implement in term of
6065 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
6066 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
6067 * dbxout.c: Include file-prefix-map.h.
6068 * varasm.c: Likewise.
6069 * vmsdbgout.c: Likewise.
6070 * xcoffout.c: Likewise.
6071 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
6072 * doc/cppopts.texi (-fmacro-prefix-map): Document.
6073 * doc/invoke.texi (-ffile-prefix-map): Document.
6074 (-fdebug-prefix-map): Update description.
6076 2018-01-18 Martin Liska <mliska@suse.cz>
6078 * config/i386/i386.c (indirect_thunk_name): Document that also
6080 (output_indirect_thunk): Document why both instructions
6081 (pause and lfence) are generated.
6083 2018-01-18 Richard Biener <rguenther@suse.de>
6085 PR tree-optimization/83887
6086 * graphite-scop-detection.c
6087 (scop_detection::get_nearest_dom_with_single_entry): Remove.
6088 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
6089 (scop_detection::merge_sese): Re-implement with a flood-fill
6090 algorithm that properly finds a SESE region if it exists.
6092 2018-01-18 Jakub Jelinek <jakub@redhat.com>
6095 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
6096 pointer_diff optimizations use view_convert instead of convert.
6098 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6100 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
6101 Generate different code for -mno-speculate-indirect-jumps.
6102 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
6103 (*call_indirect_aix<mode>): Disable for
6104 -mno-speculate-indirect-jumps.
6105 (*call_indirect_aix<mode>_nospec): New define_insn.
6106 (*call_value_indirect_aix<mode>): Disable for
6107 -mno-speculate-indirect-jumps.
6108 (*call_value_indirect_aix<mode>_nospec): New define_insn.
6109 (*sibcall_nonlocal_sysv<mode>): Generate different code for
6110 -mno-speculate-indirect-jumps.
6111 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
6113 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6115 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
6116 long double type, set the flags for noting the default long double
6117 type, even if we don't pass or return a long double type.
6119 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6122 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6125 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6127 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6128 support for merge[hl].
6129 (fold_mergehl_helper): New helper function.
6130 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6131 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6132 (altivec_vmrglw_direct): Add xxmrglw insn.
6134 2018-01-17 Andrew Waterman <andrew@sifive.com>
6136 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6137 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6139 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6142 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6143 call the lto_location_cache before reading the
6144 DECL_SOURCE_LOCATION of the types.
6146 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6147 Richard Sandiford <richard.sandiford@linaro.org>
6149 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6150 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6151 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6152 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6153 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6155 * config/aarch64/constraints.md (aarch64_movti_operand):
6157 * config/aarch64/predicates.md (Uti): Add new constraint.
6159 2018-01-17 Carl Love <cel@us.ibm.com>
6161 * config/rs6000/vsx.md (define_expand xl_len_r,
6162 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6163 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6165 (define_expand, define_insn): Move the shift left from the
6166 define_insn to the define_expand for lxvl and stxvl instructions.
6167 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6168 and XL_LEN_R definitions to PURE.
6170 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6172 * config/i386/i386.c (indirect_thunk_name): Declare regno
6173 as unsigned int. Compare regno with INVALID_REGNUM.
6174 (output_indirect_thunk): Ditto.
6175 (output_indirect_thunk_function): Ditto.
6176 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6177 in the call to output_indirect_thunk_function.
6179 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6182 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6183 rather than the size of inner_type to determine the stack slot size
6184 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6186 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6189 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6192 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6194 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6195 endian Linux systems to optionally enable multilibs for selecting
6196 the long double type if the user configured an explicit type.
6197 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6198 have no long double multilibs if not defined.
6199 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6200 warn if the user used -mabi={ieee,ibm}longdouble and we built
6201 multilibs for long double.
6202 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6203 appropriate multilib option.
6204 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6206 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6207 for building long double multilibs.
6208 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6210 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6212 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6215 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6217 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6220 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6223 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6226 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6228 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6229 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6231 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6233 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6234 different rtl trees depending on TARGET_64BIT.
6235 (rs6000_gen_lvx): Likewise.
6237 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6239 * config/visium/visium.md (nop): Tweak comment.
6240 (hazard_nop): Likewise.
6242 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6244 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6245 -mspeculate-indirect-jumps.
6246 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6247 for -mno-speculate-indirect-jumps.
6248 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6249 (*call_value_indirect_elfv2<mode>): Disable for
6250 -mno-speculate-indirect-jumps.
6251 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6252 (indirect_jump): Emit different RTL for
6253 -mno-speculate-indirect-jumps.
6254 (*indirect_jump<mode>): Disable for
6255 -mno-speculate-indirect-jumps.
6256 (*indirect_jump<mode>_nospec): New define_insn.
6257 (tablejump): Emit different RTL for
6258 -mno-speculate-indirect-jumps.
6259 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6260 (tablejumpsi_nospec): New define_expand.
6261 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6262 (tablejumpdi_nospec): New define_expand.
6263 (*tablejump<mode>_internal1): Disable for
6264 -mno-speculate-indirect-jumps.
6265 (*tablejump<mode>_internal1_nospec): New define_insn.
6266 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6269 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6271 * caller-save.c (insert_save): Drop unnecessary parameter. All
6274 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6275 Richard Biener <rguenth@suse.de>
6278 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6279 return early, inline manually is_gimple_sizepos. Make sure if we
6280 call gimplify_expr we don't end up with a gimple constant.
6281 * tree.c (variably_modified_type_p): Don't return true for
6282 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6283 * gimplify.h (is_gimple_sizepos): Remove.
6285 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6287 PR tree-optimization/83857
6288 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6289 vectorizable_live_operation for pure SLP statements.
6290 (vectorizable_live_operation): Handle PHIs.
6292 2018-01-16 Richard Biener <rguenther@suse.de>
6294 PR tree-optimization/83867
6295 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6296 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6298 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6301 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6302 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6303 If off is not INTEGER_CST, issue a may not be aligned warning
6304 rather than isn't aligned. Use isn%'t rather than isn't.
6305 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6307 <case MULT_EXPR>: Improve the case when bottom and one of the
6308 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6309 operand, in that case check if the other operand is multiple of
6310 bottom divided by the INTEGER_CST operand.
6312 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6315 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6316 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6317 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6318 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6319 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6320 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6321 (pa_function_arg_size): New function.
6323 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6325 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6326 in a separate statement.
6328 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6330 PR tree-optimization/83847
6331 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6332 group gathers and scatters.
6334 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6336 PR rtl-optimization/86620
6337 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6339 PR rtl-optimization/83213
6340 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6341 to last if both are JUMP_INSNs.
6343 PR tree-optimization/83843
6344 * gimple-ssa-store-merging.c
6345 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6346 store_immediate_info for bswap/nop orig_stores.
6348 2018-01-15 Andrew Waterman <andrew@sifive.com>
6350 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6352 <UDIV>: Increase cost if !TARGET_DIV.
6354 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6356 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6357 (define_attr "cr_logical_3op"): New.
6358 (cceq_ior_compare): Adjust.
6359 (cceq_ior_compare_complement): Adjust.
6360 (*cceq_rev_compare): Adjust.
6361 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6362 (is_cracked_insn): Adjust.
6363 (insn_must_be_first_in_group): Adjust.
6364 * config/rs6000/40x.md: Adjust.
6365 * config/rs6000/440.md: Adjust.
6366 * config/rs6000/476.md: Adjust.
6367 * config/rs6000/601.md: Adjust.
6368 * config/rs6000/603.md: Adjust.
6369 * config/rs6000/6xx.md: Adjust.
6370 * config/rs6000/7450.md: Adjust.
6371 * config/rs6000/7xx.md: Adjust.
6372 * config/rs6000/8540.md: Adjust.
6373 * config/rs6000/cell.md: Adjust.
6374 * config/rs6000/e300c2c3.md: Adjust.
6375 * config/rs6000/e500mc.md: Adjust.
6376 * config/rs6000/e500mc64.md: Adjust.
6377 * config/rs6000/e5500.md: Adjust.
6378 * config/rs6000/e6500.md: Adjust.
6379 * config/rs6000/mpc.md: Adjust.
6380 * config/rs6000/power4.md: Adjust.
6381 * config/rs6000/power5.md: Adjust.
6382 * config/rs6000/power6.md: Adjust.
6383 * config/rs6000/power7.md: Adjust.
6384 * config/rs6000/power8.md: Adjust.
6385 * config/rs6000/power9.md: Adjust.
6386 * config/rs6000/rs64.md: Adjust.
6387 * config/rs6000/titan.md: Adjust.
6389 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6391 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6392 ix86_indirect_branch_register logic.
6394 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6396 * config/i386/constraints.md (Bs): Update
6397 ix86_indirect_branch_register check. Don't check
6398 ix86_indirect_branch_register with GOT_memory_operand.
6400 * config/i386/predicates.md (GOT_memory_operand): Don't check
6401 ix86_indirect_branch_register here.
6402 (GOT32_symbol_operand): Likewise.
6404 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6406 * config/i386/predicates.md (constant_call_address_operand):
6407 Rewrite ix86_indirect_branch_register logic.
6408 (sibcall_insn_operand): Likewise.
6410 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6412 * config/i386/constraints.md (Bs): Replace
6413 ix86_indirect_branch_thunk_register with
6414 ix86_indirect_branch_register.
6416 * config/i386/i386.md (indirect_jump): Likewise.
6417 (tablejump): Likewise.
6418 (*sibcall_memory): Likewise.
6419 (*sibcall_value_memory): Likewise.
6420 Peepholes of indirect call and jump via memory: Likewise.
6421 * config/i386/i386.opt: Likewise.
6422 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6423 (GOT_memory_operand): Likewise.
6424 (call_insn_operand): Likewise.
6425 (sibcall_insn_operand): Likewise.
6426 (GOT32_symbol_operand): Likewise.
6428 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6431 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6432 type rather than type addr's type points to.
6433 (expand_omp_atomic_mutex): Likewise.
6434 (expand_omp_atomic): Likewise.
6436 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6439 * config/i386/i386.c (output_indirect_thunk_function): Use
6440 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6441 for __x86_return_thunk.
6443 2018-01-15 Richard Biener <rguenther@suse.de>
6446 * expmed.c (extract_bit_field_1): Fix typo.
6448 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6451 * config/arm/iterators.md (VF): New mode iterator.
6452 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6453 Remove integer-related logic from pattern.
6454 (neon_vabd<mode>_3): Likewise.
6456 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6459 * common.opt (fstrict-overflow): No longer an alias.
6460 (fwrapv-pointer): New option.
6461 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6462 also for pointer types based on flag_wrapv_pointer.
6463 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6464 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6465 opts->x_flag_wrapv got set.
6466 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6467 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6468 POINTER_TYPE_OVERFLOW_UNDEFINED.
6469 * match.pd: Likewise in address comparison pattern.
6470 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6472 2018-01-15 Richard Biener <rguenther@suse.de>
6475 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6476 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6477 Reset type names to their identifier if their TYPE_DECL doesn't
6478 have linkage (and thus is used for ODR and devirt).
6479 (save_debug_info_for_decl): Remove.
6480 (save_debug_info_for_type): Likewise.
6481 (add_tree_to_fld_list): Adjust.
6482 * tree-pretty-print.c (dump_generic_node): Make dumping of
6483 type names more robust.
6485 2018-01-15 Richard Biener <rguenther@suse.de>
6487 * BASE-VER: Bump to 8.0.1.
6489 2018-01-14 Martin Sebor <msebor@redhat.com>
6492 * builtins.c (check_access): Avoid warning when the no-warning bit
6495 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6497 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6498 * ira-color (allocno_hard_regs_compare): Likewise.
6500 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6503 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6504 Use .pushsection/.popsection.
6506 2018-01-14 Martin Sebor <msebor@redhat.com>
6509 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6511 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6513 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6514 entry from extra_headers.
6515 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6516 extra_headers, make the list bitwise identical to the i?86-*-* one.
6518 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6520 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6521 -mcmodel=large with -mindirect-branch=thunk,
6522 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6523 -mfunction-return=thunk-extern.
6524 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6525 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6526 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6528 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6530 * config/i386/i386.c (print_reg): Print the name of the full
6531 integer register without '%'.
6532 (ix86_print_operand): Handle 'V'.
6533 * doc/extend.texi: Document 'V' modifier.
6535 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6537 * config/i386/constraints.md (Bs): Disallow memory operand for
6538 -mindirect-branch-register.
6540 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6541 (GOT_memory_operand): Likewise.
6542 (call_insn_operand): Likewise.
6543 (sibcall_insn_operand): Likewise.
6544 (GOT32_symbol_operand): Likewise.
6545 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6546 for -mindirect-branch-register.
6547 (tablejump): Likewise.
6548 (*sibcall_memory): Likewise.
6549 (*sibcall_value_memory): Likewise.
6550 Disallow peepholes of indirect call and jump via memory for
6551 -mindirect-branch-register.
6552 (*call_pop): Replace m with Bw.
6553 (*call_value_pop): Likewise.
6554 (*sibcall_pop_memory): Replace m with Bs.
6555 * config/i386/i386.opt (mindirect-branch-register): New option.
6556 * doc/invoke.texi: Document -mindirect-branch-register option.
6558 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6560 * config/i386/i386-protos.h (ix86_output_function_return): New.
6561 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6562 set function_return_type.
6563 (indirect_thunk_name): Add ret_p to indicate thunk for function
6565 (output_indirect_thunk_function): Pass false to
6566 indirect_thunk_name.
6567 (ix86_output_indirect_branch_via_reg): Likewise.
6568 (ix86_output_indirect_branch_via_push): Likewise.
6569 (output_indirect_thunk_function): Create alias for function
6570 return thunk if regno < 0.
6571 (ix86_output_function_return): New function.
6572 (ix86_handle_fndecl_attribute): Handle function_return.
6573 (ix86_attribute_table): Add function_return.
6574 * config/i386/i386.h (machine_function): Add
6575 function_return_type.
6576 * config/i386/i386.md (simple_return_internal): Use
6577 ix86_output_function_return.
6578 (simple_return_internal_long): Likewise.
6579 * config/i386/i386.opt (mfunction-return=): New option.
6580 (indirect_branch): Mention -mfunction-return=.
6581 * doc/extend.texi: Document function_return function attribute.
6582 * doc/invoke.texi: Document -mfunction-return= option.
6584 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6586 * config/i386/i386-opts.h (indirect_branch): New.
6587 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6588 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6589 with local indirect jump when converting indirect call and jump.
6590 (ix86_set_indirect_branch_type): New.
6591 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6592 (indirectlabelno): New.
6593 (indirect_thunk_needed): Likewise.
6594 (indirect_thunk_bnd_needed): Likewise.
6595 (indirect_thunks_used): Likewise.
6596 (indirect_thunks_bnd_used): Likewise.
6597 (INDIRECT_LABEL): Likewise.
6598 (indirect_thunk_name): Likewise.
6599 (output_indirect_thunk): Likewise.
6600 (output_indirect_thunk_function): Likewise.
6601 (ix86_output_indirect_branch_via_reg): Likewise.
6602 (ix86_output_indirect_branch_via_push): Likewise.
6603 (ix86_output_indirect_branch): Likewise.
6604 (ix86_output_indirect_jmp): Likewise.
6605 (ix86_code_end): Call output_indirect_thunk_function if needed.
6606 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6608 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6609 (ix86_attribute_table): Add indirect_branch.
6610 * config/i386/i386.h (machine_function): Add indirect_branch_type
6611 and has_local_indirect_jump.
6612 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6614 (tablejump): Likewise.
6615 (*indirect_jump): Use ix86_output_indirect_jmp.
6616 (*tablejump_1): Likewise.
6617 (simple_return_indirect_internal): Likewise.
6618 * config/i386/i386.opt (mindirect-branch=): New option.
6619 (indirect_branch): New.
6622 (thunk-inline): Likewise.
6623 (thunk-extern): Likewise.
6624 * doc/extend.texi: Document indirect_branch function attribute.
6625 * doc/invoke.texi: Document -mindirect-branch= option.
6627 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6630 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6632 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6634 * ipa-inline.c (want_inline_small_function_p): Return false if
6635 inlining has already failed with CIF_FINAL_ERROR.
6636 (update_caller_keys): Call want_inline_small_function_p before
6638 (update_callee_keys): Likewise.
6640 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6642 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6644 (rs6000_quadword_masked_address_p): Likewise.
6645 (quad_aligned_load_p): Likewise.
6646 (quad_aligned_store_p): Likewise.
6647 (const_load_sequence_p): Add comment to describe the outer-most loop.
6648 (mimic_memory_attributes_and_flags): New function.
6649 (rs6000_gen_stvx): Likewise.
6650 (replace_swapped_aligned_store): Likewise.
6651 (rs6000_gen_lvx): Likewise.
6652 (replace_swapped_aligned_load): Likewise.
6653 (replace_swapped_load_constant): Capitalize argument name in
6654 comment describing this function.
6655 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6656 and stores that access quad-word aligned addresses and replace
6657 with stvx or lvx instructions when appropriate.
6658 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6659 New function prototype.
6660 (rs6000_quadword_masked_address_p): Likewise.
6661 (rs6000_gen_lvx): Likewise.
6662 (rs6000_gen_stvx): Likewise.
6663 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6664 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6665 when memory address is aligned.
6666 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6667 this split to select lvx instruction when memory address is aligned.
6668 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6669 instruction when memory address is aligned.
6670 (*vsx_le_perm_load_v16qi): Likewise.
6671 (four unnamed splitters): Modify to select the stvx instruction
6672 when memory is aligned.
6674 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6676 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6677 which appears in the queue multiple times.
6679 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6680 Alan Hayward <alan.hayward@arm.com>
6681 David Sherwood <david.sherwood@arm.com>
6683 * tree-vectorizer.h (vec_lower_bound): New structure.
6684 (_loop_vec_info): Add check_nonzero and lower_bounds.
6685 (LOOP_VINFO_CHECK_NONZERO): New macro.
6686 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6687 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6688 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6689 fields. Make seg_len the distance travelled, not including the
6691 (dr_direction_indicator): Declare.
6692 (dr_zero_step_indicator): Likewise.
6693 (dr_known_forward_stride_p): Likewise.
6694 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6696 (runtime_alias_check_p): Allow runtime alias checks with
6698 (operator ==): Compare access_size and align.
6699 (prune_runtime_alias_test_list): Rework for new distinction between
6700 the access_size and seg_len.
6701 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6703 (get_segment_min_max): New function.
6704 (create_intersect_range_checks): Use it.
6705 (dr_step_indicator): New function.
6706 (dr_direction_indicator): Likewise.
6707 (dr_zero_step_indicator): Likewise.
6708 (dr_known_forward_stride_p): Likewise.
6709 * tree-loop-distribution.c (data_ref_segment_size): Return
6710 DR_STEP * (niters - 1).
6711 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6713 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6714 (vect_preserves_scalar_order_p): New function, split out from...
6715 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6716 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6717 (vect_vfa_access_size): New function.
6718 (vect_vfa_align): Likewise.
6719 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6720 (dump_lower_bound): New function.
6721 (vect_check_lower_bound): Likewise.
6722 (vect_small_gap_p): Likewise.
6723 (vectorizable_with_step_bound_p): Likewise.
6724 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6725 depencies if the vectorization factor is 1. Convert the checks
6726 for nonzero steps into checks on the bounds of DR_STEP. Try using
6727 a bunds check for variable steps if the minimum required step is
6728 relatively small. Update calls to the dr_with_seg_len
6729 constructor and to vect_compile_time_alias.
6730 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6732 (vect_loop_versioning): Call it.
6733 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6735 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6737 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6738 Alan Hayward <alan.hayward@arm.com>
6739 David Sherwood <david.sherwood@arm.com>
6741 * doc/sourcebuild.texi (vect_scatter_store): Document.
6742 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6744 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6746 * genopinit.c (main): Add supports_vec_scatter_store and
6747 supports_vec_scatter_store_cached to target_optabs.
6748 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6749 IFN_MASK_SCATTER_STORE.
6750 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6752 * internal-fn.h (internal_store_fn_p): Declare.
6753 (internal_fn_stored_value_index): Likewise.
6754 * internal-fn.c (scatter_store_direct): New macro.
6755 (expand_scatter_store_optab_fn): New function.
6756 (direct_scatter_store_optab_supported_p): New macro.
6757 (internal_store_fn_p): New function.
6758 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6759 IFN_MASK_SCATTER_STORE.
6760 (internal_fn_mask_index): Likewise.
6761 (internal_fn_stored_value_index): New function.
6762 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6764 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6765 * optabs-query.c (supports_vec_scatter_store_p): New function.
6766 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6767 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6768 true for scatter stores.
6769 (vect_gather_scatter_fn_p): Handle scatter stores too.
6770 (vect_check_gather_scatter): Consider using scatter stores if
6771 supports_vec_scatter_store_p.
6772 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6774 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6775 internal_fn_stored_value_index.
6776 (check_load_store_masking): Handle scatter stores too.
6777 (vect_get_store_rhs): Make public.
6778 (vectorizable_call): Use internal_store_fn_p.
6779 (vectorizable_store): Handle scatter store internal functions.
6780 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6781 when deciding whether the end of the group has been reached.
6782 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6783 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6784 (mask_scatter_store<mode>): New insns.
6786 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6787 Alan Hayward <alan.hayward@arm.com>
6788 David Sherwood <david.sherwood@arm.com>
6790 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6791 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6792 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6794 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6795 Use vect_truncate_gather_scatter_offset if we can't treat the
6796 operation as a normal gather load or scatter store.
6797 (get_group_load_store_type): Take the gather_scatter_info
6798 as argument. Try using a gather load or scatter store for
6799 single-element groups.
6800 (get_load_store_type): Update calls to get_group_load_store_type
6801 and vect_use_strided_gather_scatters_p.
6803 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6804 Alan Hayward <alan.hayward@arm.com>
6805 David Sherwood <david.sherwood@arm.com>
6807 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6808 optional tree argument.
6809 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6811 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6812 but continue to use the current value as a fallback.
6813 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6814 to compare the updates.
6815 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6816 (get_load_store_type): Use it when handling a strided access.
6817 (vect_get_strided_load_store_ops): New function.
6818 (vect_get_data_ptr_increment): Likewise.
6819 (vectorizable_load): Handle strided gather loads. Always pass
6820 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6822 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6823 Alan Hayward <alan.hayward@arm.com>
6824 David Sherwood <david.sherwood@arm.com>
6826 * doc/md.texi (gather_load@var{m}): Document.
6827 (mask_gather_load@var{m}): Likewise.
6828 * genopinit.c (main): Add supports_vec_gather_load and
6829 supports_vec_gather_load_cached to target_optabs.
6830 * optabs-tree.c (init_tree_optimization_optabs): Use
6831 ggc_cleared_alloc to allocate target_optabs.
6832 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6833 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6835 * internal-fn.h (internal_load_fn_p): Declare.
6836 (internal_gather_scatter_fn_p): Likewise.
6837 (internal_fn_mask_index): Likewise.
6838 (internal_gather_scatter_fn_supported_p): Likewise.
6839 * internal-fn.c (gather_load_direct): New macro.
6840 (expand_gather_load_optab_fn): New function.
6841 (direct_gather_load_optab_supported_p): New macro.
6842 (direct_internal_fn_optab): New function.
6843 (internal_load_fn_p): Likewise.
6844 (internal_gather_scatter_fn_p): Likewise.
6845 (internal_fn_mask_index): Likewise.
6846 (internal_gather_scatter_fn_supported_p): Likewise.
6847 * optabs-query.c (supports_at_least_one_mode_p): New function.
6848 (supports_vec_gather_load_p): Likewise.
6849 * optabs-query.h (supports_vec_gather_load_p): Declare.
6850 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6851 and memory_type field.
6852 (NUM_PATTERNS): Bump to 15.
6853 * tree-vect-data-refs.c: Include internal-fn.h.
6854 (vect_gather_scatter_fn_p): New function.
6855 (vect_describe_gather_scatter_call): Likewise.
6856 (vect_check_gather_scatter): Try using internal functions for
6857 gather loads. Recognize existing calls to a gather load function.
6858 (vect_analyze_data_refs): Consider using gather loads if
6859 supports_vec_gather_load_p.
6860 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6861 (vect_get_gather_scatter_offset_type): Likewise.
6862 (vect_convert_mask_for_vectype): Likewise.
6863 (vect_add_conversion_to_patterm): Likewise.
6864 (vect_try_gather_scatter_pattern): Likewise.
6865 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6866 (vect_vect_recog_func_ptrs): Add it.
6867 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6868 internal_fn_mask_index and internal_gather_scatter_fn_p.
6869 (check_load_store_masking): Take the gather_scatter_info as an
6870 argument and handle gather loads.
6871 (vect_get_gather_scatter_ops): New function.
6872 (vectorizable_call): Check internal_load_fn_p.
6873 (vectorizable_load): Likewise. Handle gather load internal
6875 (vectorizable_store): Update call to check_load_store_masking.
6876 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6877 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6878 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6879 (aarch64_gather_scale_operand_d): New predicates.
6880 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6881 (mask_gather_load<mode>): New insns.
6883 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6884 Alan Hayward <alan.hayward@arm.com>
6885 David Sherwood <david.sherwood@arm.com>
6887 * optabs.def (fold_left_plus_optab): New optab.
6888 * doc/md.texi (fold_left_plus_@var{m}): Document.
6889 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6890 * internal-fn.c (fold_left_direct): Define.
6891 (expand_fold_left_optab_fn): Likewise.
6892 (direct_fold_left_optab_supported_p): Likewise.
6893 * fold-const-call.c (fold_const_fold_left): New function.
6894 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6895 * tree-parloops.c (valid_reduction_p): New function.
6896 (gather_scalar_reductions): Use it.
6897 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6898 (vect_finish_replace_stmt): Declare.
6899 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6900 (needs_fold_left_reduction_p): New function, split out from...
6901 (vect_is_simple_reduction): ...here. Accept reductions that
6902 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6903 (vect_force_simple_reduction): Also store the reduction type in
6904 the assignment's STMT_VINFO_REDUC_TYPE.
6905 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6906 (merge_with_identity): New function.
6907 (vect_expand_fold_left): Likewise.
6908 (vectorize_fold_left_reduction): Likewise.
6909 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6910 scalar phi in place for it. Check for target support and reject
6911 cases that would reassociate the operation. Defer the transform
6912 phase to vectorize_fold_left_reduction.
6913 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6914 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6915 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6919 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6920 call to ifc_temp_var.
6922 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6923 Alan Hayward <alan.hayward@arm.com>
6924 David Sherwood <david.sherwood@arm.com>
6926 * target.def (legitimize_address_displacement): Take the original
6927 offset as a poly_int.
6928 * targhooks.h (default_legitimize_address_displacement): Update
6930 * targhooks.c (default_legitimize_address_displacement): Likewise.
6931 * doc/tm.texi: Regenerate.
6932 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6933 as an argument, moving assert of ad->disp == ad->disp_term to...
6934 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6935 Try calling targetm.legitimize_address_displacement before expanding
6936 the address rather than afterwards, and adjust for the new interface.
6937 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6938 Match the new hook interface. Handle SVE addresses.
6939 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6942 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6944 * Makefile.in (OBJS): Add early-remat.o.
6945 * target.def (select_early_remat_modes): New hook.
6946 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6947 * doc/tm.texi: Regenerate.
6948 * targhooks.h (default_select_early_remat_modes): Declare.
6949 * targhooks.c (default_select_early_remat_modes): New function.
6950 * timevar.def (TV_EARLY_REMAT): New timevar.
6951 * passes.def (pass_early_remat): New pass.
6952 * tree-pass.h (make_pass_early_remat): Declare.
6953 * early-remat.c: New file.
6954 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6956 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6958 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6959 Alan Hayward <alan.hayward@arm.com>
6960 David Sherwood <david.sherwood@arm.com>
6962 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6963 vfm1 with a bound_epilog parameter.
6964 (vect_do_peeling): Update calls accordingly, and move the prologue
6965 call earlier in the function. Treat the base bound_epilog as 0 for
6966 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6967 this base when peeling for gaps.
6968 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6969 with fully-masked loops.
6970 (vect_estimate_min_profitable_iters): Handle the single peeled
6971 iteration in that case.
6973 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6974 Alan Hayward <alan.hayward@arm.com>
6975 David Sherwood <david.sherwood@arm.com>
6977 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6978 single-element interleaving even if the size is not a power of 2.
6979 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6980 accesses for single-element interleaving if the group size is
6983 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6984 Alan Hayward <alan.hayward@arm.com>
6985 David Sherwood <david.sherwood@arm.com>
6987 * doc/md.texi (fold_extract_last_@var{m}): Document.
6988 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6989 * optabs.def (fold_extract_last_optab): New optab.
6990 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6991 * internal-fn.c (fold_extract_direct): New macro.
6992 (expand_fold_extract_optab_fn): Likewise.
6993 (direct_fold_extract_optab_supported_p): Likewise.
6994 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6995 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6996 EXTRACT_LAST_REDUCTION.
6997 (get_initial_def_for_reduction): Do not create an initial vector
6998 for EXTRACT_LAST_REDUCTION reductions.
6999 (vectorizable_reduction): Leave the scalar phi in place for
7000 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
7001 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
7002 epilogue code for EXTRACT_LAST_REDUCTION and defer the
7003 transform phase to vectorizable_condition.
7004 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
7006 (vect_finish_stmt_generation): ...here.
7007 (vect_finish_replace_stmt): New function.
7008 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
7009 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
7011 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
7013 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7014 Alan Hayward <alan.hayward@arm.com>
7015 David Sherwood <david.sherwood@arm.com>
7017 * doc/md.texi (extract_last_@var{m}): Document.
7018 * optabs.def (extract_last_optab): New optab.
7019 * internal-fn.def (EXTRACT_LAST): New internal function.
7020 * internal-fn.c (cond_unary_direct): New macro.
7021 (expand_cond_unary_optab_fn): Likewise.
7022 (direct_cond_unary_optab_supported_p): Likewise.
7023 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
7024 loops using EXTRACT_LAST.
7025 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
7026 (extract_last_<mode>): ...this optab.
7027 (vec_extract<mode><Vel>): Update accordingly.
7029 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7030 Alan Hayward <alan.hayward@arm.com>
7031 David Sherwood <david.sherwood@arm.com>
7033 * target.def (empty_mask_is_expensive): New hook.
7034 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
7035 * doc/tm.texi: Regenerate.
7036 * targhooks.h (default_empty_mask_is_expensive): Declare.
7037 * targhooks.c (default_empty_mask_is_expensive): New function.
7038 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
7039 if the target says that empty masks are expensive.
7040 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
7042 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
7044 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7045 Alan Hayward <alan.hayward@arm.com>
7046 David Sherwood <david.sherwood@arm.com>
7048 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
7049 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
7050 (vect_use_loop_mask_for_alignment_p): New function.
7051 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
7052 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
7053 niters_skip argument. Make sure that the first niters_skip elements
7054 of the first iteration are inactive.
7055 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
7056 Update call to vect_set_loop_masks_directly.
7057 (get_misalign_in_elems): New function, split out from...
7058 (vect_gen_prolog_loop_niters): ...here.
7059 (vect_update_init_of_dr): Take a code argument that specifies whether
7060 the adjustment should be added or subtracted.
7061 (vect_update_init_of_drs): Likewise.
7062 (vect_prepare_for_masked_peels): New function.
7063 (vect_do_peeling): Skip prologue peeling if we're using a mask
7064 instead. Update call to vect_update_inits_of_drs.
7065 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7067 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
7068 alignment. Do not include the number of peeled iterations in
7069 the minimum threshold in that case.
7070 (vectorizable_induction): Adjust the start value down by
7071 LOOP_VINFO_MASK_SKIP_NITERS iterations.
7072 (vect_transform_loop): Call vect_prepare_for_masked_peels.
7073 Take the number of skipped iterations into account when calculating
7075 * tree-vect-stmts.c (vect_gen_while_not): New function.
7077 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7078 Alan Hayward <alan.hayward@arm.com>
7079 David Sherwood <david.sherwood@arm.com>
7081 * doc/sourcebuild.texi (vect_fully_masked): Document.
7082 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
7084 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
7086 (vect_analyze_loop_2): ...here. Don't check the vectorization
7087 factor against the number of loop iterations if the loop is
7090 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7091 Alan Hayward <alan.hayward@arm.com>
7092 David Sherwood <david.sherwood@arm.com>
7094 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
7095 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
7096 (dump_groups): Update accordingly.
7097 (iv_use::mem_type): New member variable.
7098 (address_p): New function.
7099 (record_use): Add a mem_type argument and initialize the new
7101 (record_group_use): Add a mem_type argument. Use address_p.
7102 Remove obsolete null checks of base_object. Update call to record_use.
7103 (find_interesting_uses_op): Update call to record_group_use.
7104 (find_interesting_uses_cond): Likewise.
7105 (find_interesting_uses_address): Likewise.
7106 (get_mem_type_for_internal_fn): New function.
7107 (find_address_like_use): Likewise.
7108 (find_interesting_uses_stmt): Try find_address_like_use before
7109 calling find_interesting_uses_op.
7110 (addr_offset_valid_p): Use the iv mem_type field as the type
7111 of the addressed memory.
7112 (add_autoinc_candidates): Likewise.
7113 (get_address_cost): Likewise.
7114 (split_small_address_groups_p): Use address_p.
7115 (split_address_groups): Likewise.
7116 (add_iv_candidate_for_use): Likewise.
7117 (autoinc_possible_for_pair): Likewise.
7118 (rewrite_groups): Likewise.
7119 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
7120 (determine_group_iv_cost): Update after split of USE_ADDRESS.
7121 (get_alias_ptr_type_for_ptr_address): New function.
7122 (rewrite_use_address): Rewrite address uses in calls that were
7123 identified by find_address_like_use.
7125 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7126 Alan Hayward <alan.hayward@arm.com>
7127 David Sherwood <david.sherwood@arm.com>
7129 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7131 * gimple-expr.h (is_gimple_addressable: Likewise.
7132 * gimple-expr.c (is_gimple_address): Likewise.
7133 * internal-fn.c (expand_call_mem_ref): New function.
7134 (expand_mask_load_optab_fn): Use it.
7135 (expand_mask_store_optab_fn): Likewise.
7137 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7138 Alan Hayward <alan.hayward@arm.com>
7139 David Sherwood <david.sherwood@arm.com>
7141 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7142 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7143 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7144 (cond_umax@var{mode}): Document.
7145 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7146 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7147 (cond_umin_optab, cond_umax_optab): New optabs.
7148 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7149 (COND_IOR, COND_XOR): New internal functions.
7150 * internal-fn.h (get_conditional_internal_fn): Declare.
7151 * internal-fn.c (cond_binary_direct): New macro.
7152 (expand_cond_binary_optab_fn): Likewise.
7153 (direct_cond_binary_optab_supported_p): Likewise.
7154 (get_conditional_internal_fn): New function.
7155 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7156 Cope with reduction statements that are vectorized as calls rather
7158 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7159 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7160 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7161 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7162 (UNSPEC_COND_EOR): New unspecs.
7163 (optab): Add mappings for them.
7164 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7165 (sve_int_op, sve_fp_op): New int attributes.
7167 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7168 Alan Hayward <alan.hayward@arm.com>
7169 David Sherwood <david.sherwood@arm.com>
7171 * optabs.def (while_ult_optab): New optab.
7172 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7173 * internal-fn.def (WHILE_ULT): New internal function.
7174 * internal-fn.h (direct_internal_fn_supported_p): New override
7175 that takes two types as argument.
7176 * internal-fn.c (while_direct): New macro.
7177 (expand_while_optab_fn): New function.
7178 (convert_optab_supported_p): Likewise.
7179 (direct_while_optab_supported_p): New macro.
7180 * wide-int.h (wi::udiv_ceil): New function.
7181 * tree-vectorizer.h (rgroup_masks): New structure.
7182 (vec_loop_masks): New typedef.
7183 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7185 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7186 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7187 (vect_max_vf): New function.
7188 (slpeel_make_loop_iterate_ntimes): Delete.
7189 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7190 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7191 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7192 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7193 internal-fn.h, stor-layout.h and optabs-query.h.
7194 (vect_set_loop_mask): New function.
7195 (add_preheader_seq): Likewise.
7196 (add_header_seq): Likewise.
7197 (interleave_supported_p): Likewise.
7198 (vect_maybe_permute_loop_masks): Likewise.
7199 (vect_set_loop_masks_directly): Likewise.
7200 (vect_set_loop_condition_masked): Likewise.
7201 (vect_set_loop_condition_unmasked): New function, split out from
7202 slpeel_make_loop_iterate_ntimes.
7203 (slpeel_make_loop_iterate_ntimes): Rename to..
7204 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7205 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7206 (vect_do_peeling): Update call accordingly.
7207 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7209 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7210 mask_compare_type, can_fully_mask_p and fully_masked_p.
7211 (release_vec_loop_masks): New function.
7212 (_loop_vec_info): Use it to free the loop masks.
7213 (can_produce_all_loop_masks_p): New function.
7214 (vect_get_max_nscalars_per_iter): Likewise.
7215 (vect_verify_full_masking): Likewise.
7216 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7217 retries, and free the mask rgroups before retrying. Check loop-wide
7218 reasons for disallowing fully-masked loops. Make the final decision
7219 about whether use a fully-masked loop or not.
7220 (vect_estimate_min_profitable_iters): Do not assume that peeling
7221 for the number of iterations will be needed for fully-masked loops.
7222 (vectorizable_reduction): Disable fully-masked loops.
7223 (vectorizable_live_operation): Likewise.
7224 (vect_halve_mask_nunits): New function.
7225 (vect_double_mask_nunits): Likewise.
7226 (vect_record_loop_mask): Likewise.
7227 (vect_get_loop_mask): Likewise.
7228 (vect_transform_loop): Handle the case in which the final loop
7229 iteration might handle a partial vector. Call vect_set_loop_condition
7230 instead of slpeel_make_loop_iterate_ntimes.
7231 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7232 (check_load_store_masking): New function.
7233 (prepare_load_store_mask): Likewise.
7234 (vectorizable_store): Handle fully-masked loops.
7235 (vectorizable_load): Likewise.
7236 (supportable_widening_operation): Use vect_halve_mask_nunits for
7238 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7239 (vect_gen_while): New function.
7240 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7241 (aarch64_uqdec<mode>): New insn.
7243 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7244 Alan Hayward <alan.hayward@arm.com>
7245 David Sherwood <david.sherwood@arm.com>
7247 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7248 (reduc_xor_scal_optab): New optabs.
7249 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7250 (reduc_xor_scal_@var{m}): Document.
7251 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7252 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7254 * fold-const-call.c (fold_const_call): Handle them.
7255 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7256 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7257 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7258 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7259 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7260 (UNSPEC_XORV): New unspecs.
7261 (optab): Add entries for them.
7262 (BITWISEV): New int iterator.
7263 (bit_reduc_op): New int attributes.
7265 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7266 Alan Hayward <alan.hayward@arm.com>
7267 David Sherwood <david.sherwood@arm.com>
7269 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7270 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7271 * optabs.def (vec_shl_insert_optab): New optab.
7272 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7273 (duplicate_and_interleave): Likewise.
7274 * tree-vect-loop.c: Include internal-fn.h.
7275 (neutral_op_for_slp_reduction): New function, split out from
7276 get_initial_defs_for_reduction.
7277 (get_initial_def_for_reduction): Handle option 2 for variable-length
7278 vectors by loading the neutral value into a vector and then shifting
7279 the initial value into element 0.
7280 (get_initial_defs_for_reduction): Replace the code argument with
7281 the neutral value calculated by neutral_op_for_slp_reduction.
7282 Use gimple_build_vector for constant-length vectors.
7283 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7284 but the first group_size elements have a neutral value.
7285 Use duplicate_and_interleave otherwise.
7286 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7287 Update call to get_initial_defs_for_reduction. Handle SLP
7288 reductions for variable-length vectors by creating one vector
7289 result for each scalar result, with the elements associated
7290 with other scalar results stubbed out with the neutral value.
7291 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7292 Require IFN_VEC_SHL_INSERT for double reductions on
7293 variable-length vectors, or SLP reductions that have
7294 a neutral value. Require can_duplicate_and_interleave_p
7295 support for variable-length unchained SLP reductions if there
7296 is no neutral value, such as for MIN/MAX reductions. Also require
7297 the number of vector elements to be a multiple of the number of
7298 SLP statements when doing variable-length unchained SLP reductions.
7299 Update call to vect_create_epilog_for_reduction.
7300 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7301 and remove initial values.
7302 (duplicate_and_interleave): Make public.
7303 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7304 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7307 Alan Hayward <alan.hayward@arm.com>
7308 David Sherwood <david.sherwood@arm.com>
7310 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7311 (can_duplicate_and_interleave_p): New function.
7312 (vect_get_and_check_slp_defs): Take the vector of statements
7313 rather than just the current one. Remove excess parentheses.
7314 Restriction rejectinon of vect_constant_def and vect_external_def
7315 for variable-length vectors to boolean types, or types for which
7316 can_duplicate_and_interleave_p is false.
7317 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7318 (duplicate_and_interleave): New function.
7319 (vect_get_constant_vectors): Use gimple_build_vector for
7320 constant-length vectors and suitable variable-length constant
7321 vectors. Use duplicate_and_interleave for other variable-length
7322 vectors. Don't defer the update when inserting new statements.
7324 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7325 Alan Hayward <alan.hayward@arm.com>
7326 David Sherwood <david.sherwood@arm.com>
7328 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7329 min_profitable_iters doesn't go negative.
7331 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7332 Alan Hayward <alan.hayward@arm.com>
7333 David Sherwood <david.sherwood@arm.com>
7335 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7336 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7337 * optabs.def (vec_mask_load_lanes_optab): New optab.
7338 (vec_mask_store_lanes_optab): Likewise.
7339 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7340 (MASK_STORE_LANES): Likewise.
7341 * internal-fn.c (mask_load_lanes_direct): New macro.
7342 (mask_store_lanes_direct): Likewise.
7343 (expand_mask_load_optab_fn): Handle masked operations.
7344 (expand_mask_load_lanes_optab_fn): New macro.
7345 (expand_mask_store_optab_fn): Handle masked operations.
7346 (expand_mask_store_lanes_optab_fn): New macro.
7347 (direct_mask_load_lanes_optab_supported_p): Likewise.
7348 (direct_mask_store_lanes_optab_supported_p): Likewise.
7349 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7351 (vect_load_lanes_supported): Likewise.
7352 * tree-vect-data-refs.c (strip_conversion): New function.
7353 (can_group_stmts_p): Likewise.
7354 (vect_analyze_data_ref_accesses): Use it instead of checking
7355 for a pair of assignments.
7356 (vect_store_lanes_supported): Take a masked_p parameter.
7357 (vect_load_lanes_supported): Likewise.
7358 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7359 vect_store_lanes_supported and vect_load_lanes_supported.
7360 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7361 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7362 parameter. Don't allow gaps for masked accesses.
7363 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7364 and vect_load_lanes_supported.
7365 (get_load_store_type): Take a masked_p parameter and update
7366 call to get_group_load_store_type.
7367 (vectorizable_store): Update call to get_load_store_type.
7368 Handle IFN_MASK_STORE_LANES.
7369 (vectorizable_load): Update call to get_load_store_type.
7370 Handle IFN_MASK_LOAD_LANES.
7372 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7373 Alan Hayward <alan.hayward@arm.com>
7374 David Sherwood <david.sherwood@arm.com>
7376 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7378 * config/aarch64/aarch64-protos.h
7379 (aarch64_sve_struct_memory_operand_p): Declare.
7380 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7381 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7382 (VPRED, vpred): Handle SVE structure modes.
7383 * config/aarch64/constraints.md (Utx): New constraint.
7384 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7385 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7386 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7387 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7388 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7389 structure modes. Split into pieces after RA.
7390 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7391 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7393 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7394 SVE structure modes.
7395 (aarch64_classify_address): Likewise.
7396 (sizetochar): Move earlier in file.
7397 (aarch64_print_operand): Handle SVE register lists.
7398 (aarch64_array_mode): New function.
7399 (aarch64_sve_struct_memory_operand_p): Likewise.
7400 (TARGET_ARRAY_MODE): Redefine.
7402 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7403 Alan Hayward <alan.hayward@arm.com>
7404 David Sherwood <david.sherwood@arm.com>
7406 * target.def (array_mode): New target hook.
7407 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7408 * doc/tm.texi: Regenerate.
7409 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7410 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7411 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7413 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7416 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7417 Alan Hayward <alan.hayward@arm.com>
7418 David Sherwood <david.sherwood@arm.com>
7420 * fold-const.c (fold_binary_loc): Check the argument types
7421 rather than the result type when testing for a vector operation.
7423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7425 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7426 * doc/tm.texi: Regenerate.
7428 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7429 Alan Hayward <alan.hayward@arm.com>
7430 David Sherwood <david.sherwood@arm.com>
7432 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7433 (sve): Document new AArch64 extension.
7434 * doc/md.texi (w): Extend the description of the AArch64
7435 constraint to include SVE vectors.
7436 (Upl, Upa): Document new AArch64 predicate constraints.
7437 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7439 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7440 (msve-vector-bits=): New option.
7441 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7442 SVE when these are disabled.
7443 (sve): New extension.
7444 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7445 modes. Adjust their number of units based on aarch64_sve_vg.
7446 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7447 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7448 aarch64_addr_query_type.
7449 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7450 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7451 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7452 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7453 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7454 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7455 (aarch64_simd_imm_zero_p): Delete.
7456 (aarch64_check_zero_based_sve_index_immediate): Declare.
7457 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7458 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7459 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7460 (aarch64_sve_float_mul_immediate_p): Likewise.
7461 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7463 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7464 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7465 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7466 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7467 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7468 (aarch64_regmode_natural_size): Likewise.
7469 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7470 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7472 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7473 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7474 for VG and the SVE predicate registers.
7475 (V_ALIASES): Add a "z"-prefixed alias.
7476 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7477 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7478 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7479 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7480 (REG_CLASS_NAMES): Add entries for them.
7481 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7482 and the predicate registers.
7483 (aarch64_sve_vg): Declare.
7484 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7485 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7486 (REGMODE_NATURAL_SIZE): Define.
7487 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7489 * config/aarch64/aarch64.c: Include cfgrtl.h.
7490 (simd_immediate_info): Add a constructor for series vectors,
7491 and an associated step field.
7492 (aarch64_sve_vg): New variable.
7493 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7494 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7495 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7496 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7497 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7498 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7499 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7500 (aarch64_get_mask_mode): New functions.
7501 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7502 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7503 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7504 predicate modes and predicate registers. Explicitly restrict
7505 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7506 to store a vector mode if it is recognized by
7507 aarch64_classify_vector_mode.
7508 (aarch64_regmode_natural_size): New function.
7509 (aarch64_hard_regno_caller_save_mode): Return the original mode
7511 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7512 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7513 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7514 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7516 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7517 does not overlap dest if the function is frame-related. Handle
7519 (aarch64_split_add_offset): New function.
7520 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7521 them aarch64_add_offset.
7522 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7523 and update call to aarch64_sub_sp.
7524 (aarch64_add_cfa_expression): New function.
7525 (aarch64_expand_prologue): Pass extra temporary registers to the
7526 functions above. Handle the case in which we need to emit new
7527 DW_CFA_expressions for registers that were originally saved
7528 relative to the stack pointer, but now have to be expressed
7529 relative to the frame pointer.
7530 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7532 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7533 IP0 and IP1 values for SVE frames.
7534 (aarch64_expand_vec_series): New function.
7535 (aarch64_expand_sve_widened_duplicate): Likewise.
7536 (aarch64_expand_sve_const_vector): Likewise.
7537 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7538 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7539 into the register, rather than emitting a SET directly.
7540 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7541 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7542 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7543 (offset_9bit_signed_scaled_p): New functions.
7544 (aarch64_replicate_bitmask_imm): New function.
7545 (aarch64_bitmask_imm): Use it.
7546 (aarch64_cannot_force_const_mem): Reject expressions involving
7547 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7548 (aarch64_classify_index): Handle SVE indices, by requiring
7549 a plain register index with a scale that matches the element size.
7550 (aarch64_classify_address): Handle SVE addresses. Assert that
7551 the mode of the address is VOIDmode or an integer mode.
7552 Update call to aarch64_classify_symbol.
7553 (aarch64_classify_symbolic_expression): Update call to
7554 aarch64_classify_symbol.
7555 (aarch64_const_vec_all_in_range_p): New function.
7556 (aarch64_print_vector_float_operand): Likewise.
7557 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7558 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7559 and the FP immediates 1.0 and 0.5.
7560 (aarch64_print_address_internal): Handle SVE addresses.
7561 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7562 (aarch64_regno_regclass): Handle predicate registers.
7563 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7565 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7566 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7567 (aarch64_convert_sve_vector_bits): New function.
7568 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7569 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7571 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7572 Handle SVE vector and predicate modes. Accept VL-based constants
7573 that need only one temporary register, and VL offsets that require
7574 no temporary registers.
7575 (aarch64_conditional_register_usage): Mark the predicate registers
7576 as fixed if SVE isn't available.
7577 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7578 Return true for SVE vector and predicate modes.
7579 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7580 rather than an unsigned int. Handle SVE modes.
7581 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7583 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7585 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7586 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7587 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7588 (aarch64_sve_float_mul_immediate_p): New functions.
7589 (aarch64_sve_valid_immediate): New function.
7590 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7591 Explicitly reject structure modes. Check for INDEX constants.
7592 Handle PTRUE and PFALSE constants.
7593 (aarch64_check_zero_based_sve_index_immediate): New function.
7594 (aarch64_simd_imm_zero_p): Delete.
7595 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7596 vector modes. Accept constants in the range of CNT[BHWD].
7597 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7598 ask for an Advanced SIMD mode.
7599 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7600 (aarch64_simd_vector_alignment): Handle SVE predicates.
7601 (aarch64_vectorize_preferred_vector_alignment): New function.
7602 (aarch64_simd_vector_alignment_reachable): Use it instead of
7604 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7605 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7607 (MAX_VECT_LEN): Delete.
7608 (expand_vec_perm_d): Add a vec_flags field.
7609 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7610 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7611 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7613 (aarch64_evpc_rev): Rename to...
7614 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7615 (aarch64_evpc_rev_global): New function.
7616 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7617 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7619 (aarch64_evpc_sve_tbl): New function.
7620 (aarch64_expand_vec_perm_const_1): Update after rename of
7621 aarch64_evpc_rev. Handle SVE permutes too, trying
7622 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7623 than aarch64_evpc_tbl.
7624 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7625 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7626 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7627 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7628 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7629 (aarch64_expand_sve_vcond): New functions.
7630 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7631 of aarch64_vector_mode_p.
7632 (aarch64_dwarf_poly_indeterminate_value): New function.
7633 (aarch64_compute_pressure_classes): Likewise.
7634 (aarch64_can_change_mode_class): Likewise.
7635 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7636 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7637 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7638 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7639 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7640 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7641 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7642 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7644 (Dn, Dl, Dr): Accept const as well as const_vector.
7645 (Dz): Likewise. Compare against CONST0_RTX.
7646 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7647 of "vector" where appropriate.
7648 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7649 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7650 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7651 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7652 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7653 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7654 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7655 (v_int_equiv): Extend to SVE modes.
7656 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7658 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7659 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7660 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7661 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7662 (SVE_COND_FP_CMP): New int iterators.
7663 (perm_hilo): Handle the new unpack unspecs.
7664 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7666 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7667 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7668 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7669 (aarch64_equality_operator, aarch64_constant_vector_operand)
7670 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7671 (aarch64_sve_nonimmediate_operand): Likewise.
7672 (aarch64_sve_general_operand): Likewise.
7673 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7674 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7675 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7676 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7677 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7678 (aarch64_sve_float_arith_immediate): Likewise.
7679 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7680 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7681 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7682 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7683 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7684 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7685 (aarch64_sve_float_arith_operand): Likewise.
7686 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7687 (aarch64_sve_float_mul_operand): Likewise.
7688 (aarch64_sve_vec_perm_operand): Likewise.
7689 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7690 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7691 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7692 as well as const_vector.
7693 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7694 in file. Use CONST0_RTX and CONSTM1_RTX.
7695 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7696 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7697 Use aarch64_simd_imm_zero.
7698 * config/aarch64/aarch64-sve.md: New file.
7699 * config/aarch64/aarch64.md: Include it.
7700 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7701 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7702 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7703 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7704 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7705 (sve): New attribute.
7706 (enabled): Disable instructions with the sve attribute unless
7708 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7709 aarch64_expand_mov_immediate.
7710 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7711 CNT[BHSD] immediates.
7712 (movti): Split CONST_POLY_INT moves into two halves.
7713 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7714 Split additions that need a temporary here if the destination
7715 is the stack pointer.
7716 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7717 (*add<mode>3_poly_1): New instruction.
7718 (set_clobber_cc): New expander.
7720 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7722 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7723 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7724 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7725 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7726 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7727 Change innermode from fixed_mode_size to machine_mode.
7728 (simplify_subreg): Update call accordingly. Handle a constant-sized
7729 subreg of a variable-length CONST_VECTOR.
7731 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7732 Alan Hayward <alan.hayward@arm.com>
7733 David Sherwood <david.sherwood@arm.com>
7735 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7736 (add_offset_to_base): New function, split out from...
7737 (create_mem_ref): ...here. When handling a scale other than 1,
7738 check first whether the address is valid without the offset.
7739 Add it into the base if so, leaving the index and scale as-is.
7741 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7744 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7745 fold_for_warn before checking if arg2 is INTEGER_CST.
7747 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7749 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7750 (store_multiple_operation): Delete.
7751 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7752 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7753 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7754 guarded by TARGET_STRING.
7755 (rs6000_output_load_multiple): Delete.
7756 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7757 OPTION_MASK_STRING / TARGET_STRING handling.
7758 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7759 (const rs6000_opt_masks) <"string">: Change mask to 0.
7760 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7761 (MASK_STRING): Delete.
7762 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7764 (load_multiple): Delete.
7771 (store_multiple): Delete.
7778 (movmemsi_8reg): Delete.
7779 (corresponding unnamed define_insn): Delete.
7780 (movmemsi_6reg): Delete.
7781 (corresponding unnamed define_insn): Delete.
7782 (movmemsi_4reg): Delete.
7783 (corresponding unnamed define_insn): Delete.
7784 (movmemsi_2reg): Delete.
7785 (corresponding unnamed define_insn): Delete.
7786 (movmemsi_1reg): Delete.
7787 (corresponding unnamed define_insn): Delete.
7788 * config/rs6000/rs6000.opt (mno-string): New.
7789 (mstring): Replace by deprecation warning stub.
7790 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7792 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7794 * regrename.c (regrename_do_replace): If replacing the same
7795 reg multiple times, try to reuse last created gen_raw_REG.
7798 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7799 main to workaround a bug in GDB.
7801 2018-01-12 Tom de Vries <tom@codesourcery.com>
7804 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7806 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7808 PR rtl-optimization/80481
7809 * ira-color.c (get_cap_member): New function.
7810 (allocnos_conflict_by_live_ranges_p): Use it.
7811 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7812 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7814 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7817 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7818 (*saddl_se_1): Ditto.
7820 (*ssubl_se_1): Ditto.
7822 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7824 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7825 rather than wi::to_widest for DR_INITs.
7826 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7827 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7828 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7830 (vect_analyze_group_access_1): Note that here.
7832 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7834 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7835 polynomial type sizes.
7837 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7839 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7840 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7841 (gimple_add_tmp_var): Likewise.
7843 2018-01-12 Martin Liska <mliska@suse.cz>
7845 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7846 (gimple_alloc_sizes): Likewise.
7847 (dump_gimple_statistics): Use PRIu64 in printf format.
7848 * gimple.h: Change uint64_t to int.
7850 2018-01-12 Martin Liska <mliska@suse.cz>
7852 * tree-core.h: Use uint64_t instead of int.
7853 * tree.c (tree_node_counts): Likewise.
7854 (tree_node_sizes): Likewise.
7855 (dump_tree_statistics): Use PRIu64 in printf format.
7857 2018-01-12 Martin Liska <mliska@suse.cz>
7859 * Makefile.in: As qsort_chk is implemented in vec.c, add
7860 vec.o to linkage of gencfn-macros.
7861 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7862 passing the info to record_node_allocation_statistics.
7863 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7865 * ggc-common.c (struct ggc_usage): Add operator== and use
7866 it in operator< and compare function.
7867 * mem-stats.h (struct mem_usage): Likewise.
7868 * vec.c (struct vec_usage): Remove operator< and compare
7869 function. Can be simply inherited.
7871 2018-01-12 Martin Jambor <mjambor@suse.cz>
7874 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7875 * tree-ssa-math-opts.c: Include domwalk.h.
7876 (convert_mult_to_fma_1): New function.
7877 (fma_transformation_info): New type.
7878 (fma_deferring_state): Likewise.
7879 (cancel_fma_deferring): New function.
7880 (result_of_phi): Likewise.
7881 (last_fma_candidate_feeds_initial_phi): Likewise.
7882 (convert_mult_to_fma): Added deferring logic, split actual
7883 transformation to convert_mult_to_fma_1.
7884 (math_opts_dom_walker): New type.
7885 (math_opts_dom_walker::after_dom_children): New method, body moved
7886 here from pass_optimize_widening_mul::execute, added deferring logic
7888 (pass_optimize_widening_mul::execute): Moved most of code to
7889 math_opts_dom_walker::after_dom_children.
7890 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7891 * config/i386/i386.c (ix86_option_override_internal): Added
7892 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7894 2018-01-12 Richard Biener <rguenther@suse.de>
7897 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7898 inline instance vars.
7900 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7903 * config/rx/rx.c (rx_is_restricted_memory_address):
7906 2018-01-12 Richard Biener <rguenther@suse.de>
7908 PR tree-optimization/80846
7909 * target.def (split_reduction): New target hook.
7910 * targhooks.c (default_split_reduction): New function.
7911 * targhooks.h (default_split_reduction): Declare.
7912 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7913 target requests first reduce vectors by combining low and high
7915 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7916 (get_vectype_for_scalar_type_and_size): Export.
7917 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7918 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7919 * doc/tm.texi: Regenerate.
7920 * config/i386/i386.c (ix86_split_reduction): Implement
7921 TARGET_VECTORIZE_SPLIT_REDUCTION.
7923 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7926 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7927 in PIC mode except for TARGET_VXWORKS_RTP.
7928 * config/sparc/sparc.c: Include cfgrtl.h.
7929 (TARGET_INIT_PIC_REG): Define.
7930 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7931 (sparc_pic_register_p): New predicate.
7932 (sparc_legitimate_address_p): Use it.
7933 (sparc_legitimize_pic_address): Likewise.
7934 (sparc_delegitimize_address): Likewise.
7935 (sparc_mode_dependent_address_p): Likewise.
7936 (gen_load_pcrel_sym): Remove 4th parameter.
7937 (load_got_register): Adjust call to above. Remove obsolete stuff.
7938 (sparc_expand_prologue): Do not call load_got_register here.
7939 (sparc_flat_expand_prologue): Likewise.
7940 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7941 (sparc_use_pseudo_pic_reg): New function.
7942 (sparc_init_pic_reg): Likewise.
7943 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7944 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7946 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7948 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7949 Add item for branch_cost.
7951 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7953 PR rtl-optimization/83565
7954 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7955 not extend the result to a larger mode for rotate operations.
7956 (num_sign_bit_copies1): Likewise.
7958 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7961 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7963 Use values-Xc.o for -pedantic.
7964 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7966 2018-01-12 Martin Liska <mliska@suse.cz>
7969 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7971 (possible_polymorphic_call_targets): Use it.
7972 (ipa_devirt): Likewise.
7974 2018-01-12 Martin Liska <mliska@suse.cz>
7976 * profile-count.h (enum profile_quality): Use 0 as invalid
7977 enum value of profile_quality.
7979 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7981 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7982 -mext-string options.
7984 2018-01-12 Richard Biener <rguenther@suse.de>
7986 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7987 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7988 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7990 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7992 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7994 * configure.ac (--with-long-double-format): Add support for the
7995 configuration option to change the default long double format on
7997 * config.gcc (powerpc*-linux*-*): Likewise.
7998 * configure: Regenerate.
7999 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
8000 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
8001 used without modification.
8003 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8005 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
8006 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
8007 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
8008 MISC_BUILTIN_SPEC_BARRIER.
8009 (rs6000_init_builtins): Likewise.
8010 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
8012 (speculation_barrier): New define_insn.
8013 * doc/extend.texi: Document __builtin_speculation_barrier.
8015 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8018 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
8019 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
8020 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
8022 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
8023 integral modes instead of "ss" and "sd".
8024 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
8025 vectors with 32-bit and 64-bit elements.
8026 (vecdupssescalarmodesuffix): New mode attribute.
8027 (vec_dup<mode>): Use it.
8029 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
8032 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
8033 frame if argument is passed on stack.
8035 2018-01-11 Jakub Jelinek <jakub@redhat.com>
8038 * ree.c (combine_reaching_defs): Optimize also
8039 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
8040 reg2=any_extend(exp); reg1=reg2;, formatting fix.
8042 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8045 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
8047 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
8050 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
8051 after they are computed.
8053 2018-01-11 Bin Cheng <bin.cheng@arm.com>
8055 PR tree-optimization/83695
8056 * gimple-loop-linterchange.cc
8057 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
8058 reset cached scev information after interchange.
8059 (pass_linterchange::execute): Remove call to scev_reset_htab.
8061 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8063 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
8064 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
8065 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
8066 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
8067 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
8068 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
8069 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
8070 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
8071 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
8072 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
8073 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
8074 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
8075 (V_lane_reg): Likewise.
8076 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
8078 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
8079 (vfmal_lane_low<mode>_intrinsic,
8080 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
8081 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
8082 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
8083 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
8084 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
8085 vfmsl_lane_high<mode>_intrinsic): New define_insns.
8087 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8089 * config/arm/arm-cpus.in (fp16fml): New feature.
8090 (ALL_SIMD): Add fp16fml.
8091 (armv8.2-a): Add fp16fml as an option.
8092 (armv8.3-a): Likewise.
8093 (armv8.4-a): Add fp16fml as part of fp16.
8094 * config/arm/arm.h (TARGET_FP16FML): Define.
8095 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
8097 * config/arm/arm-modes.def (V2HF): Define.
8098 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
8099 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
8100 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
8101 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
8102 vfmsl_low, vfmsl_high): New set of builtins.
8103 * config/arm/iterators.md (PLUSMINUS): New code iterator.
8104 (vfml_op): New code attribute.
8105 (VFMLHALVES): New int iterator.
8106 (VFML, VFMLSEL): New mode attributes.
8107 (V_reg): Define mapping for V2HF.
8108 (V_hi, V_lo): New mode attributes.
8109 (VF_constraint): Likewise.
8110 (vfml_half, vfml_half_selector): New int attributes.
8111 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
8113 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
8114 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
8116 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
8117 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
8118 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
8119 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
8121 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8122 Document new effective target and option set.
8124 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8126 * config/arm/arm-cpus.in (armv8_4): New feature.
8127 (ARMv8_4a): New fgroup.
8128 (armv8.4-a): New arch.
8129 * config/arm/arm-tables.opt: Regenerate.
8130 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8131 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8132 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8133 Add matching rules for -march=armv8.4-a and extensions.
8134 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8136 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8139 * config/rx/rx.md (BW): New mode attribute.
8140 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8142 2018-01-11 Richard Biener <rguenther@suse.de>
8144 PR tree-optimization/83435
8145 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8146 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8147 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8149 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8150 Alan Hayward <alan.hayward@arm.com>
8151 David Sherwood <david.sherwood@arm.com>
8153 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8155 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8156 (aarch64_print_address_internal): Use it to check for a zero offset.
8158 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8159 Alan Hayward <alan.hayward@arm.com>
8160 David Sherwood <david.sherwood@arm.com>
8162 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8163 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8164 Return a poly_int64 rather than a HOST_WIDE_INT.
8165 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8166 rather than a HOST_WIDE_INT.
8167 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8168 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8169 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8170 final_offset from HOST_WIDE_INT to poly_int64.
8171 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8172 to_constant when getting the number of units in an Advanced SIMD
8174 (aarch64_builtin_vectorized_function): Check for a constant number
8176 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8178 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8179 attribute instead of GET_MODE_NUNITS.
8180 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8181 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8182 GET_MODE_SIZE for fixed-size registers.
8183 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8184 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8185 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8186 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8187 (aarch64_print_operand, aarch64_print_address_internal)
8188 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8189 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8190 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8191 Handle polynomial GET_MODE_SIZE.
8192 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8193 wider than SImode without modification.
8194 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8195 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8196 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8197 passing and returning SVE modes.
8198 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8199 rather than GEN_INT.
8200 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8201 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8202 (aarch64_allocate_and_probe_stack_space): Likewise.
8203 (aarch64_layout_frame): Cope with polynomial offsets.
8204 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8205 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8207 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8208 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8209 poly_int64 rather than a HOST_WIDE_INT.
8210 (aarch64_get_separate_components, aarch64_process_components)
8211 (aarch64_expand_prologue, aarch64_expand_epilogue)
8212 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8213 (aarch64_anchor_offset): New function, split out from...
8214 (aarch64_legitimize_address): ...here.
8215 (aarch64_builtin_vectorization_cost): Handle polynomial
8216 TYPE_VECTOR_SUBPARTS.
8217 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8219 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8220 number of elements from the PARALLEL rather than the mode.
8221 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8222 rather than GET_MODE_BITSIZE.
8223 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8224 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8225 (aarch64_expand_vec_perm_const_1): Handle polynomial
8226 d->perm.length () and d->perm elements.
8227 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8228 Apply to_constant to d->perm elements.
8229 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8230 polynomial CONST_VECTOR_NUNITS.
8231 (aarch64_move_pointer): Take amount as a poly_int64 rather
8233 (aarch64_progress_pointer): Avoid temporary variable.
8234 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8235 the mode attribute instead of GET_MODE.
8237 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8238 Alan Hayward <alan.hayward@arm.com>
8239 David Sherwood <david.sherwood@arm.com>
8241 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8242 x exists before using it.
8243 (aarch64_add_constant_internal): Rename to...
8244 (aarch64_add_offset_1): ...this. Replace regnum with separate
8245 src and dest rtxes. Handle the case in which they're different,
8246 including when the offset is zero. Replace scratchreg with an rtx.
8247 Use 2 additions if there is no spare register into which we can
8248 move a 16-bit constant.
8249 (aarch64_add_constant): Delete.
8250 (aarch64_add_offset): Replace reg with separate src and dest
8251 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8252 Use aarch64_add_offset_1.
8253 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8254 an rtx rather than an int. Take the delta as a poly_int64
8255 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8256 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8257 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8258 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8259 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8261 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8262 aarch64_add_constant.
8264 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8266 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8267 Use scalar_float_mode.
8269 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8271 * config/aarch64/aarch64-simd.md
8272 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8273 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8274 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8275 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8276 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8277 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8278 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8279 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8280 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8281 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8283 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8286 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8287 targ_options->x_arm_arch_string is non NULL.
8289 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8291 * config/aarch64/aarch64.h
8292 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8294 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8297 * expmed.c (emit_store_flag_force): Swap if const op0
8298 and change VOIDmode to mode of op0.
8300 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8302 PR rtl-optimization/83761
8303 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8304 than bytes to mode_for_size.
8306 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8309 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8310 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8313 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8316 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8317 when in layout mode.
8318 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8319 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8322 2018-01-10 Michael Collison <michael.collison@arm.com>
8324 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8325 * config/aarch64/aarch64-option-extension.def: Add
8326 AARCH64_OPT_EXTENSION of 'fp16fml'.
8327 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8328 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8329 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8330 * config/aarch64/constraints.md (Ui7): New constraint.
8331 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8332 (VFMLA_SEL_W): Ditto.
8335 (VFMLA16_LOW): New int iterator.
8336 (VFMLA16_HIGH): Ditto.
8337 (UNSPEC_FMLAL): New unspec.
8338 (UNSPEC_FMLSL): Ditto.
8339 (UNSPEC_FMLAL2): Ditto.
8340 (UNSPEC_FMLSL2): Ditto.
8341 (f16mac): New code attribute.
8342 * config/aarch64/aarch64-simd-builtins.def
8343 (aarch64_fmlal_lowv2sf): Ditto.
8344 (aarch64_fmlsl_lowv2sf): Ditto.
8345 (aarch64_fmlalq_lowv4sf): Ditto.
8346 (aarch64_fmlslq_lowv4sf): Ditto.
8347 (aarch64_fmlal_highv2sf): Ditto.
8348 (aarch64_fmlsl_highv2sf): Ditto.
8349 (aarch64_fmlalq_highv4sf): Ditto.
8350 (aarch64_fmlslq_highv4sf): Ditto.
8351 (aarch64_fmlal_lane_lowv2sf): Ditto.
8352 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8353 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8354 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8355 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8356 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8357 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8358 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8359 (aarch64_fmlal_lane_highv2sf): Ditto.
8360 (aarch64_fmlsl_lane_highv2sf): Ditto.
8361 (aarch64_fmlal_laneq_highv2sf): Ditto.
8362 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8363 (aarch64_fmlalq_lane_highv4sf): Ditto.
8364 (aarch64_fmlsl_lane_highv4sf): Ditto.
8365 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8366 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8367 * config/aarch64/aarch64-simd.md:
8368 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8369 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8370 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8371 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8372 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8373 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8374 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8375 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8376 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8377 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8378 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8379 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8380 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8381 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8382 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8383 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8384 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8385 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8386 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8387 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8388 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8389 (vfmlsl_low_u32): Ditto.
8390 (vfmlalq_low_u32): Ditto.
8391 (vfmlslq_low_u32): Ditto.
8392 (vfmlal_high_u32): Ditto.
8393 (vfmlsl_high_u32): Ditto.
8394 (vfmlalq_high_u32): Ditto.
8395 (vfmlslq_high_u32): Ditto.
8396 (vfmlal_lane_low_u32): Ditto.
8397 (vfmlsl_lane_low_u32): Ditto.
8398 (vfmlal_laneq_low_u32): Ditto.
8399 (vfmlsl_laneq_low_u32): Ditto.
8400 (vfmlalq_lane_low_u32): Ditto.
8401 (vfmlslq_lane_low_u32): Ditto.
8402 (vfmlalq_laneq_low_u32): Ditto.
8403 (vfmlslq_laneq_low_u32): Ditto.
8404 (vfmlal_lane_high_u32): Ditto.
8405 (vfmlsl_lane_high_u32): Ditto.
8406 (vfmlal_laneq_high_u32): Ditto.
8407 (vfmlsl_laneq_high_u32): Ditto.
8408 (vfmlalq_lane_high_u32): Ditto.
8409 (vfmlslq_lane_high_u32): Ditto.
8410 (vfmlalq_laneq_high_u32): Ditto.
8411 (vfmlslq_laneq_high_u32): Ditto.
8412 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8413 (AARCH64_FL_FOR_ARCH8_4): New.
8414 (AARCH64_ISA_F16FML): New ISA flag.
8415 (TARGET_F16FML): New feature flag for fp16fml.
8416 (doc/invoke.texi): Document new fp16fml option.
8418 2018-01-10 Michael Collison <michael.collison@arm.com>
8420 * config/aarch64/aarch64-builtins.c:
8421 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8422 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8423 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8424 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8425 (AARCH64_ISA_SHA3): New ISA flag.
8426 (TARGET_SHA3): New feature flag for sha3.
8427 * config/aarch64/iterators.md (sha512_op): New int attribute.
8428 (CRYPTO_SHA512): New int iterator.
8429 (UNSPEC_SHA512H): New unspec.
8430 (UNSPEC_SHA512H2): Ditto.
8431 (UNSPEC_SHA512SU0): Ditto.
8432 (UNSPEC_SHA512SU1): Ditto.
8433 * config/aarch64/aarch64-simd-builtins.def
8434 (aarch64_crypto_sha512hqv2di): New builtin.
8435 (aarch64_crypto_sha512h2qv2di): Ditto.
8436 (aarch64_crypto_sha512su0qv2di): Ditto.
8437 (aarch64_crypto_sha512su1qv2di): Ditto.
8438 (aarch64_eor3qv8hi): Ditto.
8439 (aarch64_rax1qv2di): Ditto.
8440 (aarch64_xarqv2di): Ditto.
8441 (aarch64_bcaxqv8hi): Ditto.
8442 * config/aarch64/aarch64-simd.md:
8443 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8444 (aarch64_crypto_sha512su0qv2di): Ditto.
8445 (aarch64_crypto_sha512su1qv2di): Ditto.
8446 (aarch64_eor3qv8hi): Ditto.
8447 (aarch64_rax1qv2di): Ditto.
8448 (aarch64_xarqv2di): Ditto.
8449 (aarch64_bcaxqv8hi): Ditto.
8450 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8451 (vsha512h2q_u64): Ditto.
8452 (vsha512su0q_u64): Ditto.
8453 (vsha512su1q_u64): Ditto.
8454 (veor3q_u16): Ditto.
8455 (vrax1q_u64): Ditto.
8457 (vbcaxq_u16): Ditto.
8458 * config/arm/types.md (crypto_sha512): New type attribute.
8459 (crypto_sha3): Ditto.
8460 (doc/invoke.texi): Document new sha3 option.
8462 2018-01-10 Michael Collison <michael.collison@arm.com>
8464 * config/aarch64/aarch64-builtins.c:
8465 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8466 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8467 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8468 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8469 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8470 (AARCH64_ISA_SM4): New ISA flag.
8471 (TARGET_SM4): New feature flag for sm4.
8472 * config/aarch64/aarch64-simd-builtins.def
8473 (aarch64_sm3ss1qv4si): Ditto.
8474 (aarch64_sm3tt1aq4si): Ditto.
8475 (aarch64_sm3tt1bq4si): Ditto.
8476 (aarch64_sm3tt2aq4si): Ditto.
8477 (aarch64_sm3tt2bq4si): Ditto.
8478 (aarch64_sm3partw1qv4si): Ditto.
8479 (aarch64_sm3partw2qv4si): Ditto.
8480 (aarch64_sm4eqv4si): Ditto.
8481 (aarch64_sm4ekeyqv4si): Ditto.
8482 * config/aarch64/aarch64-simd.md:
8483 (aarch64_sm3ss1qv4si): Ditto.
8484 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8485 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8486 (aarch64_sm4eqv4si): Ditto.
8487 (aarch64_sm4ekeyqv4si): Ditto.
8488 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8489 (sm3part_op): Ditto.
8490 (CRYPTO_SM3TT): Ditto.
8491 (CRYPTO_SM3PART): Ditto.
8492 (UNSPEC_SM3SS1): New unspec.
8493 (UNSPEC_SM3TT1A): Ditto.
8494 (UNSPEC_SM3TT1B): Ditto.
8495 (UNSPEC_SM3TT2A): Ditto.
8496 (UNSPEC_SM3TT2B): Ditto.
8497 (UNSPEC_SM3PARTW1): Ditto.
8498 (UNSPEC_SM3PARTW2): Ditto.
8499 (UNSPEC_SM4E): Ditto.
8500 (UNSPEC_SM4EKEY): Ditto.
8501 * config/aarch64/constraints.md (Ui2): New constraint.
8502 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8503 * config/arm/types.md (crypto_sm3): New type attribute.
8504 (crypto_sm4): Ditto.
8505 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8506 (vsm3tt1aq_u32): Ditto.
8507 (vsm3tt1bq_u32): Ditto.
8508 (vsm3tt2aq_u32): Ditto.
8509 (vsm3tt2bq_u32): Ditto.
8510 (vsm3partw1q_u32): Ditto.
8511 (vsm3partw2q_u32): Ditto.
8512 (vsm4eq_u32): Ditto.
8513 (vsm4ekeyq_u32): Ditto.
8514 (doc/invoke.texi): Document new sm4 option.
8516 2018-01-10 Michael Collison <michael.collison@arm.com>
8518 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8519 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8520 (AARCH64_FL_FOR_ARCH8_4): New.
8521 (AARCH64_FL_V8_4): New flag.
8522 (doc/invoke.texi): Document new armv8.4-a option.
8524 2018-01-10 Michael Collison <michael.collison@arm.com>
8526 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8527 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8528 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8529 * config/aarch64/aarch64-option-extension.def: Add
8530 AARCH64_OPT_EXTENSION of 'sha2'.
8531 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8532 (crypto): Disable sha2 and aes if crypto disabled.
8533 (crypto): Enable aes and sha2 if enabled.
8534 (simd): Disable sha2 and aes if simd disabled.
8535 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8537 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8538 (TARGET_SHA2): New feature flag for sha2.
8539 (TARGET_AES): New feature flag for aes.
8540 * config/aarch64/aarch64-simd.md:
8541 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8542 conditional on TARGET_AES.
8543 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8544 (aarch64_crypto_sha1hsi): Make pattern conditional
8546 (aarch64_crypto_sha1hv4si): Ditto.
8547 (aarch64_be_crypto_sha1hv4si): Ditto.
8548 (aarch64_crypto_sha1su1v4si): Ditto.
8549 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8550 (aarch64_crypto_sha1su0v4si): Ditto.
8551 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8552 (aarch64_crypto_sha256su0v4si): Ditto.
8553 (aarch64_crypto_sha256su1v4si): Ditto.
8554 (doc/invoke.texi): Document new aes and sha2 options.
8556 2018-01-10 Martin Sebor <msebor@redhat.com>
8558 PR tree-optimization/83781
8559 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8562 2018-01-11 Martin Sebor <msebor@gmail.com>
8563 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8565 PR tree-optimization/83501
8566 PR tree-optimization/81703
8568 * tree-ssa-strlen.c (get_string_cst): Rename...
8569 (get_string_len): ...to this. Handle global constants.
8570 (handle_char_store): Adjust.
8572 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8573 Jim Wilson <jimw@sifive.com>
8575 * config/riscv/riscv-protos.h (riscv_output_return): New.
8576 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8577 (riscv_attribute_table, riscv_output_return),
8578 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8579 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8580 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8582 (riscv_expand_prologue): Add early return for naked function.
8583 (riscv_expand_epilogue): Likewise.
8584 (riscv_function_ok_for_sibcall): Return false for naked function.
8585 (riscv_set_current_function): New.
8586 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8587 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8588 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8589 * doc/extend.texi (RISC-V Function Attributes): New.
8591 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8593 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8594 check for 128-bit long double before checking TCmode.
8595 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8596 128-bit long doubles before checking TFmode or TCmode.
8597 (FLOAT128_IBM_P): Likewise.
8599 2018-01-10 Martin Sebor <msebor@redhat.com>
8601 PR tree-optimization/83671
8602 * builtins.c (c_strlen): Unconditionally return zero for the empty
8604 Use -Warray-bounds for warnings.
8605 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8606 for non-constant array indices with COMPONENT_REF, arrays of
8607 arrays, and pointers to arrays.
8608 (gimple_fold_builtin_strlen): Determine and set length range for
8609 non-constant character arrays.
8611 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8614 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8617 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8619 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8621 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8624 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8625 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8626 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8627 indexed_or_indirect_operand predicate.
8628 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8629 (*vsx_le_perm_load_v8hi): Likewise.
8630 (*vsx_le_perm_load_v16qi): Likewise.
8631 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8632 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8633 (*vsx_le_perm_store_v8hi): Likewise.
8634 (*vsx_le_perm_store_v16qi): Likewise.
8635 (eight unnamed splitters): Likewise.
8637 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8639 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8640 * config/rs6000/emmintrin.h: Likewise.
8641 * config/rs6000/mmintrin.h: Likewise.
8642 * config/rs6000/xmmintrin.h: Likewise.
8644 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8647 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8649 * tree.c (tree_nop_conversion): Return true for location wrapper
8651 (maybe_wrap_with_location): New function.
8652 (selftest::check_strip_nops): New function.
8653 (selftest::test_location_wrappers): New function.
8654 (selftest::tree_c_tests): Call it.
8655 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8656 (maybe_wrap_with_location): New decl.
8657 (EXPR_LOCATION_WRAPPER_P): New macro.
8658 (location_wrapper_p): New inline function.
8659 (tree_strip_any_location_wrapper): New inline function.
8661 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8664 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8665 stack_realign_offset for the largest alignment of stack slot
8667 (ix86_find_max_used_stack_alignment): New function.
8668 (ix86_finalize_stack_frame_flags): Use it. Set
8669 max_used_stack_alignment if we don't realign stack.
8670 * config/i386/i386.h (machine_function): Add
8671 max_used_stack_alignment.
8673 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8675 * config/arm/arm.opt (-mbranch-cost): New option.
8676 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8679 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8682 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8683 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8685 2018-01-10 Richard Biener <rguenther@suse.de>
8688 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8689 early out so it also covers the case where we have a non-NULL
8692 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8694 PR tree-optimization/83753
8695 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8696 for non-strided grouped accesses if the number of elements is 1.
8698 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8701 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8702 * i386.h (TARGET_USE_GATHER): Define.
8703 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8705 2018-01-10 Martin Liska <mliska@suse.cz>
8708 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8709 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8711 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8712 CLEANUP_NO_PARTITIONING is not set.
8714 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8716 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8717 for vectors, as a partial revert of r254296.
8718 * rtl.h (const_vec_p): Delete.
8719 (const_vec_duplicate_p): Don't test for vector CONSTs.
8720 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8721 * expmed.c (make_tree): Likewise.
8724 * common.md (E, F): Use CONSTANT_P instead of checking for
8726 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8727 checking for CONST_VECTOR.
8729 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8732 * predict.c (force_edge_cold): Handle in more sane way edges
8735 2018-01-09 Carl Love <cel@us.ibm.com>
8737 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8739 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8740 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8741 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8742 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8743 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8744 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8745 * config/rs6000/rs6000-protos.h: Add extern defition for
8746 rs6000_generate_float2_double_code.
8747 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8749 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8750 (float2_v2df): Add define_expand.
8752 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8755 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8756 op_mode in the force_to_mode call.
8758 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8760 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8761 instead of checking each element individually.
8762 (aarch64_evpc_uzp): Likewise.
8763 (aarch64_evpc_zip): Likewise.
8764 (aarch64_evpc_ext): Likewise.
8765 (aarch64_evpc_rev): Likewise.
8766 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8767 instead of checking each element individually. Return true without
8769 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8770 whether all selected elements come from the same input, instead of
8771 checking each element individually. Remove calls to gen_rtx_REG,
8772 start_sequence and end_sequence and instead assert that no rtl is
8775 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8777 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8778 order of HIGH and CONST checks.
8780 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8782 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8783 if the destination isn't an SSA_NAME.
8785 2018-01-09 Richard Biener <rguenther@suse.de>
8787 PR tree-optimization/83668
8788 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8790 (canonicalize_loop_form): ... here, renamed from ...
8791 (canonicalize_loop_closed_ssa_form): ... this and amended to
8792 swap successor edges for loop exit blocks to make us use
8793 the RPO order we need for initial schedule generation.
8795 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8797 PR tree-optimization/64811
8798 * match.pd: When optimizing comparisons with Inf, avoid
8799 introducing or losing exceptions from comparisons with NaN.
8801 2018-01-09 Martin Liska <mliska@suse.cz>
8804 * asan.c (shadow_mem_size): Add gcc_assert.
8806 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8808 Don't save registers in main().
8811 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8812 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8813 * config/avr/avr.c (avr_set_current_function): Don't error if
8814 naked, OS_task or OS_main are specified at the same time.
8815 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8817 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8819 * common/config/avr/avr-common.c (avr_option_optimization_table):
8820 Switch on -mmain-is-OS_task for optimizing compilations.
8822 2018-01-09 Richard Biener <rguenther@suse.de>
8824 PR tree-optimization/83572
8825 * graphite.c: Include cfganal.h.
8826 (graphite_transform_loops): Connect infinite loops to exit
8827 and remove fake edges at the end.
8829 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8831 * ipa-inline.c (edge_badness): Revert accidental checkin.
8833 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8836 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8837 symbols; not inline clones.
8839 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8842 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8843 hard registers. Formatting fixes.
8845 PR preprocessor/83722
8846 * gcc.c (try_generate_repro): Pass
8847 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8848 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8851 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8852 Kito Cheng <kito.cheng@gmail.com>
8854 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8855 (riscv_leaf_function_p): Delete.
8856 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8858 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8860 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8862 (do_ifelse): New function.
8863 (do_isel): New function.
8864 (do_sub3): New function.
8865 (do_add3): New function.
8866 (do_load_mask_compare): New function.
8867 (do_overlap_load_compare): New function.
8868 (expand_compare_loop): New function.
8869 (expand_block_compare): Call expand_compare_loop() when appropriate.
8870 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8872 (-mblock-compare-inline-loop-limit): New option.
8874 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8877 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8878 Reverse order of second and third operands in first alternative.
8879 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8880 of first and second elements in UNSPEC_VPERMR vector.
8881 (altivec_expand_vec_perm_le): Likewise.
8883 2018-01-08 Jeff Law <law@redhat.com>
8885 PR rtl-optimizatin/81308
8886 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8887 (process_switch): If group_case_labels makes a change, then set
8889 (pass_convert_switch::execute): If a switch is converted, then
8890 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8892 PR rtl-optimization/81308
8893 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8896 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8898 PR target/83663 - Revert r255946
8899 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8900 generation for cases where splatting a value is not useful.
8901 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8902 across a vec_duplicate and a paradoxical subreg forming a vector
8903 mode to a vec_concat.
8905 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8907 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8908 -march=armv8.3-a variants.
8909 * config/arm/t-multilib: Likewise.
8910 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8912 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8914 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8916 (cceq_ior_compare_complement): Give it a name so I can use it, and
8917 change boolean_or_operator predicate to boolean_operator so it can
8918 be used to generate a crand.
8919 (eqne): New code iterator.
8920 (bd/bd_neg): New code_attrs.
8921 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8922 a single define_insn.
8923 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8924 decrement (bdnzt/bdnzf/bdzt/bdzf).
8925 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8926 with the new names of the branch decrement patterns, and added the
8927 names of the branch decrement conditional patterns.
8929 2018-01-08 Richard Biener <rguenther@suse.de>
8931 PR tree-optimization/83563
8932 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8935 2018-01-08 Richard Biener <rguenther@suse.de>
8938 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8940 2018-01-08 Richard Biener <rguenther@suse.de>
8942 PR tree-optimization/83685
8943 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8944 references to abnormals.
8946 2018-01-08 Richard Biener <rguenther@suse.de>
8949 * dwarf2out.c (output_indirect_strings): Handle empty
8950 skeleton_debug_str_hash.
8951 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8953 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8955 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8956 (emit_store_direct): Likewise.
8957 (arc_trampoline_adjust_address): Likewise.
8958 (arc_asm_trampoline_template): New function.
8959 (arc_initialize_trampoline): Use asm_trampoline_template.
8960 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8961 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8962 * config/arc/arc.md (flush_icache): Delete pattern.
8964 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8966 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8967 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8970 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8973 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8974 by not USED_FOR_TARGET.
8975 (make_pass_resolve_sw_modes): Likewise.
8977 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8979 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8982 2018-01-08 Richard Biener <rguenther@suse.de>
8985 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8987 2018-01-08 Richard Biener <rguenther@suse.de>
8990 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8992 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8995 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8996 basic blocks with a small number of successors.
8997 (convert_control_dep_chain_into_preds): Improve handling of
8999 (dump_predicates): Split apart into...
9000 (dump_pred_chain): ...here...
9001 (dump_pred_info): ...and here.
9002 (can_one_predicate_be_invalidated_p): Add debugging printfs.
9003 (can_chain_union_be_invalidated_p): Improve check for invalidation
9005 (uninit_uses_cannot_happen): Avoid unnecessary if
9006 convert_control_dep_chain_into_preds yielded nothing.
9008 2018-01-06 Martin Sebor <msebor@redhat.com>
9010 PR tree-optimization/83640
9011 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
9012 subtracting negative offset from size.
9013 (builtin_access::overlap): Adjust offset bounds of the access to fall
9014 within the size of the object if possible.
9016 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
9018 PR rtl-optimization/83699
9019 * expmed.c (extract_bit_field_1): Restrict the vector usage of
9020 extract_bit_field_as_subreg to cases in which the extracted
9021 value is also a vector.
9023 * lra-constraints.c (process_alt_operands): Test for the equivalence
9024 substitutions when detecting a possible reload cycle.
9026 2018-01-06 Jakub Jelinek <jakub@redhat.com>
9029 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
9030 by default if flag_selective_schedling{,2}. Formatting fixes.
9032 PR rtl-optimization/83682
9033 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
9034 if it has non-VECTOR_MODE element mode.
9035 (vec_duplicate_p): Likewise.
9038 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
9039 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
9041 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9044 * config/i386/i386-builtin.def
9045 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
9046 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
9047 Require also OPTION_MASK_ISA_AVX512F in addition to
9048 OPTION_MASK_ISA_GFNI.
9049 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
9050 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
9051 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
9052 to OPTION_MASK_ISA_GFNI.
9053 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
9054 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
9055 OPTION_MASK_ISA_AVX512BW.
9056 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
9057 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
9058 addition to OPTION_MASK_ISA_GFNI.
9059 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
9060 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
9061 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
9062 to OPTION_MASK_ISA_GFNI.
9063 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
9064 a requirement for all ISAs rather than any of them with a few
9066 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
9068 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
9069 bitmasks to be enabled with 3 exceptions, instead of requiring any
9070 enabled ISA with lots of exceptions.
9071 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
9072 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
9073 Change avx512bw in isa attribute to avx512f.
9074 * config/i386/sgxintrin.h: Add license boilerplate.
9075 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
9076 to __AVX512F__ and __AVX512VL to __AVX512VL__.
9077 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
9078 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
9080 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
9081 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
9082 temporarily sse2 rather than sse if not enabled already.
9085 * config/i386/sse.md (VI248_VLBW): Rename to ...
9086 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
9087 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
9088 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
9089 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
9090 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
9091 mode iterator instead of VI248_VLBW.
9093 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
9095 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
9096 (record_modified): Skip clobbers; add debug output.
9097 (param_change_prob): Use sreal frequencies.
9099 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9101 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
9102 punt for user-aligned variables.
9104 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
9106 * tree-chrec.c (chrec_contains_symbols): Return true for
9109 2018-01-05 Sudakshina Das <sudi.das@arm.com>
9112 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
9113 of (x|y) == x for BICS pattern.
9115 2018-01-05 Jakub Jelinek <jakub@redhat.com>
9117 PR tree-optimization/83605
9118 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
9119 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9122 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9124 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9125 * config/epiphany/rtems.h: New file.
9127 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9128 Uros Bizjak <ubizjak@gmail.com>
9131 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9132 QIreg_operand instead of register_operand predicate.
9133 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9134 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9135 comments instead of -fmitigate[-_]rop.
9137 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9140 * cgraphunit.c (symbol_table::compile): Switch to text_section
9141 before calling assembly_start debug hook.
9142 * run-rtl-passes.c (run_rtl_passes): Likewise.
9145 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9147 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9148 range_int_cst_p rather than !symbolic_range_p before calling
9149 extract_range_from_multiplicative_op_1.
9151 2018-01-04 Jeff Law <law@redhat.com>
9153 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9154 redundant test in assertion.
9156 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9158 * doc/rtl.texi: Document machine_mode wrapper classes.
9160 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9162 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9165 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9167 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9168 the VEC_PERM_EXPR fold to fail.
9170 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9173 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9174 to switched_sections.
9176 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9179 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9182 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9185 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9186 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9188 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9191 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9192 is BLKmode and bitpos not zero or mode change is needed.
9194 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9197 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9200 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9203 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9204 instead of MULT rtx. Update all corresponding splitters.
9206 (*ssub<modesuffix>): Ditto.
9208 (*cmp_sadd_di): Update split patterns.
9209 (*cmp_sadd_si): Ditto.
9210 (*cmp_sadd_sidi): Ditto.
9211 (*cmp_ssub_di): Ditto.
9212 (*cmp_ssub_si): Ditto.
9213 (*cmp_ssub_sidi): Ditto.
9214 * config/alpha/predicates.md (const23_operand): New predicate.
9215 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9216 Look for ASHIFT, not MULT inner operand.
9217 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9219 2018-01-04 Martin Liska <mliska@suse.cz>
9221 PR gcov-profile/83669
9222 * gcov.c (output_intermediate_file): Add version to intermediate
9224 * doc/gcov.texi: Document new field 'version' in intermediate
9225 file format. Fix location of '-k' option of gcov command.
9227 2018-01-04 Martin Liska <mliska@suse.cz>
9230 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9232 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9234 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9236 2018-01-03 Martin Sebor <msebor@redhat.com>
9238 PR tree-optimization/83655
9239 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9240 checking calls with invalid arguments.
9242 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9244 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9245 (vectorizable_mask_load_store): Delete.
9246 (vectorizable_call): Return false for masked loads and stores.
9247 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9248 instead of gimple_assign_rhs1.
9249 (vectorizable_load): Handle IFN_MASK_LOAD.
9250 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9252 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9254 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9256 (vectorizable_mask_load_store): ...here.
9257 (vectorizable_load): ...and here.
9259 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9261 * tree-vect-stmts.c (vect_build_all_ones_mask)
9262 (vect_build_zero_merge_argument): New functions, split out from...
9263 (vectorizable_load): ...here.
9265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9267 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9269 (vectorizable_mask_load_store): ...here.
9270 (vectorizable_store): ...and here.
9272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9274 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9276 (vectorizable_mask_load_store): ...here.
9278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9280 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9281 (vect_model_store_cost): Take a vec_load_store_type instead of a
9283 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9284 (vect_model_store_cost): Take a vec_load_store_type instead of a
9286 (vectorizable_mask_load_store): Update accordingly.
9287 (vectorizable_store): Likewise.
9288 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9290 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9292 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9293 IFN_MASK_LOAD calls here rather than...
9294 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9297 Alan Hayward <alan.hayward@arm.com>
9298 David Sherwood <david.sherwood@arm.com>
9300 * expmed.c (extract_bit_field_1): For vector extracts,
9301 fall back to extract_bit_field_as_subreg if vec_extract
9304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9305 Alan Hayward <alan.hayward@arm.com>
9306 David Sherwood <david.sherwood@arm.com>
9308 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9309 they are variable or constant sized.
9310 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9311 slots for constant-sized data.
9313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9314 Alan Hayward <alan.hayward@arm.com>
9315 David Sherwood <david.sherwood@arm.com>
9317 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9318 handling COND_EXPRs with boolean comparisons, try to find a better
9319 basis for the mask type than the boolean itself.
9321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9323 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9324 is calculated and how it can be overridden.
9325 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9326 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9328 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9331 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9332 Alan Hayward <alan.hayward@arm.com>
9333 David Sherwood <david.sherwood@arm.com>
9335 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9336 Remove the mode argument.
9337 (aarch64_simd_valid_immediate): Remove the mode and inverse
9339 * config/aarch64/iterators.md (bitsize): New iterator.
9340 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9341 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9342 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9343 aarch64_simd_valid_immediate.
9344 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9345 (aarch64_reg_or_bic_imm): Likewise.
9346 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9347 with an insn_type enum and msl with a modifier_type enum.
9348 Replace element_width with a scalar_mode. Change the shift
9349 to unsigned int. Add constructors for scalar_float_mode and
9350 scalar_int_mode elements.
9351 (aarch64_vect_float_const_representable_p): Delete.
9352 (aarch64_can_const_movi_rtx_p)
9353 (aarch64_simd_scalar_immediate_valid_for_move)
9354 (aarch64_simd_make_constant): Update call to
9355 aarch64_simd_valid_immediate.
9356 (aarch64_advsimd_valid_immediate_hs): New function.
9357 (aarch64_advsimd_valid_immediate): Likewise.
9358 (aarch64_simd_valid_immediate): Remove mode and inverse
9359 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9360 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9361 and aarch64_float_const_representable_p on the result.
9362 (aarch64_output_simd_mov_immediate): Remove mode argument.
9363 Update call to aarch64_simd_valid_immediate and use of
9364 simd_immediate_info.
9365 (aarch64_output_scalar_simd_mov_immediate): Update call
9368 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9369 Alan Hayward <alan.hayward@arm.com>
9370 David Sherwood <david.sherwood@arm.com>
9372 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9373 (mode_nunits): Likewise CONST_MODE_NUNITS.
9374 * machmode.def (ADJUST_NUNITS): Document.
9375 * genmodes.c (mode_data::need_nunits_adj): New field.
9376 (blank_mode): Update accordingly.
9377 (adj_nunits): New variable.
9378 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9380 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9381 listed in adj_nunits.
9382 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9383 listed in adj_nunits. Don't emit case statements for such modes.
9384 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9385 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9386 nothing if adj_nunits is nonnull.
9387 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9388 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9389 (emit_mode_fbit): Update use of print_maybe_const_decl.
9390 (emit_move_size): Likewise. Treat the array as non-const
9392 (emit_mode_adjustments): Handle adj_nunits.
9394 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9396 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9397 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9398 (VECTOR_MODES): Use it.
9399 (make_vector_modes): Take the prefix as an argument.
9401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9402 Alan Hayward <alan.hayward@arm.com>
9403 David Sherwood <david.sherwood@arm.com>
9405 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9406 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9407 for MODE_VECTOR_BOOL.
9408 * machmode.def (VECTOR_BOOL_MODE): Document.
9409 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9410 (make_vector_bool_mode): New function.
9411 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9413 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9414 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9416 * stor-layout.c (int_mode_for_mode): Likewise.
9417 * tree.c (build_vector_type_for_mode): Likewise.
9418 * varasm.c (output_constant_pool_2): Likewise.
9419 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9420 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9421 for MODE_VECTOR_BOOL.
9422 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9423 of mode class checks.
9424 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9425 instead of a list of mode class checks.
9426 (expand_vector_scalar_condition): Likewise.
9427 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9430 Alan Hayward <alan.hayward@arm.com>
9431 David Sherwood <david.sherwood@arm.com>
9433 * machmode.h (mode_size): Change from unsigned short to
9435 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9436 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9437 or if measurement_type is not polynomial.
9438 (fixed_size_mode::includes_p): Check for constant-sized modes.
9439 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9440 return a poly_uint16 rather than an unsigned short.
9441 (emit_mode_size): Change the type of mode_size from unsigned short
9442 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9443 (emit_mode_adjustments): Cope with polynomial vector sizes.
9444 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9446 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9448 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9449 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9450 * caller-save.c (setup_save_areas): Likewise.
9451 (replace_reg_with_saved_mem): Likewise.
9452 * calls.c (emit_library_call_value_1): Likewise.
9453 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9454 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9455 (gen_lowpart_for_combine): Likewise.
9456 * convert.c (convert_to_integer_1): Likewise.
9457 * cse.c (equiv_constant, cse_insn): Likewise.
9458 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9459 (cselib_subst_to_values): Likewise.
9460 * dce.c (word_dce_process_block): Likewise.
9461 * df-problems.c (df_word_lr_mark_ref): Likewise.
9462 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9463 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9464 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9465 (rtl_for_decl_location): Likewise.
9466 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9467 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9468 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9469 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9470 (expand_expr_real_1): Likewise.
9471 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9472 (pad_below): Likewise.
9473 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9474 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9475 * ira.c (get_subreg_tracking_sizes): Likewise.
9476 * ira-build.c (ira_create_allocno_objects): Likewise.
9477 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9478 (ira_sort_regnos_for_alter_reg): Likewise.
9479 * ira-costs.c (record_operand_costs): Likewise.
9480 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9481 (resolve_simple_move): Likewise.
9482 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9483 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9484 (lra_constraints): Likewise.
9485 (CONST_POOL_OK_P): Reject variable-sized modes.
9486 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9487 (add_pseudo_to_slot, lra_spill): Likewise.
9488 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9489 * optabs-query.c (get_best_extraction_insn): Likewise.
9490 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9491 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9492 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9493 * recog.c (offsettable_address_addr_space_p): Likewise.
9494 * regcprop.c (maybe_mode_change): Likewise.
9495 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9496 * regrename.c (build_def_use): Likewise.
9497 * regstat.c (dump_reg_info): Likewise.
9498 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9499 (find_reloads, find_reloads_subreg_address): Likewise.
9500 * reload1.c (eliminate_regs_1): Likewise.
9501 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9502 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9503 (simplify_binary_operation_1, simplify_subreg): Likewise.
9504 * targhooks.c (default_function_arg_padding): Likewise.
9505 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9506 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9507 (verify_gimple_assign_ternary): Likewise.
9508 * tree-inline.c (estimate_move_cost): Likewise.
9509 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9510 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9511 (get_address_cost_ainc): Likewise.
9512 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9513 (vect_supportable_dr_alignment): Likewise.
9514 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9515 (vectorizable_reduction): Likewise.
9516 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9517 (vectorizable_operation, vectorizable_load): Likewise.
9518 * tree.c (build_same_sized_truth_vector_type): Likewise.
9519 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9520 * var-tracking.c (emit_note_insn_var_location): Likewise.
9521 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9522 (ADDR_VEC_ALIGN): Likewise.
9524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9525 Alan Hayward <alan.hayward@arm.com>
9526 David Sherwood <david.sherwood@arm.com>
9528 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9530 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9531 or if measurement_type is polynomial.
9532 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9533 * combine.c (make_extraction): Likewise.
9534 * dse.c (find_shift_sequence): Likewise.
9535 * dwarf2out.c (mem_loc_descriptor): Likewise.
9536 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9537 (extract_bit_field, extract_low_bits): Likewise.
9538 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9539 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9540 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9541 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9542 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9543 * reload.c (find_reloads): Likewise.
9544 * reload1.c (alter_reg): Likewise.
9545 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9546 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9547 * tree-if-conv.c (predicate_mem_writes): Likewise.
9548 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9549 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9550 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9551 * valtrack.c (dead_debug_insert_temp): Likewise.
9552 * varasm.c (mergeable_constant_section): Likewise.
9553 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9555 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9556 Alan Hayward <alan.hayward@arm.com>
9557 David Sherwood <david.sherwood@arm.com>
9559 * expr.c (expand_assignment): Cope with polynomial mode sizes
9560 when assigning to a CONCAT.
9562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9563 Alan Hayward <alan.hayward@arm.com>
9564 David Sherwood <david.sherwood@arm.com>
9566 * machmode.h (mode_precision): Change from unsigned short to
9568 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9570 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9571 or if measurement_type is not polynomial.
9572 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9573 in which the mode is already known to be a scalar_int_mode.
9574 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9575 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9577 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9578 for GET_MODE_PRECISION.
9579 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9580 for GET_MODE_PRECISION.
9581 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9583 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9584 (expand_field_assignment, make_extraction): Likewise.
9585 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9586 (get_last_value): Likewise.
9587 * convert.c (convert_to_integer_1): Likewise.
9588 * cse.c (cse_insn): Likewise.
9589 * expr.c (expand_expr_real_1): Likewise.
9590 * lra-constraints.c (simplify_operand_subreg): Likewise.
9591 * optabs-query.c (can_atomic_load_p): Likewise.
9592 * optabs.c (expand_atomic_load): Likewise.
9593 (expand_atomic_store): Likewise.
9594 * ree.c (combine_reaching_defs): Likewise.
9595 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9596 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9597 * tree.h (type_has_mode_precision_p): Likewise.
9598 * ubsan.c (instrument_si_overflow): Likewise.
9600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9601 Alan Hayward <alan.hayward@arm.com>
9602 David Sherwood <david.sherwood@arm.com>
9604 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9605 polynomial numbers of units.
9606 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9607 (valid_vector_subparts_p): New function.
9608 (build_vector_type): Remove temporary shim and take the number
9609 of units as a poly_uint64 rather than an int.
9610 (build_opaque_vector_type): Take the number of units as a
9611 poly_uint64 rather than an int.
9612 * tree.c (build_vector_from_ctor): Handle polynomial
9613 TYPE_VECTOR_SUBPARTS.
9614 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9615 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9616 (build_vector_from_val): If the number of units is variable,
9617 use build_vec_duplicate_cst for constant operands and
9618 VEC_DUPLICATE_EXPR otherwise.
9619 (make_vector_type): Remove temporary is_constant ().
9620 (build_vector_type, build_opaque_vector_type): Take the number of
9621 units as a poly_uint64 rather than an int.
9622 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9624 * cfgexpand.c (expand_debug_expr): Likewise.
9625 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9626 (store_constructor, expand_expr_real_1): Likewise.
9627 (const_scalar_mask_from_tree): Likewise.
9628 * fold-const-call.c (fold_const_reduction): Likewise.
9629 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9630 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9631 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9632 (fold_relational_const): Likewise.
9633 (native_interpret_vector): Likewise. Change the size from an
9634 int to an unsigned int.
9635 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9636 TYPE_VECTOR_SUBPARTS.
9637 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9638 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9639 duplicating a non-constant operand into a variable-length vector.
9640 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9641 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9642 * ipa-icf.c (sem_variable::equals): Likewise.
9643 * match.pd: Likewise.
9644 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9645 * print-tree.c (print_node): Likewise.
9646 * stor-layout.c (layout_type): Likewise.
9647 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9648 * tree-cfg.c (verify_gimple_comparison): Likewise.
9649 (verify_gimple_assign_binary): Likewise.
9650 (verify_gimple_assign_ternary): Likewise.
9651 (verify_gimple_assign_single): Likewise.
9652 * tree-pretty-print.c (dump_generic_node): Likewise.
9653 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9654 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9655 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9656 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9657 (vect_shift_permute_load_chain): Likewise.
9658 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9659 (expand_vector_condition, optimize_vector_constructor): Likewise.
9660 (lower_vec_perm, get_compute_type): Likewise.
9661 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9662 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9663 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9664 (vect_recog_mask_conversion_pattern): Likewise.
9665 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9666 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9667 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9668 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9669 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9670 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9671 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9672 (supportable_widening_operation): Likewise.
9673 (supportable_narrowing_operation): Likewise.
9674 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9676 * varasm.c (output_constant): Likewise.
9678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9679 Alan Hayward <alan.hayward@arm.com>
9680 David Sherwood <david.sherwood@arm.com>
9682 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9683 so that both the length == 3 and length != 3 cases set up their
9684 own permute vectors. Add comments explaining why we know the
9685 number of elements is constant.
9686 (vect_permute_load_chain): Likewise.
9688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9689 Alan Hayward <alan.hayward@arm.com>
9690 David Sherwood <david.sherwood@arm.com>
9692 * machmode.h (mode_nunits): Change from unsigned char to
9694 (ONLY_FIXED_SIZE_MODES): New macro.
9695 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9696 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9697 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9699 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9700 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9701 or if measurement_type is not polynomial.
9702 * genmodes.c (ZERO_COEFFS): New macro.
9703 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9705 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9706 Use ZERO_COEFFS when emitting initializers.
9707 * data-streamer.h (bp_pack_poly_value): New function.
9708 (bp_unpack_poly_value): Likewise.
9709 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9710 for GET_MODE_NUNITS.
9711 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9712 for GET_MODE_NUNITS.
9713 * tree.c (make_vector_type): Remove temporary shim and make
9714 the real function take the number of units as a poly_uint64
9716 (build_vector_type_for_mode): Handle polynomial nunits.
9717 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9718 * emit-rtl.c (const_vec_series_p_1): Likewise.
9719 (gen_rtx_CONST_VECTOR): Likewise.
9720 * fold-const.c (test_vec_duplicate_folding): Likewise.
9721 * genrecog.c (validate_pattern): Likewise.
9722 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9723 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9724 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9725 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9726 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9727 * rtlanal.c (subreg_get_info): Likewise.
9728 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9729 (vect_grouped_load_supported): Likewise.
9730 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9731 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9732 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9733 (simplify_const_unary_operation, simplify_binary_operation_1)
9734 (simplify_const_binary_operation, simplify_ternary_operation)
9735 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9736 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9737 instead of CONST_VECTOR_NUNITS.
9738 * varasm.c (output_constant_pool_2): Likewise.
9739 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9740 explicit-encoded elements in the XVEC for variable-length vectors.
9742 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9744 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9746 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9747 Alan Hayward <alan.hayward@arm.com>
9748 David Sherwood <david.sherwood@arm.com>
9750 * coretypes.h (fixed_size_mode): Declare.
9751 (fixed_size_mode_pod): New typedef.
9752 * builtins.h (target_builtins::x_apply_args_mode)
9753 (target_builtins::x_apply_result_mode): Change type to
9754 fixed_size_mode_pod.
9755 * builtins.c (apply_args_size, apply_result_size, result_vector)
9756 (expand_builtin_apply_args_1, expand_builtin_apply)
9757 (expand_builtin_return): Update accordingly.
9759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9761 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9762 * cselib.c (cselib_hash_rtx): Likewise.
9763 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9764 CONST_VECTOR encoding.
9766 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9767 Jeff Law <law@redhat.com>
9770 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9771 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9772 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9773 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9776 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9777 explicitly probe *sp in a noreturn function if there were any callee
9778 register saves or frame pointer is needed.
9780 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9783 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9784 BLKmode for ternary, binary or unary expressions.
9787 * var-tracking.c (delete_vta_debug_insn): New inline function.
9788 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9789 insns from get_insns () to NULL instead of each bb separately.
9790 Use delete_vta_debug_insn. No longer static.
9791 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9792 delete_vta_debug_insns callers.
9793 * rtl.h (delete_vta_debug_insns): Declare.
9794 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9795 instead of variable_tracking_main.
9797 2018-01-03 Martin Sebor <msebor@redhat.com>
9799 PR tree-optimization/83603
9800 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9801 arguments past the endof the argument list in functions declared
9802 without a prototype.
9803 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9804 Avoid checking when arguments are null.
9806 2018-01-03 Martin Sebor <msebor@redhat.com>
9809 * doc/extend.texi (attribute const): Fix a typo.
9810 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9811 issuing -Wsuggest-attribute for void functions.
9813 2018-01-03 Martin Sebor <msebor@redhat.com>
9815 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9816 offset_int::from instead of wide_int::to_shwi.
9817 (maybe_diag_overlap): Remove assertion.
9818 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9819 * gimple-ssa-sprintf.c (format_directive): Same.
9820 (parse_directive): Same.
9821 (sprintf_dom_walker::compute_format_length): Same.
9822 (try_substitute_return_value): Same.
9824 2018-01-03 Jeff Law <law@redhat.com>
9827 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9828 non-constant residual for zero at runtime and avoid probing in
9829 that case. Reorganize code for trailing problem to mirror handling
9832 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9834 PR tree-optimization/83501
9835 * tree-ssa-strlen.c (get_string_cst): New.
9836 (handle_char_store): Call get_string_cst.
9838 2018-01-03 Martin Liska <mliska@suse.cz>
9840 PR tree-optimization/83593
9841 * tree-ssa-strlen.c: Include tree-cfg.h.
9842 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9843 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9844 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9846 (strlen_dom_walker::before_dom_children): Call
9847 gimple_purge_dead_eh_edges. Dump tranformation with details
9849 (strlen_dom_walker::before_dom_children): Update call by adding
9850 new argument cleanup_eh.
9851 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9853 2018-01-03 Martin Liska <mliska@suse.cz>
9856 * cif-code.def (VARIADIC_THUNK): New enum value.
9857 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9860 2018-01-03 Jan Beulich <jbeulich@suse.com>
9862 * sse.md (mov<mode>_internal): Tighten condition for when to use
9863 vmovdqu<ssescalarsize> for TI and OI modes.
9865 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9867 Update copyright years.
9869 2018-01-03 Martin Liska <mliska@suse.cz>
9872 * ipa-visibility.c (function_and_variable_visibility): Skip
9873 functions with noipa attribure.
9875 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9877 * gcc.c (process_command): Update copyright notice dates.
9878 * gcov-dump.c (print_version): Ditto.
9879 * gcov.c (print_version): Ditto.
9880 * gcov-tool.c (print_version): Ditto.
9881 * gengtype.c (create_file): Ditto.
9882 * doc/cpp.texi: Bump @copying's copyright year.
9883 * doc/cppinternals.texi: Ditto.
9884 * doc/gcc.texi: Ditto.
9885 * doc/gccint.texi: Ditto.
9886 * doc/gcov.texi: Ditto.
9887 * doc/install.texi: Ditto.
9888 * doc/invoke.texi: Ditto.
9890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9892 * vector-builder.h (vector_builder::m_full_nelts): Change from
9893 unsigned int to poly_uint64.
9894 (vector_builder::full_nelts): Update prototype accordingly.
9895 (vector_builder::new_vector): Likewise.
9896 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9897 (vector_builder::operator ==): Likewise.
9898 (vector_builder::finalize): Likewise.
9899 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9900 Take the number of elements as a poly_uint64 rather than an
9902 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9903 from unsigned int to poly_uint64.
9904 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9905 (vec_perm_indices::new_vector): Likewise.
9906 (vec_perm_indices::length): Likewise.
9907 (vec_perm_indices::nelts_per_input): Likewise.
9908 (vec_perm_indices::input_nelts): Likewise.
9909 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9910 number of elements per input as a poly_uint64 rather than an
9911 unsigned int. Use the original encoding for variable-length
9912 vectors, rather than clamping each individual element.
9913 For the second and subsequent elements in each pattern,
9914 clamp the step and base before clamping their sum.
9915 (vec_perm_indices::series_p): Handle polynomial element counts.
9916 (vec_perm_indices::all_in_range_p): Likewise.
9917 (vec_perm_indices_to_tree): Likewise.
9918 (vec_perm_indices_to_rtx): Likewise.
9919 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9920 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9921 (tree_vector_builder::new_binary_operation): Handle polynomial
9922 element counts. Return false if we need to know the number
9923 of elements at compile time.
9924 * fold-const.c (fold_vec_perm): Punt if the number of elements
9925 isn't known at compile time.
9927 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9929 * vec-perm-indices.h (vec_perm_builder): Change element type
9930 from HOST_WIDE_INT to poly_int64.
9931 (vec_perm_indices::element_type): Update accordingly.
9932 (vec_perm_indices::clamp): Handle polynomial element_types.
9933 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9934 (vec_perm_indices::all_in_range_p): Likewise.
9935 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9937 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9938 polynomial vec_perm_indices element types.
9939 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9940 * fold-const.c (fold_vec_perm): Likewise.
9941 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9942 * tree-vect-generic.c (lower_vec_perm): Likewise.
9943 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9944 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9945 element type to HOST_WIDE_INT.
9947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9948 Alan Hayward <alan.hayward@arm.com>
9949 David Sherwood <david.sherwood@arm.com>
9951 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9952 rather than an int. Use plus_constant.
9953 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9954 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9957 Alan Hayward <alan.hayward@arm.com>
9958 David Sherwood <david.sherwood@arm.com>
9960 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9961 a HOST_WIDE_INT to a poly_int64.
9963 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9964 Alan Hayward <alan.hayward@arm.com>
9965 David Sherwood <david.sherwood@arm.com>
9967 * calls.c (load_register_parameters): Cope with polynomial
9968 mode sizes. Require a constant size for BLKmode parameters
9969 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9970 forces a parameter to be padded at the lsb end in order to
9971 fill a complete number of words, require the parameter size
9972 to be ordered wrt UNITS_PER_WORD.
9974 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9975 Alan Hayward <alan.hayward@arm.com>
9976 David Sherwood <david.sherwood@arm.com>
9978 * reload1.c (spill_stack_slot_width): Change element type
9979 from unsigned int to poly_uint64_pod.
9980 (alter_reg): Treat mode sizes as polynomial.
9982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9983 Alan Hayward <alan.hayward@arm.com>
9984 David Sherwood <david.sherwood@arm.com>
9986 * reload.c (complex_word_subreg_p): New function.
9987 (reload_inner_reg_of_subreg, push_reload): Use it.
9989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9990 Alan Hayward <alan.hayward@arm.com>
9991 David Sherwood <david.sherwood@arm.com>
9993 * lra-constraints.c (process_alt_operands): Reject matched
9994 operands whose sizes aren't ordered.
9995 (match_reload): Refer to this check here.
9997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9998 Alan Hayward <alan.hayward@arm.com>
9999 David Sherwood <david.sherwood@arm.com>
10001 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
10002 that the mode size is in the set {1, 2, 4, 8, 16}.
10004 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10005 Alan Hayward <alan.hayward@arm.com>
10006 David Sherwood <david.sherwood@arm.com>
10008 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
10009 Use plus_constant instead of gen_rtx_PLUS.
10011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10012 Alan Hayward <alan.hayward@arm.com>
10013 David Sherwood <david.sherwood@arm.com>
10015 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
10016 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
10017 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
10018 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
10019 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
10020 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
10021 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
10022 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
10023 * config/i386/i386.c (ix86_push_rounding): ...this new function.
10024 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
10026 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
10027 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
10028 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
10029 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
10030 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
10031 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
10032 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
10033 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
10034 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
10035 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
10037 * expr.c (emit_move_resolve_push): Treat the input and result
10038 of PUSH_ROUNDING as a poly_int64.
10039 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
10040 (emit_push_insn): Likewise.
10041 * lra-eliminations.c (mark_not_eliminable): Likewise.
10042 * recog.c (push_operand): Likewise.
10043 * reload1.c (elimination_effects): Likewise.
10044 * rtlanal.c (nonzero_bits1): Likewise.
10045 * calls.c (store_one_arg): Likewise. Require the padding to be
10046 known at compile time.
10048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10049 Alan Hayward <alan.hayward@arm.com>
10050 David Sherwood <david.sherwood@arm.com>
10052 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
10053 Use plus_constant instead of gen_rtx_PLUS.
10055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10056 Alan Hayward <alan.hayward@arm.com>
10057 David Sherwood <david.sherwood@arm.com>
10059 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
10060 rather than an int.
10062 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10063 Alan Hayward <alan.hayward@arm.com>
10064 David Sherwood <david.sherwood@arm.com>
10066 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
10067 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
10068 via stack temporaries. Treat the mode size as polynomial too.
10070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10071 Alan Hayward <alan.hayward@arm.com>
10072 David Sherwood <david.sherwood@arm.com>
10074 * expr.c (expand_expr_real_2): When handling conversions involving
10075 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
10076 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
10077 as a poly_uint64 too.
10079 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10080 Alan Hayward <alan.hayward@arm.com>
10081 David Sherwood <david.sherwood@arm.com>
10083 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
10085 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10086 Alan Hayward <alan.hayward@arm.com>
10087 David Sherwood <david.sherwood@arm.com>
10089 * combine.c (can_change_dest_mode): Handle polynomial
10090 REGMODE_NATURAL_SIZE.
10091 * expmed.c (store_bit_field_1): Likewise.
10092 * expr.c (store_constructor): Likewise.
10093 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10094 and polynomial REGMODE_NATURAL_SIZE.
10095 (gen_lowpart_common): Likewise.
10096 * reginfo.c (record_subregs_of_mode): Likewise.
10097 * rtlanal.c (read_modify_subreg_p): Likewise.
10099 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10100 Alan Hayward <alan.hayward@arm.com>
10101 David Sherwood <david.sherwood@arm.com>
10103 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
10104 numbers of elements.
10106 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10107 Alan Hayward <alan.hayward@arm.com>
10108 David Sherwood <david.sherwood@arm.com>
10110 * match.pd: Cope with polynomial numbers of vector elements.
10112 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10113 Alan Hayward <alan.hayward@arm.com>
10114 David Sherwood <david.sherwood@arm.com>
10116 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
10117 in a POINTER_PLUS_EXPR.
10119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10120 Alan Hayward <alan.hayward@arm.com>
10121 David Sherwood <david.sherwood@arm.com>
10123 * omp-simd-clone.c (simd_clone_subparts): New function.
10124 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10125 (ipa_simd_modify_function_body): Likewise.
10127 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10128 Alan Hayward <alan.hayward@arm.com>
10129 David Sherwood <david.sherwood@arm.com>
10131 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10132 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10133 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10134 (expand_vector_condition, vector_element): Likewise.
10135 (subparts_gt): New function.
10136 (get_compute_type): Use subparts_gt.
10137 (count_type_subparts): Delete.
10138 (expand_vector_operations_1): Use subparts_gt instead of
10139 count_type_subparts.
10141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10142 Alan Hayward <alan.hayward@arm.com>
10143 David Sherwood <david.sherwood@arm.com>
10145 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10146 (vect_compile_time_alias): ...this new function. Do the calculation
10147 on poly_ints rather than trees.
10148 (vect_prune_runtime_alias_test_list): Update call accordingly.
10150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10151 Alan Hayward <alan.hayward@arm.com>
10152 David Sherwood <david.sherwood@arm.com>
10154 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10156 (vect_schedule_slp_instance): Likewise.
10158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10159 Alan Hayward <alan.hayward@arm.com>
10160 David Sherwood <david.sherwood@arm.com>
10162 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10163 constant and extern definitions for variable-length vectors.
10164 (vect_get_constant_vectors): Note that the number of units
10165 is known to be constant.
10167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10168 Alan Hayward <alan.hayward@arm.com>
10169 David Sherwood <david.sherwood@arm.com>
10171 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10172 of units as polynomial. Choose between WIDE and NARROW based
10175 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10176 Alan Hayward <alan.hayward@arm.com>
10177 David Sherwood <david.sherwood@arm.com>
10179 * tree-vect-stmts.c (simd_clone_subparts): New function.
10180 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10183 Alan Hayward <alan.hayward@arm.com>
10184 David Sherwood <david.sherwood@arm.com>
10186 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10187 vectors as polynomial. Use build_index_vector for
10188 IFN_GOMP_SIMD_LANE.
10190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10191 Alan Hayward <alan.hayward@arm.com>
10192 David Sherwood <david.sherwood@arm.com>
10194 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10195 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10196 for variable-length vectors.
10197 (vectorizable_mask_load_store): Treat the number of units as
10198 polynomial, asserting that it is constant if the condition has
10199 already been enforced.
10200 (vectorizable_store, vectorizable_load): Likewise.
10202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10203 Alan Hayward <alan.hayward@arm.com>
10204 David Sherwood <david.sherwood@arm.com>
10206 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10207 of units as polynomial. Punt if we can't tell at compile time
10208 which vector contains the final result.
10210 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10211 Alan Hayward <alan.hayward@arm.com>
10212 David Sherwood <david.sherwood@arm.com>
10214 * tree-vect-loop.c (vectorizable_induction): Treat the number
10215 of units as polynomial. Punt on SLP inductions. Use an integer
10216 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10217 cast of such a series for variable-length floating-point
10220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10221 Alan Hayward <alan.hayward@arm.com>
10222 David Sherwood <david.sherwood@arm.com>
10224 * tree.h (build_index_vector): Declare.
10225 * tree.c (build_index_vector): New function.
10226 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10227 of units as polynomial, forcibly converting it to a constant if
10228 vectorizable_reduction has already enforced the condition.
10229 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10230 to create a {1,2,3,...} vector.
10231 (vectorizable_reduction): Treat the number of units as polynomial.
10232 Choose vectype_in based on the largest scalar element size rather
10233 than the smallest number of units. Enforce the restrictions
10236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10237 Alan Hayward <alan.hayward@arm.com>
10238 David Sherwood <david.sherwood@arm.com>
10240 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10241 number of units as polynomial.
10243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10244 Alan Hayward <alan.hayward@arm.com>
10245 David Sherwood <david.sherwood@arm.com>
10247 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10248 * target.def (autovectorize_vector_sizes): Return the vector sizes
10249 by pointer, using vector_sizes rather than a bitmask.
10250 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10251 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10252 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10254 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10255 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10256 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10257 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10258 * omp-general.c (omp_max_vf): Likewise.
10259 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10260 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10261 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10262 * tree-vect-slp.c (vect_slp_bb): Likewise.
10263 * doc/tm.texi: Regenerate.
10264 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10266 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10267 the vector size as a poly_uint64 rather than an unsigned int.
10268 (current_vector_size): Change from an unsigned int to a poly_uint64.
10269 (get_vectype_for_scalar_type): Update accordingly.
10270 * tree.h (build_truth_vector_type): Take the size and number of
10271 units as a poly_uint64 rather than an unsigned int.
10272 (build_vector_type): Add a temporary overload that takes
10273 the number of units as a poly_uint64 rather than an unsigned int.
10274 * tree.c (make_vector_type): Likewise.
10275 (build_truth_vector_type): Take the number of units as a poly_uint64
10276 rather than an unsigned int.
10278 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10279 Alan Hayward <alan.hayward@arm.com>
10280 David Sherwood <david.sherwood@arm.com>
10282 * target.def (get_mask_mode): Take the number of units and length
10283 as poly_uint64s rather than unsigned ints.
10284 * targhooks.h (default_get_mask_mode): Update accordingly.
10285 * targhooks.c (default_get_mask_mode): Likewise.
10286 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10287 * doc/tm.texi: Regenerate.
10289 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10290 Alan Hayward <alan.hayward@arm.com>
10291 David Sherwood <david.sherwood@arm.com>
10293 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10294 * omp-general.c (omp_max_vf): Likewise.
10295 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10296 (expand_omp_simd): Handle polynomial safelen.
10297 * omp-low.c (omplow_simd_context): Add a default constructor.
10298 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10299 (lower_rec_simd_input_clauses): Update accordingly.
10300 (lower_rec_input_clauses): Likewise.
10302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10303 Alan Hayward <alan.hayward@arm.com>
10304 David Sherwood <david.sherwood@arm.com>
10306 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10307 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10308 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10309 (vect_analyze_slp_cost): Likewise.
10310 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10311 (vect_model_load_cost): Likewise.
10313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10314 Alan Hayward <alan.hayward@arm.com>
10315 David Sherwood <david.sherwood@arm.com>
10317 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10318 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10319 from an unsigned int * to a poly_uint64_pod *.
10320 (calculate_unrolling_factor): New function.
10321 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10324 Alan Hayward <alan.hayward@arm.com>
10325 David Sherwood <david.sherwood@arm.com>
10327 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10328 from an unsigned int to a poly_uint64.
10329 (_loop_vec_info::slp_unrolling_factor): Likewise.
10330 (_loop_vec_info::vectorization_factor): Change from an int
10332 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10333 (vect_get_num_vectors): New function.
10334 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10335 (vect_get_num_copies): Use vect_get_num_vectors.
10336 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10337 to an unsigned int *.
10338 (vect_analyze_data_refs): Change min_vf from an int * to a
10340 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10341 than an unsigned HOST_WIDE_INT.
10342 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10343 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10344 to an unsigned int *.
10345 (vect_analyze_data_ref_dependences): Likewise.
10346 (vect_compute_data_ref_alignment): Handle polynomial vf.
10347 (vect_enhance_data_refs_alignment): Likewise.
10348 (vect_prune_runtime_alias_test_list): Likewise.
10349 (vect_shift_permute_load_chain): Likewise.
10350 (vect_supportable_dr_alignment): Likewise.
10351 (dependence_distance_ge_vf): Take the vectorization factor as a
10352 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10353 (vect_analyze_data_refs): Change min_vf from an int * to a
10355 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10356 vfm1 as a poly_uint64 rather than an int. Make the same change
10357 for the returned bound_scalar.
10358 (vect_gen_vector_loop_niters): Handle polynomial vf.
10359 (vect_do_peeling): Likewise. Update call to
10360 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10361 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10363 * tree-vect-loop.c (vect_determine_vectorization_factor)
10364 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10365 (vect_get_known_peeling_cost): Likewise.
10366 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10367 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10368 (vect_transform_loop): Likewise. Use the lowest possible VF when
10369 updating the upper bounds of the loop.
10370 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10371 rather than an int.
10372 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10373 polynomial unroll factors.
10374 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10375 (vect_make_slp_decision): Likewise.
10376 (vect_supported_load_permutation_p): Likewise, and polynomial
10378 (vect_analyze_slp_cost): Handle polynomial vf.
10379 (vect_slp_analyze_node_operations): Likewise.
10380 (vect_slp_analyze_bb_1): Likewise.
10381 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10382 than an unsigned HOST_WIDE_INT.
10383 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10384 (vectorizable_load): Handle polynomial vf.
10385 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10387 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10390 Alan Hayward <alan.hayward@arm.com>
10391 David Sherwood <david.sherwood@arm.com>
10393 * match.pd: Handle bit operations involving three constants
10394 and try to fold one pair.
10396 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10398 * tree-vect-loop-manip.c: Include gimple-fold.h.
10399 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10400 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10401 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10402 Add a path that uses a step of VF instead of 1, but disable it
10404 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10405 and niters_no_overflow parameters. Update calls to
10406 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10407 Create a new SSA name if the latter choses to use a ste other
10408 than zero, and return it via niters_vector_mult_vf_var.
10409 * tree-vect-loop.c (vect_transform_loop): Update calls to
10410 vect_do_peeling, vect_gen_vector_loop_niters and
10411 slpeel_make_loop_iterate_ntimes.
10412 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10413 (vect_gen_vector_loop_niters): Update declarations after above changes.
10415 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10417 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10418 128-bit round to integer instructions.
10419 (ceil<mode>2): Likewise.
10420 (btrunc<mode>2): Likewise.
10421 (round<mode>2): Likewise.
10423 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10425 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10426 unaligned VSX load/store on P8/P9.
10427 (expand_block_clear): Allow the use of unaligned VSX
10428 load/store on P8/P9.
10430 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10432 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10434 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10435 swap associated with both a load and a store.
10437 2018-01-02 Andrew Waterman <andrew@sifive.com>
10439 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10440 * config/riscv/riscv.md (clear_cache): Use it.
10442 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10444 * web.c: Remove out-of-date comment.
10446 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10448 * expr.c (fixup_args_size_notes): Check that any existing
10449 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10450 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10451 (emit_single_push_insn): ...here.
10453 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10455 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10456 (const_vector_encoded_nelts): New function.
10457 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10458 (const_vector_int_elt, const_vector_elt): Declare.
10459 * emit-rtl.c (const_vector_int_elt_1): New function.
10460 (const_vector_elt): Likewise.
10461 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10462 of CONST_VECTOR_ELT.
10464 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10466 * expr.c: Include rtx-vector-builder.h.
10467 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10468 directly on the tree encoding.
10469 (const_vector_from_tree): Likewise.
10470 * optabs.c: Include rtx-vector-builder.h.
10471 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10472 sequence of "u" values.
10473 * vec-perm-indices.c: Include rtx-vector-builder.h.
10474 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10475 directly on the vec_perm_indices encoding.
10477 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10479 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10480 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10481 * rtx-vector-builder.h: New file.
10482 * rtx-vector-builder.c: Likewise.
10483 * rtl.h (rtx_def::u2): Add a const_vector field.
10484 (CONST_VECTOR_NPATTERNS): New macro.
10485 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10486 (CONST_VECTOR_DUPLICATE_P): Likewise.
10487 (CONST_VECTOR_STEPPED_P): Likewise.
10488 (CONST_VECTOR_ENCODED_ELT): Likewise.
10489 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10490 (unwrap_const_vec_duplicate): Likewise.
10491 (const_vec_series_p): Check for a non-duplicated vector encoding.
10492 Say that the function only returns true for integer vectors.
10493 * emit-rtl.c: Include rtx-vector-builder.h.
10494 (gen_const_vec_duplicate_1): Delete.
10495 (gen_const_vector): Call gen_const_vec_duplicate instead of
10496 gen_const_vec_duplicate_1.
10497 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10498 (gen_const_vec_duplicate): Use rtx_vector_builder.
10499 (gen_const_vec_series): Likewise.
10500 (gen_rtx_CONST_VECTOR): Likewise.
10501 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10502 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10503 Build a new vector rather than modifying a CONST_VECTOR in-place.
10504 (handle_special_swappables): Update call accordingly.
10505 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10506 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10507 Build a new vector rather than modifying a CONST_VECTOR in-place.
10508 (handle_special_swappables): Update call accordingly.
10510 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10512 * simplify-rtx.c (simplify_const_binary_operation): Use
10513 CONST_VECTOR_ELT instead of XVECEXP.
10515 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10517 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10518 the selector elements to be different from the data elements
10519 if the selector is a VECTOR_CST.
10520 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10521 ssizetype for the selector.
10523 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10525 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10526 before testing each element individually.
10527 * tree-vect-generic.c (lower_vec_perm): Likewise.
10529 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10531 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10532 * selftest-run-tests.c (selftest::run_tests): Call it.
10533 * vector-builder.h (vector_builder::operator ==): New function.
10534 (vector_builder::operator !=): Likewise.
10535 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10536 (vec_perm_indices::all_from_input_p): New function.
10537 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10538 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10539 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10540 instead of reading the VECTOR_CST directly. Detect whether both
10541 vector inputs are the same before constructing the vec_perm_indices,
10542 and update the number of inputs argument accordingly. Use the
10543 utility functions added above. Only construct sel2 if we need to.
10545 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10547 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10548 the broadcast of the low byte.
10549 (expand_mult_highpart): Use an explicit encoding for the permutes.
10550 * optabs-query.c (can_mult_highpart_p): Likewise.
10551 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10552 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10553 (vectorizable_bswap): Likewise.
10554 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10555 explicit encoding for the power-of-2 permutes.
10556 (vect_permute_store_chain): Likewise.
10557 (vect_grouped_load_supported): Likewise.
10558 (vect_permute_load_chain): Likewise.
10560 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10562 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10563 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10564 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10565 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10566 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10567 (vect_gen_perm_mask_any): Likewise.
10569 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10571 * int-vector-builder.h: New file.
10572 * vec-perm-indices.h: Include int-vector-builder.h.
10573 (vec_perm_indices): Redefine as an int_vector_builder.
10574 (auto_vec_perm_indices): Delete.
10575 (vec_perm_builder): Redefine as a stand-alone class.
10576 (vec_perm_indices::vec_perm_indices): New function.
10577 (vec_perm_indices::clamp): Likewise.
10578 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10579 (vec_perm_indices::new_vector): New function.
10580 (vec_perm_indices::new_expanded_vector): Update for new
10581 vec_perm_indices class.
10582 (vec_perm_indices::rotate_inputs): New function.
10583 (vec_perm_indices::all_in_range_p): Operate directly on the
10584 encoded form, without computing elided elements.
10585 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10586 encoding. Update for new vec_perm_indices class.
10587 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10588 the given vec_perm_builder.
10589 (expand_vec_perm_var): Update vec_perm_builder constructor.
10590 (expand_mult_highpart): Use vec_perm_builder instead of
10591 auto_vec_perm_indices.
10592 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10593 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10594 or double series encoding as appropriate.
10595 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10596 vec_perm_indices instead of auto_vec_perm_indices.
10597 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10598 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10599 (vect_permute_store_chain): Likewise.
10600 (vect_grouped_load_supported): Likewise.
10601 (vect_permute_load_chain): Likewise.
10602 (vect_shift_permute_load_chain): Likewise.
10603 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10604 (vect_transform_slp_perm_load): Likewise.
10605 (vect_schedule_slp_instance): Likewise.
10606 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10607 (vectorizable_mask_load_store): Likewise.
10608 (vectorizable_bswap): Likewise.
10609 (vectorizable_store): Likewise.
10610 (vectorizable_load): Likewise.
10611 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10612 vec_perm_indices instead of auto_vec_perm_indices. Use
10613 tree_to_vec_perm_builder to read the vector from a tree.
10614 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10615 vec_perm_builder instead of a vec_perm_indices.
10616 (have_whole_vector_shift): Use vec_perm_builder and
10617 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10618 truncation to calc_vec_perm_mask_for_shift.
10619 (vect_create_epilog_for_reduction): Likewise.
10620 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10621 from auto_vec_perm_indices to vec_perm_indices.
10622 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10623 instead of changing individual elements.
10624 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10625 the vector in d.perm.
10626 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10627 from auto_vec_perm_indices to vec_perm_indices.
10628 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10629 instead of changing individual elements.
10630 (arm_vectorize_vec_perm_const): Use new_vector to install
10631 the vector in d.perm.
10632 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10633 Update vec_perm_builder constructor.
10634 (rs6000_expand_interleave): Likewise.
10635 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10636 (rs6000_expand_interleave): Likewise.
10638 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10640 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10641 to qimode could truncate the indices.
10642 * optabs.c (expand_vec_perm_var): Likewise.
10644 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10646 * Makefile.in (OBJS): Add vec-perm-indices.o.
10647 * vec-perm-indices.h: New file.
10648 * vec-perm-indices.c: Likewise.
10649 * target.h (vec_perm_indices): Replace with a forward class
10651 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10652 * optabs.h: Include vec-perm-indices.h.
10653 (expand_vec_perm): Delete.
10654 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10655 (expand_vec_perm_const): Declare.
10656 * target.def (vec_perm_const_ok): Replace with...
10657 (vec_perm_const): ...this new hook.
10658 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10659 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10660 * doc/tm.texi: Regenerate.
10661 * optabs.def (vec_perm_const): Delete.
10662 * doc/md.texi (vec_perm_const): Likewise.
10663 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10664 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10665 expand_vec_perm for constant permutation vectors. Assert that
10666 the mode of variable permutation vectors is the integer equivalent
10667 of the mode that is being permuted.
10668 * optabs-query.h (selector_fits_mode_p): Declare.
10669 * optabs-query.c: Include vec-perm-indices.h.
10670 (selector_fits_mode_p): New function.
10671 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10672 is defined, instead of checking whether the vec_perm_const_optab
10673 exists. Use targetm.vectorize.vec_perm_const instead of
10674 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10675 fit in the vector mode before using a variable permute.
10676 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10677 vec_perm_indices instead of an rtx.
10678 (expand_vec_perm): Replace with...
10679 (expand_vec_perm_const): ...this new function. Take the selector
10680 as a vec_perm_indices rather than an rtx. Also take the mode of
10681 the selector. Update call to shift_amt_for_vec_perm_mask.
10682 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10683 Use vec_perm_indices::new_expanded_vector to expand the original
10684 selector into bytes. Check whether the indices fit in the vector
10685 mode before using a variable permute.
10686 (expand_vec_perm_var): Make global.
10687 (expand_mult_highpart): Use expand_vec_perm_const.
10688 * fold-const.c: Includes vec-perm-indices.h.
10689 * tree-ssa-forwprop.c: Likewise.
10690 * tree-vect-data-refs.c: Likewise.
10691 * tree-vect-generic.c: Likewise.
10692 * tree-vect-loop.c: Likewise.
10693 * tree-vect-slp.c: Likewise.
10694 * tree-vect-stmts.c: Likewise.
10695 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10697 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10698 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10699 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10700 (aarch64_vectorize_vec_perm_const): ...this new function.
10701 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10702 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10703 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10704 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10705 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10706 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10707 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10709 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10710 check for NEON modes.
10711 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10712 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10713 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10714 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10716 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10717 the old VEC_PERM_CONST conditions.
10718 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10719 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10720 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10721 (ia64_vectorize_vec_perm_const_ok): Merge into...
10722 (ia64_vectorize_vec_perm_const): ...this new function.
10723 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10724 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10725 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10726 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10727 * config/mips/mips.c (mips_expand_vec_perm_const)
10728 (mips_vectorize_vec_perm_const_ok): Merge into...
10729 (mips_vectorize_vec_perm_const): ...this new function.
10730 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10731 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10732 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10733 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10734 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10735 (rs6000_expand_vec_perm_const): Delete.
10736 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10738 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10739 (altivec_expand_vec_perm_const_le): Take each operand individually.
10740 Operate on constant selectors rather than rtxes.
10741 (altivec_expand_vec_perm_const): Likewise. Update call to
10742 altivec_expand_vec_perm_const_le.
10743 (rs6000_expand_vec_perm_const): Delete.
10744 (rs6000_vectorize_vec_perm_const_ok): Delete.
10745 (rs6000_vectorize_vec_perm_const): New function.
10746 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10747 an element count and rtx array.
10748 (rs6000_expand_extract_even): Update call accordingly.
10749 (rs6000_expand_interleave): Likewise.
10750 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10751 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10752 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10753 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10754 (rs6000_expand_vec_perm_const): Delete.
10755 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10756 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10757 (altivec_expand_vec_perm_const_le): Take each operand individually.
10758 Operate on constant selectors rather than rtxes.
10759 (altivec_expand_vec_perm_const): Likewise. Update call to
10760 altivec_expand_vec_perm_const_le.
10761 (rs6000_expand_vec_perm_const): Delete.
10762 (rs6000_vectorize_vec_perm_const_ok): Delete.
10763 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10764 reference to the SPE evmerge intructions.
10765 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10766 an element count and rtx array.
10767 (rs6000_expand_extract_even): Update call accordingly.
10768 (rs6000_expand_interleave): Likewise.
10769 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10770 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10772 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10774 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10776 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10777 vector mode and that that mode matches the mode of the data
10779 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10780 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10781 directly using expand_vec_perm_1 when forcing selectors into
10783 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10785 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10787 * optabs-query.h (can_vec_perm_p): Delete.
10788 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10789 * optabs-query.c (can_vec_perm_p): Split into...
10790 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10791 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10792 particular selector is valid.
10793 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10794 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10795 (vect_grouped_load_supported): Likewise.
10796 (vect_shift_permute_load_chain): Likewise.
10797 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10798 (vect_transform_slp_perm_load): Likewise.
10799 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10800 (vectorizable_bswap): Likewise.
10801 (vect_gen_perm_mask_checked): Likewise.
10802 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10803 implementations of variable permutation vectors into account
10804 when deciding which selector to use.
10805 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10806 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10807 with a false third argument.
10808 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10809 to test whether the constant selector is valid and can_vec_perm_var_p
10810 to test whether a variable selector is valid.
10812 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10814 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10815 * optabs-query.c (can_vec_perm_p): Likewise.
10816 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10817 instead of vec_perm_indices.
10818 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10819 (vect_gen_perm_mask_checked): Likewise,
10820 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10821 (vect_gen_perm_mask_checked): Likewise,
10823 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10825 * optabs-query.h (qimode_for_vec_perm): Declare.
10826 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10827 (qimode_for_vec_perm): ...this new function.
10828 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10830 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10832 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10833 does not have a conditional at the top.
10835 2018-01-02 Richard Biener <rguenther@suse.de>
10837 * ipa-inline.c (big_speedup_p): Fix expression.
10839 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10842 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10845 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10849 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10850 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10851 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10852 cond_taken_branch_cost 3->4.
10854 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10856 PR tree-optimization/83581
10857 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10858 TODO_cleanup_cfg if any changes have been made.
10860 PR middle-end/83608
10861 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10862 convert_modes if target mode has the right side, but different mode
10865 PR middle-end/83609
10866 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10867 last argument when extracting from CONCAT. If either from_real or
10868 from_imag is NULL, use expansion through memory. If result is not
10869 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10870 the parts directly to inner mode, if even that fails, use expansion
10873 PR middle-end/83623
10874 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10875 check for bswap in mode rather than HImode and use that in expand_unop
10878 Copyright (C) 2018 Free Software Foundation, Inc.
10880 Copying and distribution of this file, with or without modification,
10881 are permitted in any medium without royalty provided the copyright
10882 notice and this notice are preserved.