1 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
4 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
5 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
8 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
11 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
12 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
13 Ensure negative shifts result in {0}.
15 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
17 PR rtl-optimization/79916
18 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
19 regs (if any) to define how to gnerate SD moves when LRA is in
22 2018-04-13 Jakub Jelinek <jakub@redhat.com>
24 PR rtl-optimization/85393
25 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
26 * except.c (expand_dw2_landing_pad_for_region): Make static.
27 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
28 a label and unconditional jump to old_bb, rather than
29 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
32 PR rtl-optimization/85376
33 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
34 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
35 instead of a specific value.
37 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
38 Bin Cheng <bin.cheng@arm.com>
40 PR tree-optimization/82965
41 PR tree-optimization/83991
42 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
43 by_profile_only parameter.
44 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
45 information if the loop was predicted to iterate too many times.
46 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
48 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
51 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
54 2018-04-13 Martin Liska <mliska@suse.cz>
55 Jakub Jelinek <jakub@redhat.com>
58 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
59 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
60 * builtins.c (expand_builtin_memory_copy_args): Use
61 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
62 handle dest_addr == pc_rtx.
64 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
67 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
68 asked to not generate direct moves.
69 (fix_trunc<mode>si2_stfiwx): Similar.
70 (fix_trunc<mode>si2_internal): Similar.
72 2018-04-12 Jakub Jelinek <jakub@redhat.com>
75 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
76 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
77 lookup if dest in some wider mode is known to be const0_rtx and
78 if so, record permanent equivalence for it to be ZERO_EXTEND of
79 the narrower mode destination.
81 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
83 * lto-streamer-out.c (output_function): Revert 259346.
84 * omp-expand.c (expand_oacc_for): Likewise.
86 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
88 PR rtl-optimization/85354
89 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
90 * sel-sched.c (sel_global_init): ... here.
92 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
95 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
96 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
97 mode for PE-COFF targets.
98 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
99 (i386_pe_asm_lto_end): Likewise.
100 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
101 (TARGET_ASM_LTO_END): Likewise.
102 * config/i386/winnt.c (saved_debug_info_level): New static variable.
103 (i386_pe_asm_lto_start): New function.
104 (i386_pe_asm_lto_end): Likewise.
106 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
107 Richard Biener <rguenther@suse.de>
110 * lto-streamer-out.c (output_function): Fix CFG loop state before
112 * omp-expand.c (expand_oacc_for): Handle calls to internal
113 functions like regular functions.
115 2018-04-12 Richard Biener <rguenther@suse.de>
118 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
119 for the early LTO debug to properly generate references to it
120 during DIE emission. Do not re-use that for the skeleton for
122 (dwarf2out_early_finish): Likewise.
124 2018-04-12 Jakub Jelinek <jakub@redhat.com>
128 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
129 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
130 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
131 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
132 and output is a reg, avoid creating invalid lowpart subreg, but
133 instead split into a 512-bit move. Don't split if not AVX512VL,
134 input is xmm16+ reg and output is a mem.
135 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
136 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
137 xmm16+ reg and output is a mem.
139 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
141 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
142 also for flag_dwarf2_cfi_asm.
144 2018-04-12 Jakub Jelinek <jakub@redhat.com>
146 PR rtl-optimization/85342
147 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
148 a bool scalar var inside of the loop instead. Don't try to update
149 recog_data.operand after failed apply_change_group.
151 2018-04-12 Tom de Vries <tom@codesourcery.com>
154 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
155 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
156 array with flexible array member as array without given dimension.
157 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
158 argument for undefined param to true.
160 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
163 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
164 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
165 from PowerPC section.
166 * config/rs6000/sysv4.opt (mcall-): Improve help text.
167 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
168 help text that is too long.
169 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
170 help text that is too long.
171 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
172 help text that is too long.
174 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
176 * config/alpha/alpha.md (stack_probe_internal): Rename
177 from "probe_stack". Update all callers.
179 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
181 PR rtl-optimization/84566
182 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
183 sched_macro_fuse_insns.
185 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
188 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
189 (compute_block_dependences): ... from here.
191 2018-04-11 Jakub Jelinek <jakub@redhat.com>
193 PR tree-optimization/85331
194 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
195 from int to HOST_WIDE_INT.
197 2018-04-11 Martin Jambor <mjambor@suse.cz>
200 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
201 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
202 not the same as the source val.
203 (cgraph_edge_brings_value_p): New parameter.
204 (gather_edges_for_value): Pass destination value to
205 cgraph_edge_brings_value_p.
206 (perhaps_add_new_callers): Likewise.
207 (get_info_about_necessary_edges): Likewise and exclude values brought
208 only by self-recursive edges.
209 (create_specialized_node): Redirect only clones of self-calling edges.
210 (+self_recursive_pass_through_p): New function.
211 (find_more_scalar_values_for_callers_subset): Use it.
212 (find_aggregate_values_for_callers_subset): Likewise.
213 (known_aggs_to_agg_replacement_list): Removed.
214 (decide_whether_version_node): Re-calculate known constants for all
215 remaining context clones.
217 2018-04-11 Richard Biener <rguenther@suse.de>
220 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
221 from early DWARF output.
222 (dwarf2out_early_finish): Output line info unconditionally into
223 early DWARF and add reference to it.
225 2018-04-11 Jakub Jelinek <jakub@redhat.com>
228 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
229 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
230 other than V2DFmode using iptr mode attribute.
231 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
233 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
235 PR rtl-optimization/84659
236 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
238 2018-04-11 Jakub Jelinek <jakub@redhat.com>
241 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
243 (output_loc_list): Pass address of a dummy size variable even in the
244 locview handling loop.
245 (index_location_lists): Add comment on why skip_loc_list_entry can't
248 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
251 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
254 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
257 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
258 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
259 and -mstring-compare-inline-limit.
261 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
264 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
265 for stack clash protection in a register whenever we need it to be in
268 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
270 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
271 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
273 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
276 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
278 (mlong-double-): Ditto.
279 * config/rs6000/sysv4.opt (msdata=): Ditto.
282 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
284 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
285 erroneous entries for
286 "vector int vec_ldl (int, long int *)", and
287 "vector unsigned int vec_ldl (int, unsigned long int *)".
288 Add comments and entries for
289 "vector bool char vec_ldl (int, bool char *)",
290 "vector bool short vec_ldl (int, bool short *)",
291 "vector bool int vec_ldl (int, bool int *)",
292 "vector bool long long vec_ldl (int, bool long long *)",
293 "vector pixel vec_ldl (int, pixel *)",
294 "vector long long vec_ldl (int, long long *)",
295 "vector unsigned long long vec_ldl (int, unsigned long long *)".
296 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
297 type tree bool_long_long_type_node and correct definition of
298 bool_V2DI_type_node to make reference to this new type tree.
299 (rs6000_mangle_type): Replace erroneous reference to
300 bool_long_type_node with bool_long_long_type_node.
301 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
302 comments to emphasize sign distinctions for char and int types and
303 replace RS6000_BTI_bool_long constant with
304 RS6000_BTI_bool_long_long constant. Also add comment to restrict
305 use of RS6000_BTI_pixel.
306 (bool_long_type_node): Remove this macro definition.
307 (bool_long_long_type_node): New macro definition
309 2018-04-10 Jakub Jelinek <jakub@redhat.com>
311 PR rtl-optimization/85300
312 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
313 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
314 simplify_unary_operation fails.
316 2018-04-10 Martin Liska <mliska@suse.cz>
318 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
319 cgraph_edge and ipa_ref.
321 2018-04-10 Jakub Jelinek <jakub@redhat.com>
326 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
327 computation of the VEC_MERGE selector from mask.
328 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
329 Fix decoding of the VEC_MERGE selector into mask.
331 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
333 PR tree-optimization/85286
334 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
336 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
338 * final.c (final_1): Set insn_last_address as well as
339 insn_current_address.
341 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
344 * explow.c (emit_stack_probe): Call validize_mem on memory location
345 before passing it to gen_probe_stack. Create address operand and
346 legitimize it for the probe_stack_address case.
348 2018-04-09 Jan Hubicka <jh@suse.cz>
351 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
352 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
353 * tree.c (free_lang_data_in_type): Fix handling of binfos;
355 (free_lang_data): Rebuild type inheritance graph.
357 2018-04-09 Martin Sebor <msebor@redhat.com>
359 * invoke.texi (-finline-small-functions): Mention other optimization
361 (-findirect-inlining, -fpartial-inlining): Same.
362 (-finline-functions-called-once): Same.
363 (-freorder-blocks-and-partition): Same.
365 2018-04-09 Jan Hubicka <jh@suse.cz>
368 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
369 jumps; choose last target that matches the criteria (i.e.
370 no partition changes for non-crossing jumps).
371 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
372 support for redirecting crossing jumps to non-crossing.
374 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
376 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
377 also for naked functions.
379 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
381 * config/arc/arc.md (add_shift): New pattern.
382 (add_shift2): Likewise.
383 (sub_shift): Likewise.
384 (sub_shift_cmp0_noout): Likewise.
385 (compare_si_ashiftsi): Likewise.
386 (xbfu_cmp0_noout): New combine pattern.
387 (xbfu_cmp0"): Likewise.
388 (movsi_set_cc_insn): Place the predicable variant first.
389 (commutative_binary_cmp0_noout): Remove clobber.
390 (commutative_binary_cmp0): New pattern.
391 (noncommutative_binary_cmp0): Likewise.
392 (noncommutative_binary_cmp0_noout): Likewise.
393 (noncommutative_binary_comparison_result_used): Removed.
394 (rsub_cmp0): New pattern.
395 (rsub_cmp0_noout): Likewise.
396 (extzvsi): Changed, keep only meaningful variants.
397 (SQH, SEZ): New iterators.
398 (SQH_postfix): New mode attribute.
399 (SEZ_prefix): New code attribute.
400 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
401 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
402 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
404 (noncommutative_operator): Check the availability of barrel
407 2018-04-09 Richard Biener <rguenther@suse.de>
409 PR tree-optimization/85284
410 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
411 Only use the niter constraining form of simple_iv when the exit
414 2018-04-09 Tom de Vries <tom@codesourcery.com>
417 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
418 (define_expand "*memory_barrier"): New define_expand.
419 (define_insn "memory_barrier"): New insn.
421 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
423 PR rtl-optimization/80463
424 PR rtl-optimization/83972
425 PR rtl-optimization/83480
427 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
428 correct producer for the insn.
429 (tidy_control_flow): Fixup seqnos in case of debug insns.
431 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
433 PR rtl-optimization/83913
435 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
436 different sched-times when merging exprs.
438 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
440 PR rtl-optimization/83962
442 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
443 tidy_fallthru_edge and tidy_control_flow.
445 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
447 PR rtl-optimization/83530
449 * sel-sched.c (force_next_insn): New global variable.
450 (remove_insn_for_debug): When force_next_insn is true, also leave only
451 next insn in the ready list.
452 (sel_sched_region): When the region wasn't scheduled, make another pass
453 over it with force_next_insn set to 1.
455 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
457 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
459 * config/nds32/constants.md (unspec_volatile_element): Add enum values
460 for interrupt control.
461 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
462 functions for interrupt control.
463 * config/nds32/nds32-intrinsic.md: Likewise.
464 * config/nds32/nds32_intrinsic.h: Likewise.
465 * config/nds32/nds32.h (nds32_builtins): Likewise.
467 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
469 * config/nds32/nds32.c (nds32_init_machine_status,
470 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
471 strict_aligned_p field.
472 (nds32_expand_to_rtl_hook): New function.
473 (TARGET_EXPAND_TO_RTL_HOOK): Define.
474 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
476 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
477 Chung-Ju Wu <jasonwucj@gmail.com>
479 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
480 * config/nds32/nds32-n7.md: New file.
481 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
482 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
484 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
485 * config/nds32/nds32.md (pipeline_model): Add n7.
486 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
487 * config/nds32/pipelines.md: Include n7 settings.
489 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
490 Chung-Ju Wu <jasonwucj@gmail.com>
492 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
493 * config/nds32/nds32-e8.md: New file.
494 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
495 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
497 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
498 * config/nds32/nds32.md (pipeline_model): Add e8.
499 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
500 * config/nds32/pipelines.md: Include e8 settings.
502 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
503 Chung-Ju Wu <jasonwucj@gmail.com>
505 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
506 * config/nds32/nds32-n8.md: New file.
507 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
508 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
510 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
511 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
512 * config/nds32/nds32.md (pipeline_model): Add n8.
513 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
514 * config/nds32/pipelines.md: Include n8 settings.
516 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
517 Chung-Ju Wu <jasonwucj@gmail.com>
519 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
520 * config/nds32/nds32-n9-2r1w.md: New file.
521 * config/nds32/nds32-n9-3r2w.md: New file.
522 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
523 nds32_register_ports): New or modify for cpu n9.
524 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
526 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
527 * config/nds32/nds32-utils.c: New file.
528 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
529 TARGET_MUL_SLOW): Define.
530 * config/nds32/nds32.md (pipeline_model): New attribute.
531 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
532 New options that support cpu n9.
533 * config/nds32/pipelines.md: Include n9 settings.
534 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
536 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
538 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
539 information if necessary.
540 (output_cond_branch_compare_zero): Likewise.
541 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
542 (nds32_target_alignment): Refine for alignment.
543 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
544 (FUNCTION_BOUNDARY): Modify.
545 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
547 * config/nds32/nds32.opt (malways-align, malign-functions): New.
549 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
551 * config/nds32/constants.md (unspec_volatile_element): Add values for
552 TLB operation and data prefetch.
553 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
554 functions for TLB operation and data prefetch.
555 * config/nds32/nds32-intrinsic.md: Likewise.
556 * config/nds32/nds32_intrinsic.h: Likewise.
557 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
558 (nds32_print_operand): Likewise.
559 * config/nds32/nds32.h (nds32_builtins): Likewise.
561 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
562 Andrew Pinski <pinsika@gcc.gnu.org>
565 * match.pd: Use constant_boolean_node of correct type instead of
566 boolean_true_node or boolean_false_node for simplifying
567 pointer comparisons to zero.
569 2018-04-07 Jakub Jelinek <jakub@redhat.com>
571 PR tree-optimization/80021
572 * tree.c (verify_type_variant): Make error call in verify_variant_match
573 translatable and remove final full stop.
575 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
577 * config/nds32/constants.md (unspec_volatile_element): Add
578 UNSPEC_VOLATILE_EH_RETURN.
579 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
580 nds32_output_stack_pop): Support dwarf exception handling process.
581 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
582 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
583 exception handling process.
584 (nds32_compute_stack_frame): Likewise.
585 (nds32_return_addr_rtx): Likewise.
586 (nds32_initial_elimination_offset): Likewise.
587 (nds32_expand_prologue): Likewise.
588 (nds32_expand_epilogue): Likewise.
589 (nds32_dynamic_chain_address): New function.
590 * config/nds32/nds32.h (machine_function): Add fields for dwarf
592 (DYNAMIC_CHAIN_ADDRESS): Define.
593 (EH_RETURN_DATA_REGNO): Define.
594 (EH_RETURN_STACKADJ_RTX): Define.
595 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
596 patterns for dwarf exception handling.
598 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
600 * config/nds32/nds32.h: Clean up obsolete macros.
602 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
604 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
605 Add enum values for particular instructions.
606 * config/nds32/nds32-intrinsic.c: Implementation of expanding
607 particular intrinsic functions.
608 * config/nds32/nds32-intrinsic.md: Likewise.
609 * config/nds32/nds32_intrinsic.h: Likewise.
610 * config/nds32/nds32.h (nds32_builtins): Likewise.
611 * config/nds32/nds32.md (type): Add pbsad and pbsada.
612 (btst, ave): New patterns for particular instructions.
614 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
616 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
617 Add enum values for atomic load/store and memory sync.
618 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
620 * config/nds32/nds32-intrinsic.md: Likewise.
621 * config/nds32/nds32_intrinsic.h: Likewise.
622 * config/nds32/nds32.h (nds32_builtins): Likewise.
624 2018-04-07 Jakub Jelinek <jakub@redhat.com>
626 PR tree-optimization/85257
627 * fold-const.c (native_encode_vector): If not all elts could fit
628 and off is -1, return 0 rather than offset.
629 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
630 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
631 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
632 adjust buffer in native_interpret_expr call.
634 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
636 * config/nds32/constants.md (unspec_volatile_element): Add cache
638 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
639 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
640 * config/nds32/nds32.c (nds32_cctl_names): New.
641 (nds32_print_operand): Handle cache control register names.
642 * config/nds32/nds32.h (nds32_builtins): New enum values.
643 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
645 * config/nds32/nds32.md (type): Add mmu.
646 * config/nds32/pipelines.md (simple_insn): Add mmu.
648 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
650 * config/nds32/nds32.md (type): Remove call.
651 * config/nds32/pipelines.md (simple_insn): Likewise.
653 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
655 * config/nds32/constants.md (unspec_volatile_element): Add
656 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
657 UNSPEC_VOLATILE_FMFCFG.
658 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
659 description for fmfcfg and fmfcsr.
660 (bdesc_1arg): Add fmtcsr.
661 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
662 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
663 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
664 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
665 unspec_fmfcfg): New patterns.
666 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
667 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
668 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
669 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
670 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
671 __nds32__fmfcfg): Define.
673 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
675 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
676 intrinsic register names.
677 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
678 intrinsic register enum values and macros.
680 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
682 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
683 for load/store addressing form.
684 (nds32_print_operand_address): Likewise.
686 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
689 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
690 based on LABEL_REF. Remove useless assertion.
691 (pic_address_needs_scratch): Fix formatting.
692 (sparc_legitimize_pic_address): Minor tweaks.
693 (sparc_delegitimize_address): Adjust assertion accordingly.
694 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
695 into symbolic_operand.
696 (movsi_high_pic_label_ref): Likewise.
697 (movsi_lo_sum_pic_label_ref): Likewise.
698 (movdi_pic_label_ref): Likewise.
699 (movdi_high_pic_label_ref): Likewise.
700 (movdi_lo_sum_pic_label_ref): Likewise.
702 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
704 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
705 custom LIB_SPEC setup.
707 2018-04-06 Ruslan Bukin <br@bsdpad.com>
708 Kito Cheng <kito.cheng@gmail.com>
710 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
711 * config/riscv/freebsd.h: New.
713 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
715 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
716 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
719 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
720 Kito Cheng <kito.cheng@gmail.com>
722 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
723 nds32_output_call, nds32_symbol_binds_local_p): New functions.
724 * config/nds32/nds32-protos.h (nds32_output_call,
725 nds32_output_return): Declare.
726 * config/nds32/nds32.md: Refine all the call and return patterns.
728 2018-04-06 Jakub Jelinek <jakub@redhat.com>
731 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
732 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
734 PR rtl-optimization/84872
735 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
736 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
739 2018-04-06 Tamar Christina <tamar.christina@arm.com>
741 * expr.c (copy_blkmode_to_reg): Revert 254862.
742 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
744 2018-04-06 Richard Biener <rguenther@suse.de>
747 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
748 after seeing a component reference with an adjacent field. Treat
749 refs to arrays at struct end of external decls similar to
750 refs to unconstrained commons.
752 2018-04-06 Jakub Jelinek <jakub@redhat.com>
755 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
756 look through SAVE_EXPRs with non-side-effects argument. Adjust
758 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
761 2018-04-06 Richard Biener <rguenther@suse.de>
764 * alias.c (find_base_term): New wrapper around find_base_term
765 unwinding CSELIB_VAL_PTR changes.
766 (find_base_term): Do not restore CSELIB_VAL_PTR during the
769 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
771 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
773 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
774 constant definitions.
775 ("nop"): lr 0,0 -> nopr r0
776 ("nop_lr0", "nop_lr1"): New insn definitions.
778 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
780 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
781 NDS32_V3PUSH_AVAILABLE_P macro.
783 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
784 Chung-Ju Wu <jasonwucj@gmail.com>
786 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
787 (nds32*-*-*): Add float and fpu_config into supported_defaults.
788 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
789 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
790 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
791 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
792 * config/nds32/constraints.md: New constraints and checking for hard
794 * config/nds32/iterators.md: New mode iterator and attribute for hard
796 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
798 * config/nds32/nds32-fpu.md: New file.
799 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
800 deal with hard float code generation.
801 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
803 (abi_type, float_reg_number): New enum type.
804 * config/nds32/nds32-predicates.c: New predicates for hard float.
805 * config/nds32/nds32-protos.h: Declare functions for hard float.
806 * config/nds32/nds32.c: Implementation for hard float configuration.
807 * config/nds32/nds32.h: Definitions for hard float configuration.
808 * config/nds32/nds32.md: Include hard float machine description and
809 modify patterns for hard float configuration.
810 * config/nds32/nds32.opt: New options for hard float configuration.
811 * config/nds32/predicates.md: New predicates for hard float
814 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
816 * common/config/nds32/nds32-common.c
817 (nds32_option_optimization_table): Enable -mreleax-hint by default.
819 2018-04-05 Jakub Jelinek <jakub@redhat.com>
822 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
823 CONSTRUCTOR_ELT (ctor, ...)->value.
825 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
828 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
830 2018-04-05 Tom de Vries <tom@codesourcery.com>
833 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
836 2018-04-05 Shiva Chen <shiva0217@gmail.com>
837 Kito Cheng <kito.cheng@gmail.com>
839 * config/nds32/constraints.md (U33): Fine-tune checking condition.
840 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
841 * config/nds32/nds32.h (nds32_16bit_address_type): Add
842 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
844 2018-04-05 Shiva Chen <shiva0217@gmail.com>
845 Kito Cheng <kito.cheng@gmail.com>
847 * config/nds32/constraints.md (Ufe): New memory constraint.
848 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
849 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
850 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
852 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
853 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
855 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
857 * config/nds32/nds32.md: Use optimize_size in the condition for
858 alu-shift instructions.
860 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
862 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
864 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
866 * config/nds32/nds32.md (negsi2): Refine pattern.
868 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
869 Chung-Ju Wu <jasonwucj@gmail.com>
871 * config/nds32/iterators.md (shift_rotate): New code iterator.
872 (shift): New code attribute.
873 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
874 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
875 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
876 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
878 (andsi3, *andsi3): Ditto.
879 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
880 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
881 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
882 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
883 nds32_ior_operand, nds32_xor_operand): New predicates.
885 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
887 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
888 (addsi3, subsi3): ... this.
890 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
892 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
894 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
896 * config/nds32/nds32.md: Adjust indention.
898 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
900 * config/nds32/nds32.md (feature): New attribute.
902 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
904 * config/nds32/nds32.md (subtype): New attribute.
906 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
909 * config/arm/arm-builtins.c (arm_expand_builtin): Change
910 expansion to perform a bitwise AND of the argument followed by a
911 boolean negation of the result.
913 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
915 PR rtl-optimization/84878
916 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
917 the basic block. Assert the use reference is not artificial and that
918 it has an associated insn.
920 2018-04-04 Michael Matz <matz@suse.de>
922 * builtins.c (compute_objsize): Pass correct operand
923 to array_at_struct_end_p.
925 2018-04-04 Richard Biener <rguenther@suse.de>
928 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
929 from contexts for DINFO_LEVEL_TERSE and below.
931 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
933 * config/nds32/nds32-doubleword.md (move_<mode>): Require
934 resiter_operand condition.
935 * config/nds32/nds32.md (*move<mode>): Ditto.
937 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
938 Monk Chiang <sh.chiang04@gmail.com>
940 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
942 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
944 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
946 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
947 Kito Cheng <kito.cheng@gmail.com>
949 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
950 nds32_cond_code_str, output_cond_branch,
951 output_cond_branch_compare_zero, nds32_expand_cbranch,
952 nds32_expand_cstore, nds32_expand_movcc,
953 nds32_output_cbranchsi4_equality_zero,
954 nds32_output_cbranchsi4_equality_reg,
955 nds32_output_cbranchsi4_equality_reg_or_const_int,
956 nds32_output_cbranchsi4_greater_less_zero: New functions.
957 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
958 nds32_expand_cstore, nds32_expand_movcc,
959 nds32_output_cbranchsi4_equality_zero,
960 nds32_output_cbranchsi4_equality_reg,
961 nds32_output_cbranchsi4_equality_reg_or_const_int,
962 nds32_output_cbranchsi4_greater_less_zero): Declare.
963 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
964 nds32_rimm11s_operand): New predicates.
965 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
966 * config/nds32/nds32.md: Rewrite all the branch and conditional move
969 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
971 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
972 * config/nds32/nds32.md: Ditto.
973 * config/nds32/pipelines.md: Ditto.
975 2018-04-04 Richard Biener <rguenther@suse.de>
977 PR tree-optimization/85168
978 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
979 propagating abnormals.
981 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
983 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
985 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
986 Kito Cheng <kito.cheng@gmail.com>
988 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
989 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
990 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
991 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
992 * config/nds32/nds32.md (sibcall_internal): New.
993 (sibcall_register): Remove.
994 (sibcall_immediate): Remove.
995 (sibcall_value_internal): New.
996 (sibcall_value_register): Remove.
997 (sibcall_value_immediate): Remove.
998 * config/nds32/predicates.md (nds32_general_register_operand): New.
999 (nds32_call_address_operand): New.
1001 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1003 PR rtl-optimization/85167
1004 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
1005 bb_defs if *split_p, instead preinitialize it to NULL.
1007 PR tree-optimization/85156
1008 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
1009 evaluating the argument multiple times.
1011 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
1013 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
1015 (_mm_cvtpd_ps): Likewise.
1016 (_mm_cvttpd_epi32): Likewise.
1017 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
1018 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
1019 vector, pixel, and bool following altivec.h include.
1021 2018-04-03 Martin Sebor <msebor@redhat.com>
1023 * doc/extend.texi (Common Function Attributes): Clarify.
1024 (const attribute): Likewise.
1025 (pure attribute): Likewise.
1027 2018-04-03 Jakub Jelinek <jakub@redhat.com>
1030 * config/i386/i386.c (ix86_expand_vector_set): Use
1031 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
1033 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
1035 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
1036 instructions when changing rounding bits to preserve precision bits
1037 in the x87 control word.
1039 2018-04-03 Martin Liska <mliska@suse.cz>
1041 PR tree-optimization/82491
1042 * rtl.h (strip_offset_and_add): Replace += suboffset with
1043 poly_uint64 () + suboffset.
1045 2018-03-29 Martin Liska <mliska@suse.cz>
1046 Martin Jambor <mjambor@suse.cz>
1049 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
1050 param_type is not an integral or pointer type.
1052 2018-04-03 Richard Biener <rguenther@suse.de>
1054 * sese.h (recompute_all_dominators): Remove.
1056 2018-04-02 Martin Sebor <msebor@redhat.com>
1058 * doc/invoke.texi (-Wrestrict): Fix typos.
1060 2018-04-02 Jim Wilson <jimw@sifive.com>
1062 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
1063 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
1064 (<optab>di3, <optab>si3_extend): Likewise.
1065 (<optab>si3_mask, <optab>si3_mask_1): New.
1066 (<optab>di3_mask, <optab>di3_mask_1): New.
1067 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
1068 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
1069 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
1071 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
1073 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
1076 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
1078 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
1079 (nds32_canonicalize_comparison): New function.
1081 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1082 Kito Cheng <kito.cheng@gmail.com>
1083 Kuan-Lin Chen <kuanlinchentw@gmail.com>
1085 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
1086 * config/nds32/constants.md (unspec_volatile_element): Add
1087 UNSPEC_VOLATILE_RELAX_GROUP.
1088 * config/nds32/nds32-relax-opt.c: New file.
1089 * config/nds32/nds32-predicates.c
1090 (nds32_symbol_load_store_p): New function.
1091 * config/nds32/nds32-protos.h
1092 (nds32_symbol_load_store_p): Declare function.
1093 (make_pass_nds32_relax_opt): Declare new rtl pass function.
1094 * config/nds32/nds32.c
1095 (nds32_register_pass): New function to register pass.
1096 (nds32_register_passes): New function to register passes.
1097 * config/nds32/nds32.md (relax_group): New pattern.
1098 * config/nds32/nds32.opt (mrelax-hint): New option.
1099 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
1101 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
1103 * config/nds32/t-nds32: Modify files dependency.
1105 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1107 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
1108 (PROFILE_HOOK): Define its implementation.
1110 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
1112 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
1113 type and 32-bit size.
1115 2018-04-01 Jakub Jelinek <jakub@redhat.com>
1118 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
1119 (V_128_256): New mode iterator.
1120 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
1121 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
1122 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1124 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1125 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1127 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1130 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1131 NaN inputs correctly.
1133 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1136 * config/rs6000/vsx.md (??r): New mode attribute.
1137 (*vsx_mov<mode>_64bit): Use it.
1138 (*vsx_mov<mode>_32bit): Likewise.
1140 2018-03-30 Martin Sebor <msebor@redhat.com>
1142 PR tree-optimization/84818
1143 * builtins.c (check_access): Use warning_n.
1145 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1148 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1150 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1153 2018-03-30 Julia Koval <julia.koval@intel.com>
1156 * x86-tune.def (movx, partial_reg_dependency): Enable for
1159 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1162 * lra-constraints.c (process_alt_operands): Move setting
1163 this_alternative_matches below.
1165 2018-03-29 Martin Liska <mliska@suse.cz>
1168 * doc/invoke.texi: Document how LTO works with debug info.
1169 Describe auto-load support of binutils. Mention 'x86-64'
1170 as valid option value of -march option.
1172 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1174 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1177 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1178 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1179 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1182 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1185 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1186 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1187 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1188 for RS6000_BTM_POWERPC64.
1189 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1190 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1191 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1196 2018-03-28 Carl Love <cel@us.ibm.com>
1199 2017-09-27 Carl Love <cel@us.ibm.com>
1201 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1202 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1203 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1206 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1208 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1209 instead of __vector bool.
1210 (_mm_max_pu8): Likewise.
1211 (_mm_min_pi16): Likewise.
1213 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1216 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1217 (DIVWEUO): Likewise.
1219 (DIVDEUO): Likewise.
1220 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1221 DIVWEUO and DIVDEUO.
1222 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1223 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1224 (div_extend): Likewise.
1225 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1227 (__builtin_divweuo): Likewise.
1228 (__builtin_divdeo): Likewise.
1229 (__builtin_divdeuo): Likewise.
1231 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1234 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1235 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1237 PR tree-optimization/82004
1238 * gimple-match-head.c (optimize_pow_to_exp): New function.
1239 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1240 Don't fold to exp if optimize_pow_to_exp is false.
1242 2018-03-28 Martin Liska <mliska@suse.cz>
1245 * calls.c (initialize_argument_information): Fix trailing space.
1246 * common.opt: Fix typo and provide better explanation for
1247 -fsanitize-coverage option.
1248 * config/i386/i386.opt: Fix typo.
1250 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1251 Martin Liska <mliska@suse.cz>
1254 * gimplify.c (asan_poison_variable): Don't do the check for
1255 gimplify_omp_ctxp here.
1256 (gimplify_decl_expr): Do it here.
1257 (gimplify_target_expr): Likewise.
1259 2018-03-28 Martin Liska <mliska@suse.cz>
1262 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1263 chkp_type_bounds_count if MPX is not enabled.
1265 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1267 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1269 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1272 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1273 function to create the function decl for complex long double
1274 multiply and divide for -mabi=ieeelongdouble.
1275 (init_float128_ieee): Call it.
1277 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1280 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1281 -fcf-protection=branch -mibt.
1282 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1284 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1287 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1289 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1292 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1293 extern array declarations.
1295 2018-03-27 Richard Biener <rguenther@suse.de>
1298 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1299 explicit single_use checks.
1301 2018-03-27 Richard Biener <rguenther@suse.de>
1303 PR tree-optimization/85082
1304 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1307 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1309 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1310 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1311 Turn on fasynchronous-unwind-tables and funwind-tables.
1313 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1316 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1317 (*bmi_blsr_<mode>_ccz): Ditto.
1319 2018-03-26 Tom de Vries <tom@codesourcery.com>
1321 PR tree-optimization/85063
1322 * omp-general.c (offloading_function_p): New function. Factor out
1324 * omp-offload.c (pass_omp_target_link::gate): ... here.
1325 * omp-general.h (offloading_function_p): Declare.
1326 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1327 with attribute omp declare target for offloading functions.
1329 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1331 PR tree-optimization/84005
1332 * tree-data-ref.h (get_base_for_alignment): Declare.
1333 * tree-data-ref.c (get_base_for_alignment_1): New function.
1334 (get_base_for_alignment): Likewise.
1335 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1336 get_base_for_alignment to find a suitable base object, instead
1337 of always using drb->base_address.
1339 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1342 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1343 known size by default.
1345 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1348 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1349 and non BLKmode operands.
1351 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1354 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1355 Clean up attributes.
1357 2018-03-23 Richard Biener <rguenther@suse.de>
1360 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1361 we are going to emit early debug for LTO.
1363 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1366 * function.c (match_asm_constraints_1): Don't optimize if input
1367 doesn't satisfy general_operand predicate for output's mode.
1370 * alias.c (write_dependence_p): Don't require for x_canonicalized
1371 non-VOIDmode if x has VOIDmode.
1374 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1375 just don't try to optimize it rather than assert it never happens.
1377 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1379 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1380 macro expansions for definition of ST_INTERNAL_<mode> and
1381 LD_INTERNAL_<mode> builtins.
1382 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1384 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1386 (altivec_expand_st_builtin): Likewise.
1387 (altivec_expand_builtin): Remove calls to deleted functions.
1388 (rs6000_address_for_altivec): Delete this function.
1389 * config/rs6000/vector.md: Remove expands for
1390 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1392 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1395 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1396 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1397 re-computing once computed.
1398 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1399 (arm_init_machine_status): Initialize
1400 machine->static_chain_stack_bytes.
1402 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1405 * doc/extend.texi: Add four new prototypes for vec_ld.
1406 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1407 definitions for more logical presentation.
1408 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1409 entries for V1TI variants of __builtin_altivec_ld builtin.
1410 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1411 handling of V1TI variant of LVX icode pattern.
1412 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1413 (rs6000_gimple_fold_builtin): Likewise.
1414 (altivec_init_builtins): Add code to define
1415 __builtin_altivec_lvx_v1ti function.
1417 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1420 * function.c (match_asm_constraints_1): Don't do the optimization
1421 if input isn't a REG, SUBREG, MEM or constant.
1423 2018-03-22 Tom de Vries <tom@codesourcery.com>
1425 PR tree-optimization/84956
1426 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1427 bb_has_abnormal_pred.
1429 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1432 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1433 DECL_INITIAL (decl) to decl at the end.
1434 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1437 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1439 * doc/extend.texi (__builtin_tgmath): Document when complex
1440 integer types are treated as _Complex _Float64.
1442 2018-03-21 Tom de Vries <tom@codesourcery.com>
1444 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1446 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1448 PR tree-optimization/84960
1449 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1450 if it is ENTRY block, move them into single succ of ENTRY in that case.
1452 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1454 PR tree-optimization/84811
1455 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1456 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1457 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1458 (known_subrange_p): Update accordingly. Cast each value involved
1459 in the size comparison, rather than casting the result of the
1462 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1464 PR tree-optimization/84982
1465 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1466 by flipping the least significant bit rather than all bits from
1467 bitpos to bitpos + bitsize - 1.
1469 2018-03-21 Nathan Sidwell <nathan@acm.org>
1471 * doc/extend.texi (Deprecated Features): Remove mention of
1472 long-deleted deprecations.
1474 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1477 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1478 * configure: Regenerate.
1480 2018-03-21 Tom de Vries <tom@codesourcery.com>
1482 PR tree-optimization/83126
1483 * tree-parloops.c (num_phis): New function.
1484 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1486 2018-03-21 Nathan Sidwell <nathan@acm.org>
1488 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1489 mention anon-struct/union members and trailing attributes.
1491 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1493 PR tree-optimization/84969
1494 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1495 builtin memset partitions if they set different rhs values.
1497 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1499 PR rtl-optimization/84989
1500 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1501 VEC_DUPLICATE with scalar result mode.
1503 2018-03-21 Martin Liska <mliska@suse.cz>
1506 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1507 not intended return statement.
1509 2018-03-21 Martin Liska <mliska@suse.cz>
1512 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1513 (chkp_find_bound_slots_1): Limit number of iterations.
1515 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1518 * Minor grammar fixes for x86 options.
1520 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1523 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1524 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1526 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1529 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1530 (altivec_lvx_<mode>_1op): Likewise.
1531 (altivec_stvx_<mode>_2op): Likewise.
1532 (altivec_stvx_<mode>_1op): Likewise.
1533 (altivec_lvx_<VM2:mode>): New define_expand.
1534 (altivec_stvx_<VM2:mode>): Likewise.
1535 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1536 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1537 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1538 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1539 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1540 (rs6000_gen_lvx): Likewise.
1541 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1542 (altivec_expand_stv_builtin): Likewise.
1543 (altivec_expand_builtin): Likewise.
1544 * config/rs6000/vector.md: Likewise.
1546 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1549 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1552 2018-03-20 Richard Biener <rguenther@suse.de>
1555 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1556 sign-conversions as zero, fall back to standard scalar_stmt
1559 2018-03-20 Martin Liska <mliska@suse.cz>
1562 * predict.c (rebuild_frequencies): Handle case when we have
1563 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1565 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1568 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1569 flag_section_anchors.
1570 * varasm.c (use_blocks_for_decl_p): Remove hack for
1571 dw2_force_const_mem.
1574 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1576 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1577 be created, use lowpart_subreg of operands[0] rather than operands[0]
1579 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1580 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1581 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1582 and n constraint instead of aarch64_shift_imm_di and Usd.
1583 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1584 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1586 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1589 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1590 to favor GPR over NEON registers.
1591 (<shift>di3_neon): Likewise.
1593 2018-03-20 Tom de Vries <tom@codesourcery.com>
1596 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1597 (nvptx_process_pars): Emit bar.sync asap and alap.
1599 2018-03-20 Tom de Vries <tom@codesourcery.com>
1602 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1603 seen_label if seen_label is already set.
1605 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1608 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1609 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1610 Use 1U instead of 1. Formatting fixes.
1613 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1614 instead of TREE_TYPE (s1) for the return value.
1616 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1618 PR tree-optimization/84946
1619 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1620 bitsize + bitsize in poly_uint64 rather than poly_int64.
1623 * dwarf2asm.c: Include fold-const.c.
1624 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1625 of decl rather than decl itself.
1627 PR rtl-optimization/84643
1628 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1630 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1633 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1634 calling assemble_variable.
1636 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1639 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1640 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1642 2018-03-19 Jim Wilson <jimw@sifive.com>
1645 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1646 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1647 (riscv_first_stack_step): Likewise.
1648 (riscv_option_override): Use STACK_BOUNDARY instead of
1650 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1652 (BIGGEST_ALIGNMENT): Set to 128.
1653 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1654 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1657 2018-03-19 Richard Biener <rguenther@suse.de>
1659 PR tree-optimization/84933
1660 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1661 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1663 2018-03-19 Richard Biener <rguenther@suse.de>
1665 PR tree-optimization/84859
1666 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1667 (cond_if_else_store_replacement): Perform sinking operation on
1668 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1669 Generalize what a BB with a single eligible store is.
1671 2018-03-19 Richard Biener <rguenther@suse.de>
1673 PR tree-optimization/84929
1674 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1675 chrec_is_positive against non-chrec arg.
1677 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1680 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1682 2018-03-18 Martin Liska <mliska@suse.cz>
1684 PR rtl-optimization/84635
1685 * regrename.c (build_def_use): Use matches_mode only when
1688 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1690 PR tree-optimization/84913
1691 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1692 vectorize chains of COND_EXPRs.
1694 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1696 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1698 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1700 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1702 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1704 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1706 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1707 Kito Cheng <kito.cheng@gmail.com>
1709 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1710 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1711 (nds32_adjust_reg_alloc_order): New function.
1712 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1714 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1716 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1717 nds32_print_operand, nds32_print_operand_address): Use
1718 HOST_WIDE_INT_PRINT_DEC instead.
1720 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1722 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1724 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1727 * config/i386/i386.c (initial_ix86_tune_features,
1728 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1730 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1731 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1732 rather than 1u << ix86_tune. Formatting fix.
1733 (ix86_option_override_internal): Change ix86_arch_mask from
1734 unsigned int to unsigned HOST_WIDE_INT, initialize to
1735 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1736 (ix86_function_specific_restore): Likewise.
1738 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1741 * postreload.c (reload_combine_recognize_pattern): Perform
1742 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1743 truncate_int_for_mode the result for the destination's mode.
1746 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1747 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1749 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1752 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1755 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1756 regno_allocno_class_array and sorted_pseudos.
1757 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1758 insns where regno is used.
1760 2018-03-16 Martin Liska <mliska@suse.cz>
1763 * multiple_target.c (create_dispatcher_calls): Redirect
1764 reference in the symbol table.
1766 2018-03-16 Martin Liska <mliska@suse.cz>
1769 * multiple_target.c (create_dispatcher_calls): Redirect also
1772 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1776 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1777 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1778 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1780 2018-03-16 Julia Koval <julia.koval@intel.com>
1782 * doc/invoke.texi (Skylake Server): Add CLWB.
1783 Cannonlake): Remove CLWB.
1785 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1787 PR tree-optimization/84841
1788 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1790 (FLOAT_ONE_CONST_TYPE): Define.
1791 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1792 (sort_by_operand_rank): Put entries with higher constant_type last
1793 rather than first to match comments.
1795 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1797 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1800 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1803 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1804 check and use address of target always.
1806 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1809 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1810 (indirect_thunk_bnd_needed): Likewise.
1811 (indirect_thunks_used): Likewise.
1812 (indirect_thunks_bnd_used): Likewise.
1813 (indirect_return_needed): New.
1814 (indirect_return_bnd_needed): Likewise.
1815 (output_indirect_thunk_function): Add a bool argument for
1817 (output_indirect_thunk_function): Don't generate alias for
1818 function return thunk.
1819 (ix86_code_end): Call output_indirect_thunk_function to generate
1820 function return thunks.
1821 (ix86_output_function_return): Set indirect_return_bnd_needed
1822 and indirect_return_needed instead of indirect_thunk_bnd_needed
1823 and indirect_thunk_needed.
1825 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1827 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1828 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1829 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1831 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1832 Paul Hua <paul.hua.gm@gmail.com>
1835 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1837 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1839 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1840 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1843 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1846 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1847 instead of GET_MODE_SIZE when comparing Units.
1849 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1852 * varasm.c (hash_section): Return an unchangeble hash value
1853 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1854 Return !aarch64_can_use_per_function_literal_pools_p ().
1856 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1859 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1860 rather than address of cmode as last argument to prepare_cmp_insn.
1862 2018-03-15 Julia Koval <julia.koval@intel.com>
1864 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1865 F_AVX512VNNI, F_AVX512BITALG): New.
1867 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1870 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1871 insn for floating-point loads and stores.
1873 2018-03-14 Carl Love <cel@us.ibm.com>
1875 * config/rs6000/rs6000-c.c: Add macro definitions for
1876 ALTIVEC_BUILTIN_VEC_PERMXOR.
1877 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1878 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1879 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1880 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1882 * config/doc/extend.texi: Add prototypes for vec_permxor.
1884 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1887 * diagnostic-show-locus.c (class layout_point): Convert m_line
1888 from int to linenum_type.
1889 (line_span::comparator): Use linenum "compare" function when
1890 comparing line numbers.
1891 (test_line_span): New function.
1892 (layout_range::contains_point): Convert param "row" from int to
1894 (layout_range::intersects_line_p): Likewise.
1895 (layout::will_show_line_p): Likewise.
1896 (layout::print_source_line): Likewise.
1897 (layout::should_print_annotation_line_p): Likewise.
1898 (layout::print_annotation_line): Likewise.
1899 (layout::print_leading_fixits): Likewise.
1900 (layout::annotation_line_showed_range_p): Likewise.
1901 (struct line_corrections): Likewise for field m_row.
1902 (line_corrections::line_corrections): Likewise for param "row".
1903 (layout::print_trailing_fixits): Likewise.
1904 (layout::get_state_at_point): Likewise.
1905 (layout::get_x_bound_for_row): Likewise.
1906 (layout::print_line): Likewise.
1907 (diagnostic_show_locus): Likewise for locals "last_line" and
1909 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1910 * input.c (selftest::test_linenum_comparisons): New function.
1911 (selftest::input_c_tests): Call it.
1912 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1913 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1914 * selftest.h (ASSERT_GT): New macro.
1915 (ASSERT_GT_AT): New macro.
1916 (ASSERT_LT): New macro.
1917 (ASSERT_LT_AT): New macro.
1919 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1921 PR rtl-optimization/84780
1922 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1924 2018-03-14 Martin Liska <mliska@suse.cz>
1926 * tree.c (record_node_allocation_statistics): Use
1927 get_stats_node_kind.
1928 (get_stats_node_kind): New function extracted from
1929 record_node_allocation_statistics.
1930 (free_node): Use get_stats_node_kind.
1932 2018-03-14 Richard Biener <rguenther@suse.de>
1934 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1935 that the value-set of ANTIC_IN doesn't grow.
1938 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1940 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1941 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1943 2018-03-14 Julia Koval <julia.koval@intel.com>
1945 * config.gcc (icelake-client, icelake-server): New.
1947 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1948 (initial_ix86_arch_features): Ditto.
1949 (PTA_SKYLAKE): Add SGX.
1950 (PTA_ICELAKE): Remove.
1951 (PTA_ICELAKE_CLIENT): New.
1952 (PTA_ICELAKE_SERVER): New.
1953 (ix86_option_override_internal): Split up icelake on icelake client and
1955 (get_builtin_code_for_version): Ditto.
1956 (fold_builtin_cpu): Ditto.
1957 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1958 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1959 * config/i386/i386.h (processor_type): Ditto.
1960 * doc/invoke.texi: Ditto.
1962 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1965 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1966 INTEGER_CST offset, add it together with bitpos / 8 and
1967 sign extend based on POINTER_SIZE.
1971 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1974 * config/i386/constraints.md (Yc): New register constraint.
1975 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1976 Use Yc constraint for alternative 2 of operand 0. Remove
1977 preferred_for_speed attribute.
1979 2018-03-14 Richard Biener <rguenther@suse.de>
1981 PR tree-optimization/84830
1982 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1983 with the old one to avoid oscillations.
1985 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1988 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1990 (assign_by_spills): Return a flag of reload assignment failure.
1991 Do not process the reload assignment failures. Do not spill other
1992 reload pseudos if they has the same reg class. Update n if
1994 (lra_assign): Add a return arg. Set up from the result of
1995 assign_by_spills call.
1996 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1997 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1998 usage_insns if it is not NULL.
1999 (spill_hard_reg_in_range): New function.
2000 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2001 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2002 function prototypes.
2003 (lra_assign): Change prototype.
2004 * lra.c (lra): Add code to deal with fails by splitting hard reg
2007 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
2009 * config/riscv/riscv.opt (mrelax): New option.
2010 * config/riscv/riscv.c (riscv_file_start): Emit ".option
2011 "norelax" when riscv_mrelax is disabled.
2012 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
2014 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2017 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
2018 reassociation for int modes.
2020 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2022 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
2023 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
2025 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
2026 * config/aarch64/aarch64-sve.md
2027 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
2028 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
2029 (*extend<mode><Vwide>2): Rename to...
2030 (aarch64_sve_extend<mode><Vwide>2): ...this.
2031 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
2032 renaming the old pattern to...
2033 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
2035 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
2036 define_expand, renaming the old pattern to...
2037 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
2038 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
2039 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
2040 account when deciding which SVE instruction the optab should use.
2041 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
2043 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2045 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
2046 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
2047 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
2048 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
2049 (tlsdesc_small_<mode>): Turn a define_expand and use
2050 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
2051 (tlsdesc_small_advsimd_<mode>): ...this.
2052 (tlsdesc_small_sve_<mode>): New pattern.
2054 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
2056 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
2057 (UNSPEC_UMUL_HIGHPART): New constants.
2058 (MUL_HIGHPART): New int iteraor.
2059 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
2060 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
2062 (*<su>mul<mode>3_highpart): New define_insn.
2064 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
2067 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
2070 2018-03-13 Martin Liska <mliska@suse.cz>
2073 * (sem_item_optimizer::sem_item_optimizer): Initialize new
2075 (sem_item_optimizer::~sem_item_optimizer): Release it.
2076 (sem_item_optimizer::merge_classes): Register variable aliases.
2077 (sem_item_optimizer::fixup_pt_set): New function.
2078 (sem_item_optimizer::fixup_points_to_sets): Likewise.
2079 * ipa-icf.h: Declare new variables and functions.
2081 2018-03-13 Jakub Jelinek <jakub@redhat.com>
2084 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
2085 integer_pow2p@2 and test integer_pow2p in condition.
2086 (A < 0 ? C : 0): Similarly for @1.
2089 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
2090 characters starting at p contain '\0' character, don't look beyond
2094 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
2095 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
2098 * reg-stack.c (change_stack): Change update_end var from int to
2099 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
2100 also call set_block_for_insn on the newly added insns and rescan.
2103 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
2104 on the last operand.
2107 * tree.c (stabilize_reference_1): Return save_expr (e) for
2108 STATEMENT_LIST even if it doesn't have side-effects.
2110 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
2112 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
2114 2018-03-12 Renlin Li <renlin.li@arm.com>
2116 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
2117 aarch64_output_scalar_simd_mov_immediate.
2119 2018-03-12 Martin Sebor <msebor@redhat.com>
2121 PR tree-optimization/83456
2122 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2123 for perfectly overlapping calls to memcpy.
2124 (gimple_fold_builtin_memory_chk): Same.
2125 (gimple_fold_builtin_strcpy): Handle no-warning.
2126 (gimple_fold_builtin_stxcpy_chk): Same.
2127 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2129 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2131 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2132 parameter. Use it for SFmode.
2133 (rs6000_function_arg_advance_1): Adjust.
2134 (rs6000_function_arg): Adjust.
2135 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2137 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2139 PR rtl-optimization/84169
2140 PR rtl-optimization/84780
2141 * combine.c (can_combine_p): Check for a 2-insn combination whether
2142 the destination register is used between the two insns, too.
2144 2018-03-12 Richard Biener <rguenther@suse.de>
2146 PR tree-optimization/84803
2147 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2148 for refs DR analysis didn't process.
2150 2018-03-12 Richard Biener <rguenther@suse.de>
2152 PR tree-optimization/84777
2153 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2154 force-vectorize loops ignore whether we are optimizing for size.
2156 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2158 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2159 (TARGET_MD_ASM_ADJUST): Define.
2161 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2162 Kito Cheng <kito.cheng@gmail.com>
2163 Chung-Ju Wu <jasonwucj@gmail.com>
2165 * config/nds32/nds32.c (nds32_compute_stack_frame,
2166 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2167 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2168 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2169 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2170 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2171 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2172 * config/nds32/nds32.md (prologue, epilogue): Use macro
2173 NDS32_V3PUSH_AVAILABLE_P to do checking.
2175 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2178 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2179 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2180 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2181 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2182 addition of most attributes on !orig_type_die or the attribute not
2183 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2185 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2186 Chung-Ju Wu <jasonwucj@gmail.com>
2188 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2190 * config/nds32/nds32.opt (mvh): New option.
2192 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2193 Chung-Ju Wu <jasonwucj@gmail.com>
2195 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2197 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2198 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2201 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2202 Chung-Ju Wu <jasonwucj@gmail.com>
2204 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2206 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2207 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2209 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2210 Kito Cheng <kito.cheng@gmail.com>
2211 Chung-Ju Wu <jasonwucj@gmail.com>
2213 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2214 UNSPEC_FFMISM and UNSPEC_FLMISM.
2215 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2216 for ffb, ffmism and flmism.
2217 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2218 (unspec_ffmism): Ditto.
2219 (unspec_flmism): Ditto.
2220 (nds32_expand_builtin_impl): Check if string extension is available.
2221 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2222 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2224 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2227 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2230 * lra-assigns.c (assign_by_spills): Return a flag of reload
2231 assignment failure. Do not process the reload assignment
2232 failures. Do not spill other reload pseudos if they has the same
2234 (lra_assign): Add a return arg. Set up from the result of
2235 assign_by_spills call.
2236 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2237 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2238 usage_insns if it is not NULL.
2239 (spill_hard_reg_in_range): New function.
2240 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2241 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2242 function prototypes.
2243 (lra_assign): Change prototype.
2244 * lra.c (lra): Add code to deal with fails by splitting hard reg
2247 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2250 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2252 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2255 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2256 (dw_val_node): Add val_symbolic_view.
2257 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2258 (symview_upper_bound): New.
2259 (new_line_info_table): Initialize symviews_since_reset.
2260 (dwarf2out_source_line): Count symviews_since_reset and set
2261 symview_upper_bound.
2262 (dw_val_equal_p): Handle symview.
2263 (add_AT_symview): New.
2264 (print_dw_val): Handle symview.
2265 (attr_checksum, attr_checksum_ordered): Likewise.
2266 (same_dw_val_p, size_of_die): Likewise.
2267 (value_format, output_die): Likewise.
2268 (add_high_low_attributes): Use add_AT_symview for entry_view.
2269 (dwarf2out_finish): Reset symview_upper_bound, clear
2272 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2275 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2276 Add strict argument and use it.
2277 (rs6000_split_multireg_move): Update for new strict argument.
2278 (mem_operand_gpr): Disallow all non-offsettable addresses.
2279 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2281 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2284 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2285 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2286 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2289 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2290 decl, use remap_type if we want to use the type.
2292 2018-03-09 Martin Sebor <msebor@redhat.com>
2294 PR tree-optimization/84526
2295 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2297 (builtin_access::generic_overlap): Be prepared to handle non-array
2300 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2302 PR rtl-optimization/84682
2303 * lra-constraints.c (process_address_1): Check is_address flag
2304 for address constraints.
2305 (process_alt_operands): Likewise.
2306 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2307 preprocess_constraints.
2308 * recog.h (preprocess_constraints): Add oploc parameter.
2310 * recog.c (preprocess_constraints): Test address_operand for
2311 CT_ADDRESS constraints.
2313 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2316 * lra-assigns.c (assign_by_spills): Return a flag of reload
2317 assignment failure. Do not process the reload assignment
2318 failures. Do not spill other reload pseudos if they has the same
2320 (lra_assign): Add a return arg. Set up from the result of
2321 assign_by_spills call.
2322 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2323 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2324 usage_insns if it is not NULL.
2325 (spill_hard_reg_in_range): New function.
2326 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2327 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2328 function prototypes.
2329 (lra_assign): Change prototype.
2330 * lra.c (lra): Add code to deal with fails by splitting hard reg
2333 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2336 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2337 Accept complain bool parameter. Only emit errors if it is true.
2338 (arm_parse_cpu_option_name): Likewise.
2339 (arm_target_thumb_only): Adjust callers of the above.
2340 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2341 prototype to take a default true bool parameter.
2342 (arm_parse_arch_option_name): Likewise.
2344 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2345 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2349 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2350 * configure: Regenerate.
2351 * configure.ac ("linker --version-script option"): New.
2352 ("linker soname option"): New.
2354 2018-03-09 Richard Biener <rguenther@suse.de>
2356 PR tree-optimization/84775
2357 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2358 immediate uses of predicate stmts and mark them modified.
2361 PR tree-optimization/84178
2362 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2364 (version_loop_for_if_conversion): Delay update_ssa call.
2365 (tree_if_conversion): Delay update_ssa until after predicate
2368 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2371 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2372 when the function accesses prior frames.
2374 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2377 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2378 gen_llsym, otherwise call maybe_gen_llsym.
2381 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2382 has ',' character inside of it.
2384 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2387 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2388 as clobbering CC_REGNUM.
2390 2018-03-08 Richard Biener <rguenther@suse.de>
2393 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2394 (follow_copies_to_constant): Do not follow SSA names registered
2397 2018-03-08 Richard Biener <rguenther@suse.de>
2399 PR tree-optimization/84178
2400 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2402 (version_loop_for_if_conversion): Delay update_ssa call.
2403 (tree_if_conversion): Delay update_ssa until after predicate
2406 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2408 PR tree-optimization/84178
2409 * tree-if-conv.c (release_bb_predicate): Remove the
2410 the assertion that the stmts have NULL use_ops.
2411 Discard the statements, asserting that they haven't
2412 yet been added to a BB.
2414 2018-03-08 Richard Biener <rguenther@suse.de>
2416 PR tree-optimization/84746
2417 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2418 (phi_translate): Pass in destination ANTIC_OUT set.
2419 (phi_translate_1): Likewise. For a simplified result lookup
2420 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2421 (phi_translate_set): Adjust.
2422 (do_pre_regular_insertion): Likewise.
2423 (do_pre_partial_partial_insertion): Likewise.
2425 2018-03-08 Martin Liska <mliska@suse.cz>
2427 PR gcov-profile/84735
2428 * doc/gcov.texi: Document usage of profile files.
2429 * gcov-io.h: Document changes in the format.
2431 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2435 * dwarf2out.c (struct dw_line_info_table): Update comments for
2437 (FORCE_RESET_NEXT_VIEW): New.
2438 (FORCE_RESETTING_VIEW_P): New.
2439 (RESETTING_VIEW_P): Check for -1 too.
2440 (ZERO_VIEW_P): Likewise.
2441 (new_line_info_table): Force-reset next view.
2442 (dwarf2out_begin_function): Likewise.
2443 (dwarf2out_source_line): Simplify zero_view_p initialization.
2444 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2445 view directly. Omit view when omitting .loc at line 0.
2447 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2449 PR tree-optimization/84740
2450 * tree-switch-conversion.c (process_switch): Call build_constructors
2451 only if info.phi_count is non-zero.
2453 PR tree-optimization/84739
2454 * tree-tailcall.c (find_tail_calls): Check call arguments against
2455 DECL_ARGUMENTS (current_function_decl) rather than
2456 DECL_ARGUMENTS (func) when checking for tail recursion.
2458 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2460 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2461 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2462 Volker Reichelt's entry and add entries for people that perform
2463 GCC fuzzy testing and report numerous bugs.
2465 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2468 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2469 readonly data in sdata, if that is disabled.
2470 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2471 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2472 -mreadonly-in-sdata option.
2474 2018-03-07 Martin Sebor <msebor@redhat.com>
2476 PR tree-optimization/84468
2477 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2478 basic block when looking for nul assignment.
2480 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2483 * except.h (output_function_exception_table): Adjust prototype.
2484 * except.c (output_function_exception_table): Remove FNNAME parameter
2485 and add SECTION parameter. Ouput one part of the table at a time.
2486 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2487 the first part of the exception table and emit unwind directives.
2488 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2489 (i386_pe_seh_cold_init): Likewise.
2490 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2491 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2492 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2493 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2494 * config/i386/winnt.c: Include except.h.
2495 (struct seh_frame_state): Add reg_offset, after_prologue and
2496 in_cold_section fields.
2497 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2498 (i386_pe_seh_cold_init): New function.
2499 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2500 to seh->in_cold_section.
2501 (seh_emit_push): Record the offset of the push.
2502 (seh_emit_save): Record the offet of the save.
2503 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2504 Test seh->after_prologue to disregard the epilogue.
2505 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2506 (i386_pe_end_cold_function): New function.
2508 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2511 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2512 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2515 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2517 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2518 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2521 * multiple_target.c: Include tree-inline.h and intl.h.
2522 (expand_target_clones): Diagnose and fail if node->definition and
2523 !tree_versionable_function_p (node->decl).
2525 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2527 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2529 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2530 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2531 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2533 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2536 * combine.c (try_combine): Use reg_or_subregno instead of handling
2537 just paradoxical SUBREGs and REGs.
2539 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2541 * config/arc/arc.c (arc_finalize_pic): Remove function.
2542 (arc_must_save_register): We use single base PIC register, remove
2543 checks to save/restore the PIC register.
2544 (arc_expand_prologue): Likewise.
2545 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2547 (arc_verify_short): Likewise.
2548 (arc_attr_type): Likewise.
2549 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2550 (walk_stores): Likewise.
2551 (arc_address_cost): Make it static.
2552 (arc_verify_short): Likewise.
2553 (branch_dest): Likewise.
2554 (arc_attr_type): Likewise.
2555 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2556 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2557 (arc_final_prescan_insn): Remove inserting the nops due to
2558 hardware hazards. It is done in reorg step.
2559 (insn_length_variant_t): Remove.
2560 (insn_length_parameters_t): Likewise.
2561 (arc_insn_length_parameters): Likewise.
2562 (arc_get_insn_variants): Likewise.
2563 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2565 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2568 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2571 PR tree-optimization/84687
2572 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2574 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2576 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2578 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2579 Rename to ppc_speculation_barrier.
2580 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2581 __builtin_ppc_speculation_barrier.
2583 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2586 * combine.c (combine_simplify_rtx): Don't try to simplify if
2587 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2590 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2592 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2593 to 32 bytes when compiling for POWER9.
2595 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2598 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2599 regparm >= 3 with no arg reg available also for calls with
2600 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2603 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2605 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2607 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2610 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2612 2018-03-05 Richard Biener <rguenther@suse.de>
2614 PR tree-optimization/84486
2615 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2616 When inserting a __builtin_assume_aligned call set the LHS
2617 SSA name alignment info accordingly.
2619 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2621 PR tree-optimization/84114
2622 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2623 Avoid reassociation of FLOAT_MODE addition.
2625 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2627 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2628 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2629 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2630 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2631 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2632 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2633 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2635 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2636 __builtin_ia32_wbinvd): New builtins.
2637 (SPECIAL_ARGS2): New.
2638 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2639 (SPECIAL_ARGS2): New.
2640 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2641 (ix86_valid_target_attribute_inner_p): Ditto.
2642 (ix86_init_mmx_sse_builtins): Add special_args2.
2643 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2644 TARGET_WBNOINVD_P): New.
2645 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2646 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2647 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2648 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2649 * config/i386/pconfigintrin.h: New file.
2650 * config/i386/wbnoinvdintrin.h: Ditto.
2651 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2652 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2654 2018-03-05 Richard Biener <rguenther@suse.de>
2656 PR tree-optimization/84670
2657 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2659 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2660 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2661 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2662 doesn't grow if all successors (recursively) were visited at least
2665 2018-03-05 Richard Biener <rguenther@suse.de>
2667 PR tree-optimization/84650
2668 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2669 if executed in the loop pipeline.
2671 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2673 * doc/configfiles.texi (Configuration Files): Move info about
2674 conditionalizing $target-protos.h to...
2675 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2676 differs from $target-protos.h.
2678 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2679 Chung-Ju Wu <jasonwucj@gmail.com>
2681 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2682 * config/nds32/nds32-multiple.md (setmemsi): Define.
2683 * config/nds32/nds32-memory-manipulation.c
2684 (nds32_gen_dup_4_byte_to_word_value): New.
2685 (emit_setmem_word_loop): New.
2686 (emit_setmem_byte_loop): New.
2687 (nds32_expand_setmem_loop): New.
2688 (nds32_expand_setmem_loop_v3m): New.
2689 (nds32_expand_setmem_unroll): New.
2690 (nds32_expand_setmem): New.
2692 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2693 Chung-Ju Wu <jasonwucj@gmail.com>
2695 * config/nds32/nds32-memory-manipulation.c
2696 (nds32_emit_load_store): New.
2697 (nds32_emit_post_inc_load_store): New.
2698 (nds32_emit_mem_move): New.
2699 (nds32_emit_mem_move_block): New.
2700 (nds32_expand_movmemsi_loop_unknown_size): New.
2701 (nds32_expand_movmemsi_loop_known_size): New.
2702 (nds32_expand_movmemsi_loop): New.
2703 (nds32_expand_movmemsi_unroll): New.
2704 (nds32_expand_movmemqi): Rename ...
2705 (nds32_expand_movmemsi): ... to this.
2706 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2707 (movmemsi): ... to this.
2708 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2709 (nds32_expand_movmemsi): ... to this.
2711 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2712 Monk Chiang <sh.chiang04@gmail.com>
2713 Chung-Ju Wu <jasonwucj@gmail.com>
2715 * config/nds32/nds32-protos.h
2716 (nds32_expand_load_multiple): New arguments.
2717 (nds32_expand_store_multiple): Ditto.
2718 (nds32_valid_multiple_load_store): Rename ...
2719 (nds32_valid_multiple_load_store_p): ... to this.
2720 * config/nds32/nds32-memory-manipulation.c
2721 (nds32_expand_load_multiple): Refine implementation.
2722 (nds32_expand_store_multiple): Ditto.
2723 * config/nds32/nds32-multiple.md
2724 (load_multiple): Update nds32_expand_load_multiple interface.
2725 (store_multiple): Update nds32_expand_store_multiple interface.
2726 * config/nds32/nds32-predicates.c
2727 (nds32_valid_multiple_load_store): Rename ...
2728 (nds32_valid_multiple_load_store_p): ... to this and refine
2730 * config/nds32/predicates.md
2731 (nds32_load_multiple_and_update_address_operation): New predicate.
2732 (nds32_store_multiple_and_update_address_operation): New predicate.
2734 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2735 Chung-Ju Wu <jasonwucj@gmail.com>
2737 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2738 (combo): New attribute.
2739 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2741 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2743 * config/nds32/nds32.opt: Change -mcmodel= default value.
2745 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2746 Monk Chiang <sh.chiang04@gmail.com>
2747 Chung-Ju Wu <jasonwucj@gmail.com>
2749 * config/nds32/constants.md (unspec_element): New enum.
2750 * config/nds32/constraints.md (Umw): New constraint.
2751 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2752 * config/nds32/nds32-intrinsic.md: Likewise.
2753 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2754 (nds32_valid_smw_lwm_base_p): New.
2755 (nds32_output_smw_single_word): New.
2756 (nds32_output_lmw_single_word): New.
2757 (nds32_expand_unaligned_load): New.
2758 (nds32_expand_unaligned_store): New.
2759 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2760 (nds32_output_smw_single_word): Declare.
2761 (nds32_output_lmw_single_word): Declare.
2762 (nds32_expand_unaligned_load): Declare.
2763 (nds32_expand_unaligned_store): Declare.
2764 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2765 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2766 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2767 NDS32_BUILTIN_UASTORE_DW.
2768 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2771 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2772 Kito Cheng <kito.cheng@gmail.com>
2773 Chung-Ju Wu <jasonwucj@gmail.com>
2775 * config/nds32/nds32-intrinsic.c
2776 (nds32_expand_builtin_null_ftype_reg): Delete.
2777 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2778 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2779 (nds32_read_argument): New.
2780 (nds32_legitimize_target): Ditto.
2781 (nds32_legitimize_argument): Ditto.
2782 (nds32_check_constant_argument): Ditto.
2783 (nds32_expand_unop_builtin): Ditto.
2784 (nds32_expand_unopimm_builtin): Ditto.
2785 (nds32_expand_binop_builtin): Ditto.
2786 (nds32_builtin_decl_impl): Ditto.
2787 (builtin_description): Ditto.
2788 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2789 (nds32_init_builtins_impl): Ditto.
2790 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2791 (nds32_builtin_decl): New.
2792 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2793 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2795 2018-03-02 Jeff Law <law@redhat.com>
2797 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2798 (redundant_insn, fill_simple_delay_slots): Likewise.
2799 (fill_slots_from_thread): Likewise.
2800 * resource.c (mark_referenced_resources): Likewise.
2801 (mark_set_resources, find_dead_or_set_registers): Likewise.
2803 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2805 * substring-locations.h (format_warning_va): Formatting fix for
2807 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2809 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2810 * substring-locations.c: Include intl.h.
2811 (format_warning_va): Turned into small wrapper around
2812 format_warning_n_va, renamed to ...
2813 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2814 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2816 (format_warning_at_substring_n): New function.
2817 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2818 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2819 format_warning_at_substring with just a shorter name instead of
2820 const function pointer.
2821 (fmtwarn_n): New function.
2822 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2823 appropriate, get rid of all the fmtstr temporaries, move conditionals
2824 with G_() wrapped string literals directly into fmtwarn arguments,
2825 cast dir.len to (int), formatting fixes.
2827 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2829 * doc/invoke.texi: Remove "Cilk Plus" references.
2831 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2832 Richard Biener <rguenther@suse.de>
2835 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2836 for error or warning attributes if CALL_FROM_THUNK_P is set.
2839 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2842 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2843 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2846 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2847 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2849 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2850 instead of -1U in last predictors element's probability member.
2852 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2855 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2856 arguments if they are comparable.
2858 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2860 PR tree-optimization/84634
2861 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2862 masks and masked_loop_p with a single loop_masks, making sure it's
2863 null for bb vectorization.
2865 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2867 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2868 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2869 loop->force_vectorize to check whether there is no alias.
2871 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2874 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2876 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2878 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2879 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2880 instead of a loop around prev_real_insn.
2881 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2885 * config/i386/i386.c (ix86_print_operand): Use conditional
2886 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2889 2018-03-02 Richard Biener <rguenther@suse.de>
2891 PR tree-optimization/84427
2892 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2893 (bitmap_set_subtract_values): Rewrite to handle multiple
2896 (prune_clobbered_mems): Likewise.
2897 (phi_translate): Take edge instead of pred/phiblock.
2898 (phi_translate_1): Likewise.
2899 (phi_translate_set): Likewise. Insert all translated
2900 exprs for a value into the set, keeping possibly multiple
2901 expressions per value.
2902 (compute_antic_aux): Adjust for phi_translate changes.
2903 When intersecting union the expressions and prune those
2904 not in the final value set, keeping possibly multiple
2905 expressions per value. Do not use value-insertion
2906 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2907 all expressions. Add verification that the value-sets
2908 only shrink during iteration.
2909 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2910 (do_pre_regular_insertion): Likewise.
2911 (do_pre_partial_partial_insertion): Likewise.
2913 2018-03-02 Richard Biener <rguenther@suse.de>
2916 * config/darwin.c (saved_debug_info_level): New static global.
2917 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2918 (darwin_asm_lto_end): Restore debug info generation settings.
2920 2018-03-01 Martin Liska <mliska@suse.cz>
2923 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2926 2018-03-01 Richard Biener <rguenther@suse.de>
2929 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2930 type annotation with LTO when debug was disabled at compile-time.
2932 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2934 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2936 (mips_final_postscan_insn): Likewise.
2938 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2940 PR rtl-optimization/84528
2941 * alias.c (init_alias_target): Add commentary.
2942 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2943 a unique base value if the frame pointer is not eliminated
2944 to the stack pointer.
2946 2018-03-01 Tom de Vries <tom@codesourcery.com>
2948 PR rtl-optimization/83327
2949 * lra-int.h (hard_regs_spilled_into): Declare.
2950 * lra.c (hard_regs_spilled_into): Define.
2951 (init_reg_info): Init hard_regs_spilled_into.
2952 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2953 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2954 (process_bb_lives): Handle hard_regs_spilled_into.
2955 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2956 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2958 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2960 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2961 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2962 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2963 * config/rs6000/aix72.h: New file.
2965 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2967 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2968 instead of warning_at with conditional singular and plural messages
2972 * stor-layout.c (update_alignment_for_field): For
2973 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2974 && !DECL_PACKED (field), do the alignment update, just use
2975 only desired_align instead of MAX (type_align, desired_align)
2977 (place_field): Don't do known_align < desired_align handling
2978 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2979 is non-NULL, instead do it after rli->prev_field handling and
2980 only if not within a bitfield word. For DECL_PACKED (field)
2981 use type_align of BITS_PER_UNIT.
2983 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2985 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2986 superfluous parentheses and trailing spaces.
2988 2018-02-28 Richard Biener <rguenther@suse.de>
2990 PR tree-optimization/84584
2991 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2992 SCoPs with fake exit edge.
2994 2018-02-28 Martin Liska <mliska@suse.cz>
2997 * timevar.c (timer::print): Fix format to properly print 100%
3000 2018-02-28 Richard Biener <rguenther@suse.de>
3003 * genmatch.c (capture_info::walk_match): Do not mark
3004 captured expressions without operands as expr_p given
3005 they act more like predicates and should be subject to
3006 "lost tail" side-effect preserving.
3008 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
3010 PR rtl-optimization/81611
3011 * auto-inc-dec.c (attempt_change): Move dead note from
3012 mem_insn if it's the next use of regno
3013 (find_address): Take address use of reg holding
3014 non-incremented value. Add parm to limit search to the named
3016 (merge_in_block): Attempt to use a mem insn that is the next
3017 use of the original regno.
3019 2018-02-27 Martin Sebor <msebor@redhat.com>
3022 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
3023 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
3025 2018-02-27 Martin Sebor <msebor@redhat.com>
3027 PR translation/84207
3028 * diagnostic-core.h (warning_n, error_n, inform_n): Change
3029 n argument to unsigned HOST_WIDE_INT.
3030 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
3031 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
3032 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
3033 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
3035 2018-02-27 Richard Biener <rguenther@suse.de>
3037 PR tree-optimization/84512
3038 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
3039 Do not use the estimate returned from record_stmt_cost for
3040 the scalar iteration cost but sum properly using add_stmt_cost.
3042 2018-02-27 Richard Biener <rguenther@suse.de>
3044 PR tree-optimization/84466
3045 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3046 Adjust last change to less strictly validate use operands.
3048 2018-02-27 Martin Liska <mliska@suse.cz>
3050 PR gcov-profile/84548
3051 * gcov.c (process_file): Allow partial overlap and consider it
3052 also as group functions.
3053 (output_lines): Properly calculate range of lines for a group.
3055 2018-02-27 Martin Liska <mliska@suse.cz>
3057 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
3058 'ggc' suffixes. Change first column width.
3059 (timer::print): Fix formatting of the column.
3061 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
3063 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
3064 preserve inline entry blocks for the sake of debug inline
3065 entry point markers alone.
3066 (remove_unused_locals): Suggest in comments a better place to
3067 force the preservation of inline entry blocks that are
3068 otherwise unused, but do not preserve them.
3070 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3072 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
3074 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3077 * config/i386/constraints.md (Bs): Replace
3078 ix86_indirect_branch_register with
3079 TARGET_INDIRECT_BRANCH_REGISTER.
3081 * config/i386/i386.md (indirect_jump): Likewise.
3082 (tablejump): Likewise.
3083 (*sibcall_memory): Likewise.
3084 (*sibcall_value_memory): Likewise.
3085 Peepholes of indirect call and jump via memory: Likewise.
3086 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
3087 (*sibcall_value_GOT_32): Likewise.
3088 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3089 (GOT_memory_operand): Likewise.
3090 (call_insn_operand): Likewise.
3091 (sibcall_insn_operand): Likewise.
3092 (GOT32_symbol_operand): Likewise.
3093 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
3095 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3097 PR rtl-optimization/83496
3098 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
3099 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
3100 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
3101 redundant insn, if any.
3102 (relax_delay_slots): Likewise.
3103 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
3105 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
3107 PR tree-optimization/83965
3108 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
3109 that grouped statements are part of a reduction chain. Return
3110 true if the statement is not marked as a reduction itself but
3112 (vect_recog_dot_prod_pattern): Don't check whether the statement
3113 is part of a group here.
3114 (vect_recog_sad_pattern): Likewise.
3115 (vect_recog_widen_sum_pattern): Likewise.
3117 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
3120 * final.c (rest_of_clean_state): Also look for calls inside sequences.
3122 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3125 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3127 (ix86_output_indirect_function_return): New prototype.
3128 (ix86_split_simple_return_pop_internal): Likewise.
3129 * config/i386/i386.c (indirect_return_via_cx): New.
3130 (indirect_return_via_cx_bnd): Likewise.
3131 (indirect_thunk_name): Handle return va CX_REG.
3132 (output_indirect_thunk_function): Create alias for
3133 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3134 (ix86_output_indirect_jmp): Remove the bool argument.
3135 (ix86_output_indirect_function_return): New function.
3136 (ix86_split_simple_return_pop_internal): Likewise.
3137 * config/i386/i386.md (*indirect_jump): Don't pass false
3138 to ix86_output_indirect_jmp.
3139 (*tablejump_1): Likewise.
3140 (simple_return_pop_internal): Change it to define_insn_and_split.
3141 Call ix86_split_simple_return_pop_internal to split it for
3143 (simple_return_indirect_internal): Call
3144 ix86_output_indirect_function_return instead of
3145 ix86_output_indirect_jmp.
3147 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3150 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3151 memset and value initialization afterwards.
3153 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3155 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3157 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3160 * common/config/aarch64/aarch64-common.c
3161 (aarch_option_optimization_table[]): Switch
3162 off fomit-frame-pointer
3164 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3165 Chung-Ju Wu <jasonwucj@gmail.com>
3167 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3169 (store_multiple): Ditto.
3171 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3173 * config.gcc: Add --with-cpu support for nds32 target.
3174 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3175 * config/nds32/nds32.opt: Add -mcpu= option.
3177 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3179 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3180 isel=yes): Warn for these deprecated options.
3182 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3184 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3185 ISA_2_5_MASKS_EMBEDDED.
3187 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3189 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3190 p->max as pointers rather than using iterative_hash_expr.
3192 2018-02-23 Carl Love <cel@us.ibm.com>
3194 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3195 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3197 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3198 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3199 P8V_BUILTIN_VEC_VUNSIGNED2.
3201 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3204 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3205 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3207 * lra-constraints.c (curr_insn_transform): Set up
3208 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3210 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3211 * lra-eliminations.c (spill_pseudos): Ditto.
3212 (process_insn_for_elimination): Ditto.
3213 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3214 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3215 LRA_NON_CLOBBERED_ALT.
3217 2018-02-22 Martin Sebor <msebor@redhat.com>
3219 PR tree-optimization/84480
3220 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3221 to maybe_diag_stxncpy_trunc. Call it.
3222 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3223 from gimple_fold_builtin_strcpy. Print inlining stack.
3224 (handle_builtin_stxncpy): Print inlining stack.
3225 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3227 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3230 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3231 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3232 and -fcheck-pointer-bounds are used together.
3233 (indirect_thunk_prefix): New enum.
3234 (indirect_thunk_need_prefix): New function.
3235 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3236 "_nt" instead of "_bnd" for NOTRACK prefix.
3237 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3238 (output_indirect_thunk_function): Likewise.
3240 (ix86_code_end): Update output_indirect_thunk_function calls.
3241 (ix86_output_indirect_branch_via_reg): Replace
3242 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3243 (ix86_output_indirect_branch_via_push): Likewise.
3244 (ix86_output_function_return): Likewise.
3245 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3246 incompatible with -fcf-protection=branch and
3247 -fcheck-pointer-bounds.
3249 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3252 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3253 Change gcc_assert call to output_operand_lossage.
3255 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3257 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3259 2018-02-22 DJ Delorie <dj@redhat.com>
3260 Sebastian Perta <sebastian.perta@renesas.com>
3261 Oleg Endo <olegendo@gcc.gnu.org>
3263 * config/rx/rx.c (rx_rtx_costs): New function.
3264 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3266 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3268 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3270 2018-02-22 Martin Liska <mliska@suse.cz>
3273 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3274 Add "native" as a possible value.
3276 2018-02-22 Martin Liska <mliska@suse.cz>
3279 * config/i386/i386.c (ix86_option_override_internal):
3280 Add "native" as a possible value for -march and -mtune.
3282 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3285 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3286 to all type variants.
3288 PR tree-optimization/84503
3289 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3290 width as info->bitpos + info->bitsize - start.
3291 (merged_store_group::merge_overlapping): Simplify width computation.
3292 (check_no_overlap): New function.
3293 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3294 start + width and last_order of the group, fail if check_no_overlap
3296 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3297 to group if check_no_overlap fails.
3299 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3301 * config/rs6000/altivec.md: Delete contraint arguments to
3302 define_expand, define_split, and define_peephole2, and in
3303 define_insn_and_split if always unused.
3304 * config/rs6000/darwin.md: Ditto.
3305 * config/rs6000/dfp.md: Ditto.
3306 * config/rs6000/rs6000.md: Ditto.
3307 * config/rs6000/sync.md: Ditto.
3308 * config/rs6000/vector.md: Ditto.
3309 * config/rs6000/vsx.md: Ditto.
3311 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3313 * config/rs6000/altivec.md: Write output control strings as braced
3314 blocks instead of double-quoted strings.
3315 * config/rs6000/darwin.md: Ditto.
3316 * config/rs6000/rs6000.md: Ditto.
3317 * config/rs6000/vector.md: Ditto.
3318 * config/rs6000/vsx.md: Ditto.
3320 2018-02-21 Jason Merrill <jason@redhat.com>
3322 PR c++/84314 - ICE with templates and fastcall attribute.
3323 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3325 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3327 * ipa-cp.c (determine_versionability): Fix comment typos.
3329 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3332 * ipa-cp.c (determine_versionability): Do not version functions caling
3335 2018-02-21 Martin Liska <mliska@suse.cz>
3338 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3339 Add "native" as a possible value.
3340 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3341 the macro when native cpu detection is available.
3343 2018-02-21 Martin Liska <mliska@suse.cz>
3346 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3347 Add "native" as a possible value.
3348 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3349 when native cpu detection is available.
3351 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3352 Martin Sebor <msebor@redhat.com>
3354 PR tree-optimization/84478
3355 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3357 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3358 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3359 support which is conservatively correct, for 2 only stay conservative
3360 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3361 argument to the 2 argument get_range_strlen, adjust 6 arg
3362 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3364 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3365 (gimple_fold_builtin_strlen): Pass true as last argument to
3368 2018-02-20 Martin Sebor <msebor@redhat.com>
3371 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3372 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3373 (builtin_memref::builtin_memref): Factor out parts into
3374 set_base_and_offset and call it.
3376 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3379 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3380 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3381 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3382 search at the associated MODE_INT.
3384 2018-02-20 Jeff Law <law@redhat.com>
3387 PR tree-optimization/81592
3389 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3390 for range data rather than using global data.
3391 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3392 range data rather than using global data.
3393 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3394 pass it to children as needed.
3395 (struct directive::fmtresult): Similarly.
3396 (struct directive::set_width): Similarly.
3397 (struct directive::set_precision): Similarly.
3398 (format_integer, format_directive, parse_directive): Similarly.
3399 (format_none): Accept unnamed vr_values parameter.
3400 (format_percent, format_floating, format_character): Similarly.
3401 (format_string, format_plain): Similarly.
3402 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3403 the EVRP range analyzer for range data rather than using global data.
3404 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3405 gimple-ssa-evrp-analyze.h
3406 (class sprintf_dom_walker): Add after_dom_children member function.
3407 Add evrp_range_analyzer member.
3408 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3409 range analyzer as needed.
3410 (sprintf_dom_walker::after_dom_children): New member function.
3411 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3413 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3414 (evrp_range_analyzer::pop_to_marker): Likewise.
3416 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3418 PR tree-optimization/84419
3419 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3420 with the required type if its current type is compatible but
3423 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3426 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3427 after vectorization.
3429 2018-02-20 Martin Liska <mliska@suse.cz>
3432 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3433 possible values if we don't have a hint.
3435 2018-02-20 Martin Liska <mliska@suse.cz>
3439 * final.c (shorten_branches): Build align_tab array with one
3441 * opts.c (finish_options): Add alignment option limit check.
3442 (MAX_CODE_ALIGN): Likewise.
3443 (MAX_CODE_ALIGN_VALUE): Likewise.
3444 * doc/invoke.texi: Document maximum allowed option value for
3445 all -falign-* options.
3447 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3450 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3451 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3452 * var-tracking.c (emit_note_insn_var_location): Remove all references
3453 to NOTE_INSN_CALL_ARG_LOCATION.
3454 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3455 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3456 Use copy_rtx_if_shared.
3457 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3458 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3459 (dwarf2out_var_location): Remove handling of
3460 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3462 * final.c (final_scan_insn): Remove all references to
3463 NOTE_INSN_CALL_ARG_LOCATION.
3464 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3465 before dumping final insns.
3466 * except.c (emit_note_eh_region_end): Remove all references to
3467 NOTE_INSN_CALL_ARG_LOCATION.
3468 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3469 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3470 * config/arc/arc.c (hwloop_optimize): Likewise.
3471 * config/arm/arm.c (create_fix_barrier): Likewise.
3472 * config/s390/s390.c (s390_chunkify_start): Likewise.
3473 * config/sh/sh.c (find_barrier): Likewise.
3474 * config/i386/i386.c (rest_of_insert_endbranch,
3475 ix86_seh_fixup_eh_fallthru): Likewise.
3476 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3477 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3478 * config/frv/frv.c (frv_function_prologue): Likewise.
3479 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3481 (note_outside_basic_block_p): Remove all references to
3482 NOTE_INSN_CALL_ARG_LOCATION.
3483 * gengtype.c (adjust_field_rtx_def): Likewise.
3484 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3486 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3487 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3490 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3493 PR tree-optimization/84452
3494 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3495 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3498 2018-02-19 Martin Liska <mliska@suse.cz>
3501 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3503 2018-02-19 Martin Liska <mliska@suse.cz>
3504 Richard Sandiford <richard.sandiford@linaro.org>
3506 PR tree-optimization/82491
3507 * gimple-fold.c (get_base_constructor): Make earlier bail out
3510 2018-02-19 Carl Love <cel@us.ibm.com>
3512 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3513 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3515 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3516 P8V_BUILTIN_VEC_NEG.
3518 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3520 * config/rl78/rl78.md (movdf): New define expand.
3522 2018-02-19 Martin Liska <mliska@suse.cz>
3525 * doc/invoke.texi: Fix typo.
3526 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3528 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3530 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3531 handle rs6000_single_float and rs6000_double_float specially for
3534 2018-02-16 Jeff Law <law@redhat.com>
3536 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3539 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3542 * ipa-inline.c (inline_small_functions): Fix a typo.
3544 2018-02-16 Nathan Sidwell <nathan@acm.org>
3546 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3548 2018-02-16 Carl Love <cel@us.ibm.com>
3550 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3551 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3552 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3553 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3554 expansion to P8V_BUILTIN_VEC_FLOAT2.
3556 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3558 PR rtl-optimization/70023
3559 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3560 src_regno into account.
3562 2018-02-16 Carl Love <cel@us.ibm.com>
3564 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3565 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3566 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3567 * config/rs6000/rs6000.c: Remove case statements for
3568 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3569 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3570 and P9V_BUILTIN_VEC_VINSERT4B.
3571 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3572 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3573 * config/rs6000/vsx.md:
3574 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3577 2018-02-16 Carl Love <cel@us.ibm.com>
3579 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3581 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3583 * config/rs6000/rs6000-c.c: Add the definitions for
3584 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3585 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3586 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3587 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3588 definition for insert4b and define insn *insert3b_internal.
3589 * doc/extend.texi: Add documentation for vec_extract4b.
3591 2018-02-16 Nathan Sidwell <nathan@acm.org>
3593 * doc/extend.texi (Backwards Compatibility): Mention friend
3594 injection. Note for-scope is deprecated.
3595 * doc/invoke.texi (-ffriend-injection): Deprecate.
3597 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3599 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3600 that moved to I2, also allow destinations that are a paradoxical
3601 subreg (instead of a normal reg).
3603 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3606 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3609 2018-02-16 Richard Biener <rguenther@suse.de>
3611 PR tree-optimization/84037
3612 PR tree-optimization/84016
3614 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3615 Adjust vec_construct for the fact we need additional higher latency
3616 128bit inserts for AVX256 and AVX512 vector builds.
3617 (ix86_add_stmt_cost): Scale vector construction cost for
3620 2018-02-16 Richard Biener <rguenther@suse.de>
3622 PR tree-optimization/84417
3623 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3624 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3625 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3627 2018-02-16 Martin Liska <mliska@suse.cz>
3630 * internal-fn.def (ASAN_CHECK): Set proper flags.
3631 (ASAN_MARK): Likewise.
3633 2018-02-16 Julia Koval <julia.koval@intel.com>
3635 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3636 from PTA_CANNONLAKE.
3638 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3641 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3642 Use ++iter rather than iter++ for std::list iterators.
3643 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3644 defer deleting them until all nodes in the forest are processed. Do
3645 free even leaf nodes. Change to_process into auto_vec.
3648 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3649 * vec.h (vec_default_construct): Use memset instead of placement new
3650 if BROKEN_VALUE_INITIALIZATION is defined.
3651 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3652 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3655 PR rtl-optimization/83723
3656 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3657 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3658 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3660 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3662 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3664 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3666 PR rtl-optimization/81443
3667 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3668 from inner REGs to paradoxical SUBREGs.
3670 2018-02-16 Richard Biener <rguenther@suse.de>
3672 PR tree-optimization/84399
3673 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3674 For operands we can analyze at their definition make sure we can
3675 analyze them at each use as well.
3677 2018-02-16 Richard Biener <rguenther@suse.de>
3679 PR tree-optimization/84190
3680 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3681 volatile accesses if the decl isn't volatile.
3683 2018-02-15 Jason Merrill <jason@redhat.com>
3685 PR c++/84314 - ICE with templates and fastcall attribute.
3686 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3687 TYPE_CANONICAL on an existing type.
3689 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3691 PR tree-optimization/84383
3692 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3693 dstoff nor call operand_equal_p if dstbase is NULL.
3695 PR tree-optimization/84334
3696 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3697 also a CONSTANT_CLASS_P, punt.
3699 2018-02-14 Jim Wilson <jimw@sifive.com>
3701 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3702 first SMALL_OPERAND check. New local min_second_step. Move assert
3703 to where locals are set. Add TARGET_RVC support.
3704 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3706 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3708 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3710 2018-02-14 Martin Sebor <msebor@redhat.com>
3712 PR tree-optimization/83698
3713 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3714 arrays constrain the offset range to their bounds.
3715 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3716 (builtin_access::overlap): Avoid setting the size of overlap if it's
3718 (maybe_diag_overlap): Also consider arrays when deciding what values
3719 of offsets to include in diagnostics.
3721 2018-02-14 Martin Sebor <msebor@redhat.com>
3724 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3725 that correspond to the kind of a declaration.
3727 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3730 * config/pa/pa.md: Load address of PIC label using the linkage table
3731 if the label is nonlocal.
3733 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3735 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3736 warning message if user requests -maltivec=be.
3737 * doc/invoke.texi: Document deprecation of -maltivec=be.
3739 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3742 * config/rs6000/rs6000-c.c: Update definitions for
3743 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3744 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3746 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3749 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3750 add _get_ssp intrinsics. Remove argument from
3751 __builtin_ia32_rdssp[d|q].
3752 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3753 * config/i386/i386-builtin.def: Remove argument from
3754 __builtin_ia32_rdssp[d|q].
3755 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3756 ix86_expand_special_args_builtin for _rdssp[d|q].
3757 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3758 Clear register before usage.
3759 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3760 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3762 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3764 PR tree-optimization/84357
3765 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3766 operand 1 of an ARRAY_REF too.
3768 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3771 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3772 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3774 (set_of_reg): New struct.
3775 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3776 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3777 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3779 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3780 Split into bitclr, bitset, bitinvert patterns if appropriate.
3781 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3782 use rx_fuse_in_memory_bitop.
3783 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3784 to named insn, correct maximum insn length.
3786 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3789 * machmode.def: Define a complex mode for PARTIAL_INT.
3790 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3792 * doc/rtl.texi: Document CSPImode.
3793 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3795 (msp430_hard_regno_nregs_with_padding): Likewise.
3797 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3800 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3802 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3804 PR rtl-optimization/84169
3805 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3806 we generated a parallel as new i3 and we split that to new i2 and i3
3807 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3808 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3809 those to i2, not i1. Partially rewrite this scan code.
3811 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3814 * stor-layout.c (place_field): For variable length fields, adjust
3815 offset_align afterwards not just based on the field's alignment,
3816 but also on the size.
3819 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3820 of exps and logs in the use_exp2 case.
3822 2018-02-13 Jeff Law <law@redhat.com>
3824 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3827 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3830 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3834 * common.opt (gas-loc-support, gas-locview-support): New.
3835 (ginline-points, ginternal-reset-location-views): New.
3836 * doc/invoke.texi: Document them. Use @itemx where intended.
3837 (gvariable-location-views): Adjust.
3838 * target.def (reset_location_view): New.
3839 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3840 (TARGET_RESET_LOCATION_VIEW): New.
3841 * doc/tm.texi: Rebuilt.
3842 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3843 (dwarf2out_default_as_locview_support): New.
3844 (output_asm_line_debug_info): Use option variables.
3845 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3846 (output_loc_list): Likewise.
3847 (add_high_low_attributes): Check option variables.
3848 Don't output entry view attribute in strict mode.
3849 (gen_inlined_subroutine_die): Check option variables.
3850 (dwarf2out_inline_entry): Likewise.
3851 (init_sections_and_labels): Likewise.
3852 (dwarf2out_early_finish): Likewise.
3853 (maybe_reset_location_view): New, from...
3854 (dwarf2out_var_location): ... here. Call it.
3855 * debug.h (dwarf2out_default_as_loc_support): Declare.
3856 (dwarf2out_default_as_locview_support): Declare.
3857 * hooks.c (hook_int_rtx_insn_0): New.
3858 * hooks.h (hook_int_rtx_insn_0): Declare.
3859 * toplev.c (process_options): Take -gas-loc-support and
3860 -gas-locview-support from dwarf2out. Enable
3861 -gvariable-location-views by default only with locview
3862 assembler support. Enable -ginternal-reset-location-views by
3863 default only if the target defines the corresponding hook.
3864 Enable -ginline-points by default if location views are
3865 enabled; force it disabled if statement frontiers are
3867 * tree-inline.c (expand_call_inline): Check option variables.
3868 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3870 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3872 PR tree-optimization/84321
3873 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3874 handling. Also check whether the anti-range contains any values
3875 that satisfy the mask; switch to a VR_RANGE if not.
3877 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3880 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3882 2018-02-13 Martin Jambor <mjambor@suse.cz>
3885 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3886 of call statements, also set location of a load to a temporary.
3888 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3890 * config/rl78/rl78.c (add_vector_labels): New function.
3891 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3892 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3893 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3894 which checks that no arguments are passed.
3895 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3896 * doc/extend.texi: Documentation for the new attribute.
3898 2018-02-13 Andreas Schwab <schwab@suse.de>
3900 * config/riscv/linux.h (CPP_SPEC): Define.
3902 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3905 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3906 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3907 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3908 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3909 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3910 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3911 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3912 temporarily for AES and PCLMUL builtins.
3914 PR tree-optimization/84339
3915 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3916 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3920 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3921 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3922 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3924 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3926 * omp-simd-clone.h: New file.
3927 * omp-simd-clone.c: Include omp-simd-clone.h.
3928 (expand_simd_clones): No longer static.
3929 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3930 cgraph.h and omp-simd-clone.h.
3931 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3932 (vect_recog_widen_shift_pattern): Formatting fix.
3933 (vect_pattern_recog_1): Don't check optab for calls.
3936 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3937 operands[2] into a REG before using gen_lowpart on it.
3939 2018-02-12 Jeff Law <law@redhat.com>
3942 * config/sh/sh.c (find_barrier): Consider a sibling call
3945 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3946 successfully back substituting a reg.
3948 2018-02-12 Richard Biener <rguenther@suse.de>
3950 PR tree-optimization/84037
3951 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3952 parameter, move visited init to caller.
3953 (vect_slp_analyze_operations): Separate cost from validity
3954 check, initialize visited once for all instances.
3955 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3957 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3959 (vect_model_promotion_demotion_cost): Likewise.
3960 (vectorizable_bswap): Guard cost modeling with !slp_node
3961 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3963 (vectorizable_call): Likewise.
3964 (vectorizable_conversion): Likewise.
3965 (vectorizable_assignment): Likewise.
3966 (vectorizable_shift): Likewise.
3967 (vectorizable_operation): Likewise.
3968 (vectorizable_store): Likewise.
3969 (vectorizable_load): Likewise.
3970 (vectorizable_condition): Likewise.
3971 (vectorizable_comparison): Likewise.
3973 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3976 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3977 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3978 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3980 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3983 * params.def (inline-min-speedup): Increase from 8 to 15.
3984 (max-inline-insns-auto): Decrease from 40 to 30.
3985 * ipa-split.c (consider_split): Add some buffer for function to
3986 be considered inlining candidate.
3987 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3990 2018-02-12 Richard Biener <rguenther@suse.de>
3992 PR tree-optimization/84037
3993 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3994 matched stmts if we cannot swap the non-matched ones.
3996 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3998 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3999 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
4000 _mm_maskz_scalef_round_ss): New intrinsics.
4001 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
4002 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
4003 __builtin_ia32_scalefss_round): Remove.
4004 (__builtin_ia32_scalefsd_mask_round,
4005 __builtin_ia32_scalefss_mask_round): New intrinsics.
4006 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
4007 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
4008 ((match_operand:VF_128 2 "<round_nimm_predicate>"
4009 "<round_constraint>")): Changed to ...
4010 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
4011 "<round_scalar_constraint>")): ... this.
4012 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
4013 %0, %1, %2<round_op3>}"): Changed to ...
4014 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
4015 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
4016 %2<round_scalar_mask_op3>}"): ... this.
4017 * config/i386/subst.md (round_scalar_nimm_predicate): New.
4019 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
4021 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
4022 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
4023 (_mm_maskz_sqrt_round_ss): New intrinsics.
4024 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
4025 (__builtin_ia32_sqrtsd_mask_round)
4026 (__builtin_ia32_sqrtss_mask_round): New builtins.
4027 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
4028 (__builtin_ia32_sqrtss_round): Remove.
4029 (__builtin_ia32_sqrtsd_mask_round)
4030 (__builtin_ia32_sqrtss_mask_round): New builtins.
4031 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
4032 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
4033 ((match_operand:VF_128 1 "vector_operand"
4034 "xBm,<round_constraint>")): Changed to ...
4035 ((match_operand:VF_128 1 "vector_operand"
4036 "xBm,<round_scalar_constraint>")): ... this.
4037 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
4038 %0, %2, %<iptr>1<round_op3>}): Changed to ...
4039 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
4040 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
4041 %<iptr>1<round_scalar_mask_op3>}): ... this.
4042 ((set_attr "prefix" "<round_prefix>")): Changed to ...
4043 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
4045 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
4048 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
4049 Cast vec_cmpeq result to correct type.
4050 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
4051 Cast vec_cmpgt result to correct type.
4053 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
4055 * final.c (final_scan_insn_1): Renamed from...
4056 (final_scan_insn): ... this. New wrapper, to recover
4057 seen from the outermost call in recursive ones.
4058 * config/sparc/sparc.c (output_return): Drop seen from call.
4059 (output_sibcall): Likewise.
4060 * config/visium/visium.c (output_branch): Likewise.
4062 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
4064 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
4067 2018-02-10 Alan Modra <amodra@gmail.com>
4070 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
4071 Specify LR as an input.
4073 2018-02-10 Jakub Jelinek <jakub@redhat.com>
4076 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
4077 remove_member_access_dummy_vars): New functions.
4078 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
4079 lower_omp_1, execute_lower_omp): Use them.
4081 PR rtl-optimization/84308
4082 * shrink-wrap.c (spread_components): Release todo vector.
4084 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
4086 PR rtl-optimization/57193
4087 * ira-color.c (struct allocno_color_data): Add member
4088 conflict_allocno_hard_prefs.
4089 (update_conflict_allocno_hard_prefs): New.
4090 (bucket_allocno_compare_func): Add a preference based on
4091 conflict_allocno_hard_prefs.
4092 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
4093 (color_allocnos): Remove a dead code. Initiate
4094 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
4096 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4099 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
4100 constraint from =wa to wa. Avoid a subreg on the output operand,
4101 instead use a pseudo and subreg it in a move.
4102 (p9_xxbrd_<mode>): Changed to ...
4103 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
4104 (p9_xxbrd_v2df): New expander.
4105 (p9_xxbrw_<mode>): Changed to ...
4106 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
4107 (p9_xxbrw_v4sf): New expander.
4109 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4111 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
4113 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
4116 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
4117 multiply in 32-bit mode.
4118 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
4119 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
4122 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4124 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4125 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4126 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4127 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4129 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4132 * dwarf2out.c (is_trivial_indirect_ref): New function.
4133 (dwarf2out_late_global_decl): Do not generate a location
4134 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4135 and that are not defined in the current unit.
4137 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4139 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4140 instead of a libcall for UNORDERED.
4142 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4145 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4146 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4148 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4151 * config/s390/s390.c (s390_set_current_function): Invoke
4152 s390_indirect_branch_settings also if fndecl didn't change.
4154 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4156 * config/rs6000/rs6000.md (blockage): Set length to zero.
4158 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4160 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4162 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4165 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4166 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4170 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4171 PARALLEL incoming that failed vt_get_decl_and_offset check.
4174 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4175 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4177 (get_variable_section): For decls in named .bss* sections pass true as
4178 second argument to bss_initializer_p.
4180 2018-02-09 Marek Polacek <polacek@redhat.com>
4181 Jakub Jelinek <jakub@redhat.com>
4184 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4185 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4186 Sync some changes from cxx_fold_indirect_ref.
4188 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4190 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4192 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4193 (BLOCK_INLINE_ENTRY_LABEL): New.
4194 (dwarf2out_var_location): Disregard inline entry markers.
4195 (inline_entry_data): New struct.
4196 (inline_entry_data_hasher): New hashtable type.
4197 (inline_entry_data_hasher::hash): New.
4198 (inline_entry_data_hasher::equal): New.
4199 (inline_entry_data_table): New variable.
4200 (add_high_low_attributes): Add DW_AT_entry_pc and
4201 DW_AT_GNU_entry_view attributes if a pending entry is found
4202 in inline_entry_data_table. Add old entry_pc attribute only
4203 if debug nonbinding markers are disabled.
4204 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4205 markers are enabled.
4206 (block_within_block_p, dwarf2out_inline_entry): New.
4207 (dwarf2out_finish): Check that no entries remained in
4208 inline_entry_data_table.
4209 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4210 (final_scan_insn, notice_source_line): Likewise.
4211 (rest_of_clean_state): Skip inline entry markers.
4212 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4214 * gimple.c (gimple_build_debug_inline_entry): New.
4215 * gimple.h (enum gimple_debug_subcode): Add
4216 GIMPLE_DEBUG_INLINE_ENTRY.
4217 (gimple_build_debug_inline_entry): Declare.
4218 (gimple_debug_inline_entry_p): New.
4219 (gimple_debug_nonbind_marker_p): Adjust.
4220 * insn-notes.def (INLINE_ENTRY): New.
4221 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4222 inline entry marker notes.
4223 (print_insn): Likewise.
4224 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4225 (INSN_DEBUG_MARKER_KIND): Likewise.
4226 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4227 * tree-inline.c (expand_call_inline): Build and insert
4228 debug_inline_entry stmt.
4229 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4230 inline entry blocks early, if nonbind markers are enabled.
4231 (dump_scope_block): Dump fragment info.
4232 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4233 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4234 (gimple_build_debug_inline_entry): New.
4235 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4236 Enable/disable inline entry points too.
4237 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4238 (DEBUG_INSN): Describe inline entry markers.
4240 * common.opt (gvariable-location-views): New.
4241 (gvariable-location-views=incompat5): New.
4242 * config.in: Rebuilt.
4243 * configure: Rebuilt.
4244 * configure.ac: Test assembler for view support.
4245 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4246 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4247 * dwarf2out.c (var_loc_view): New typedef.
4248 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4249 (dwarf2out_locviews_in_attribute): New.
4250 (dwarf2out_locviews_in_loclist): New.
4251 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4252 (enum dw_line_info_opcode): Add LI_adv_address.
4253 (struct dw_line_info_table): Add view.
4254 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4255 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4256 (zero_view_p): New variable.
4257 (ZERO_VIEW_P): New macro.
4258 (output_asm_line_debug_info): New.
4259 (struct var_loc_node): Add view.
4260 (add_AT_view_list, AT_loc_list): New.
4261 (add_var_loc_to_decl): Add view param. Test it against last.
4262 (new_loc_list): Add view params. Record them.
4263 (AT_loc_list_ptr): Handle loc and view lists.
4264 (view_list_to_loc_list_val_node): New.
4265 (print_dw_val): Handle dw_val_class_view_list.
4266 (size_of_die): Likewise.
4267 (value_format): Likewise.
4268 (loc_list_has_views): New.
4269 (gen_llsym): Set vl_symbol too.
4270 (maybe_gen_llsym, skip_loc_list_entry): New.
4271 (dwarf2out_maybe_output_loclist_view_pair): New.
4272 (output_loc_list): Output view list or entries too.
4273 (output_view_list_offset): New.
4274 (output_die): Handle dw_val_class_view_list.
4275 (output_dwarf_version): New.
4276 (output_compilation_unit_header): Use it.
4277 (output_skeleton_debug_sections): Likewise.
4278 (output_rnglists, output_line_info): Likewise.
4279 (output_pubnames, output_aranges): Update version comments.
4280 (output_one_line_info_table): Output view numbers in asm comments.
4281 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4282 Call maybe_gen_llsym.
4283 (loc_list_from_tree_1): Adjust.
4284 (add_AT_location_description): Create view list attribute if
4285 needed, check it's absent otherwise.
4286 (convert_cfa_to_fb_loc_list): Adjust.
4287 (maybe_emit_file): Call output_asm_line_debug_info for test.
4288 (dwarf2out_var_location): Reset views as needed. Precompute
4289 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4290 attribute. Set view.
4291 (new_line_info_table): Reset next view.
4292 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4293 (dwarf2out_source_line): Likewise. Output view resets and labels to
4294 the assembler, or select appropriate line info opcodes.
4295 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4296 (optimize_string_length): Catch it. Adjust.
4297 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4298 dw_val_class_view_list, and remove it if no longer needed.
4299 (hash_loc_list): Hash view numbers.
4300 (loc_list_hasher::equal): Compare them.
4301 (optimize_location_lists): Check whether a view list symbol is
4302 needed, and whether the locview attribute is present, and
4303 whether they match. Remove the locview attribute if no longer
4305 (index_location_lists): Call skip_loc_list_entry for test.
4306 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4307 Use output_dwarf_version.
4308 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4309 (struct dw_val_node): Add val_view_list.
4310 * final.c (SEEN_NEXT_VIEW): New.
4311 (set_next_view_needed): New.
4312 (clear_next_view_needed): New.
4313 (maybe_output_next_view): New.
4314 (final_start_function): Rename to...
4315 (final_start_function_1): ... this. Take pointer to FIRST,
4316 add SEEN parameter. Emit param bindings in the initial view.
4317 (final_start_function): Reintroduce SEEN-less interface.
4318 (final): Rename to...
4319 (final_1): ... this. Take SEEN parameter. Output final pending
4320 next view at the end.
4321 (final): Reintroduce seen-less interface.
4322 (final_scan_insn): Output pending next view before switching
4323 sections or ending a block. Mark the next view as needed when
4324 outputting variable locations. Notify debug backend of section
4325 changes, and of location view changes.
4326 (rest_of_handle_final): Adjust.
4327 * toplev.c (process_options): Autodetect value for debug variable
4328 location views option. Warn on incompat5 without -gdwarf-5.
4329 * doc/invoke.texi (gvariable-location-views): New.
4330 (gvariable-location-views=incompat5): New.
4331 (gno-variable-location-views): New.
4333 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4335 PR tree-optimization/84136
4336 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4337 that the result of find_edge is non-NULL.
4339 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4342 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4343 storing integer register in SImode. Fix cost of 256 and 512
4344 byte aligned SSE register store.
4346 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4348 * config/i386/i386.c (ix86_multiplication_cost): Fix
4349 multiplication cost for TARGET_AVX512DQ.
4351 2018-02-08 Marek Polacek <polacek@redhat.com>
4353 PR tree-optimization/84238
4354 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4357 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4359 PR tree-optimization/84265
4360 * tree-vect-stmts.c (vectorizable_store): Don't treat
4361 VMAT_CONTIGUOUS accesses as grouped.
4362 (vectorizable_load): Likewise.
4364 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4366 PR tree-optimization/81635
4367 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4368 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4369 (test_round_for_mask): New functions.
4370 (wide_int_cc_tests): Call test_round_for_mask.
4371 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4372 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4373 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4374 range returned by get_range_info.
4376 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4379 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4380 * symtab.c: Include builtins.h
4381 (symtab_node::output_to_lto_symbol_table_p): Move here
4382 from lto-streamer-out.c:output_symbol_p.
4383 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4384 (output_symbol_p): Move all logic to symtab.c
4385 (produce_symtab): Update.
4387 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4389 * config/s390/s390-opts.h (enum indirect_branch): Define.
4390 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4391 (s390_indirect_branch_via_thunk)
4392 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4393 (enum s390_indirect_branch_type): Define.
4394 * config/s390/s390.c (struct s390_frame_layout, struct
4395 machine_function): Remove.
4396 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4397 (indirect_branch_table_label_no, indirect_branch_table_name):
4399 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4400 (enum s390_indirect_branch_option): Define.
4401 (s390_return_addr_from_memory): New function.
4402 (s390_handle_string_attribute): New function.
4403 (s390_attribute_table): Add new attribute handler.
4404 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4405 (s390_indirect_branch_via_thunk): New function.
4406 (s390_indirect_branch_via_inline_thunk): New function.
4407 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4408 sibling call optimization for non z10 compiles.
4409 (s390_emit_call): Force indirect branch target to be a single
4410 register. Add r1 clobber for non-z10 compiles.
4411 (s390_emit_epilogue): Emit return jump via return_use expander.
4412 (s390_reorg): Handle JUMP_INSNs as execute targets.
4413 (s390_option_override_internal): Perform validity checks for the
4414 new command line options.
4415 (s390_indirect_branch_attrvalue): New function.
4416 (s390_indirect_branch_settings): New function.
4417 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4418 (s390_output_indirect_thunk_function): New function.
4419 (s390_code_end): Implement target hook.
4420 (s390_case_values_threshold): Implement target hook.
4421 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4423 * config/s390/s390.h (struct s390_frame_layout)
4424 (struct machine_function): Move here from s390.c.
4425 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4426 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4427 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4428 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4429 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4430 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4431 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4432 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4433 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4434 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4435 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4436 (mnemonic attribute): Add values which aren't recognized
4438 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4439 pattern for branch conversion. Fix mnemonic attribute.
4440 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4441 indirect branch via thunk if requested.
4442 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4443 ("*indirect_jump"): Disable for branch conversion using out of
4445 ("indirect_jump_via_thunk<mode>_z10")
4446 ("indirect_jump_via_thunk<mode>")
4447 ("indirect_jump_via_inlinethunk<mode>_z10")
4448 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4449 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4450 ("casesi_jump_via_inlinethunk<mode>_z10")
4451 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4452 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4453 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4454 ("*indirect2_jump"): Disable for branch conversion.
4455 ("casesi_jump"): Turn into expander and expand patterns for branch
4457 ("return_use"): New expander.
4458 ("*return"): Emit return via thunk and rename it to ...
4459 ("*return<mode>"): ... this one.
4460 * config/s390/s390.opt: Add new options and and enum for the
4463 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4465 * lra-constraints.c (match_reload): Unconditionally use
4466 gen_lowpart_SUBREG, rather than selecting between that
4467 and equivalent gen_rtx_SUBREG code.
4469 2018-02-08 Richard Biener <rguenther@suse.de>
4471 PR tree-optimization/84233
4472 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4473 changed flag instead of boguously re-using phi_inserted.
4475 2018-02-08 Martin Jambor <mjambor@suse.cz>
4477 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4478 static local variables.
4480 2018-02-08 Richard Biener <rguenther@suse.de>
4482 PR tree-optimization/84278
4483 * tree-vect-stmts.c (vectorizable_store): When looking for
4484 smaller vector types to perform grouped strided loads/stores
4485 make sure the mode is supported by the target.
4486 (vectorizable_load): Likewise.
4488 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4490 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4491 Increase LDP/STP opportunities by adding adjacent callee-saves.
4493 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4495 PR rtl-optimization/84068
4496 PR rtl-optimization/83459
4497 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4499 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4501 PR tree-optimization/84224
4502 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4503 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4506 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4509 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4510 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4511 count, remove one USE.
4513 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4515 * doc/install.texi (Configuration): Document the
4516 --with-long-double-format={ibm,ieee} PowerPC configuration
4520 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4521 Convert from define_expand to be define_insn_and_split. Rework
4522 float/double/_Float128 conversions to QI/HI/SImode to work with
4523 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4524 conversions to QI/HImode types did a store and then a load to
4525 truncate the value. For conversions to VSX registers, don't split
4526 the insn, instead emit the code directly. Use the code iterator
4527 any_fix to combine signed and unsigned conversions.
4528 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4529 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4530 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4531 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4532 (fix_<mode>di2_hw): Likewise.
4533 (fixuns_<mode>di2_hw): Likewise.
4534 (fix_<mode>si2_hw): Likewise.
4535 (fixuns_<mode>si2_hw): Likewise.
4536 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4537 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4538 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4539 fix<uns>_trunc<SFDF:mode>si2_p8.
4540 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4542 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4543 (fix<uns>_<mode>_mem): Likewise.
4544 (fctiw<u>z_<mode>_mem): Likewise.
4545 (fix<uns>_<mode>_mem): Likewise.
4546 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4547 the register allocator from doing a direct move to the GPRs to do
4548 a store, and instead use the ISA 3.0 store byte/half-word from
4549 vector register instruction. For IEEE 128-bit floating point,
4550 also optimize stores of 32-bit ints.
4551 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4553 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4555 * genextract.c (push_pathstr_operand): New function to support
4557 (walk_rtx): Call push_pathstr_operand.
4558 (print_path): Support [a-zA-Z].
4560 2018-02-07 Richard Biener <rguenther@suse.de>
4562 PR tree-optimization/84037
4563 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4564 (cse_and_gimplify_to_preheader): Declare.
4565 (vect_get_place_in_interleaving_chain): Likewise.
4566 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4568 (_loop_vec_info::~_loop_vec_info): Delete it.
4569 (cse_and_gimplify_to_preheader): New function.
4570 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4571 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4572 (vectorizable_load): Likewise. For grouped stores always base
4573 the IV on the first element.
4574 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4575 condition before gimplifying.
4577 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4579 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4580 *DIV_EXPR and *MOD_EXPR.
4582 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4585 * config/i386/i386.c (ix86_option_override_internal): Mask out
4586 the CF_SET bit when checking -fcf-protection.
4588 2018-02-07 Tom de Vries <tom@codesourcery.com>
4591 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4594 2018-02-07 Richard Biener <rguenther@suse.de>
4596 PR tree-optimization/84204
4597 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4600 PR tree-optimization/84205
4601 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4602 special-case isl_ast_op_zdiv_r.
4604 PR tree-optimization/84223
4605 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4606 Only add conditions from within the region.
4607 (gather_bbs::after_dom_children): Adjust.
4609 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4612 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4613 * config/avr/avr.md: Only post-reload split REG-REG moves if
4614 either register is GENERAL_REG_P.
4616 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4618 PR tree-optimization/84235
4619 * tree-ssa-scopedtables.c
4620 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4621 if the subtraction is performed in floating point type where NaNs are
4622 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4623 build 1. Formatting fix.
4625 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4628 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4629 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4630 and skip it regardless of bb boundaries. Use CALL_P macro,
4631 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4632 unnecessarily, formatting fix.
4634 2018-02-06 Michael Collison <michael.collison@arm.com>
4636 * config/arm/thumb2.md:
4637 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4638 (*thumb_mov_notscc): Ditto.
4640 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4643 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4644 unsigned_fix, not "s".
4646 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4648 * configure.ac (gcc_fn_eh_frame_ro): New function.
4649 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4650 correct .eh_frame permissions.
4651 * configure: Regenerate.
4653 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4655 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4658 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4660 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4661 Display warning message for -mno-speculate-indirect-jumps.
4663 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4665 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4667 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4669 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4671 PR tree-optimization/84225
4672 * tree-eh.c (find_trapping_overflow): Only call
4673 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4675 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4678 * config/i386/i386.c: Reimplement the check of possible options
4679 -mibt/-mshstk conbination. Change error messages.
4680 * doc/invoke.texi: Fix a typo: remove extra '='.
4682 2018-02-06 Marek Polacek <polacek@redhat.com>
4684 PR tree-optimization/84228
4685 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4687 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4690 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4691 emitted arch directives.
4692 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4693 __ARM_FEATURE_COPROC before changing architectures.
4695 2018-02-06 Richard Biener <rguenther@suse.de>
4697 * config/i386/i386.c (print_reg): Fix typo.
4698 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4700 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4702 * configure: Regenerate.
4704 2018-02-05 Martin Sebor <msebor@redhat.com>
4706 PR tree-optimization/83369
4707 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4710 2018-02-05 Martin Liska <mliska@suse.cz>
4712 * doc/invoke.texi: Cherry-pick upstream r323995.
4714 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4716 * ira.c (ira_init_register_move_cost): Adjust comment.
4718 2018-02-05 Martin Liska <mliska@suse.cz>
4720 PR gcov-profile/84137
4721 * doc/gcov.texi: Fix typo in documentation.
4723 2018-02-05 Martin Liska <mliska@suse.cz>
4725 PR gcov-profile/83879
4726 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4727 using dlopen functionality.
4729 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4731 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4732 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4733 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4734 _mm_maskz_range_round_ss): New intrinsics.
4735 (__builtin_ia32_rangesd128_round)
4736 (__builtin_ia32_rangess128_round): Remove.
4737 (__builtin_ia32_rangesd128_mask_round,
4738 __builtin_ia32_rangess128_mask_round): New builtins.
4739 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4740 __builtin_ia32_rangess128_round): Remove.
4741 (__builtin_ia32_rangesd128_mask_round,
4742 __builtin_ia32_rangess128_mask_round): New builtins.
4743 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4744 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4745 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4746 "<round_saeonly_constraint>")): Changed to ...
4747 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4748 "<round_saeonly_scalar_constraint>")): ... this.
4749 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4750 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4751 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4752 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4753 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4755 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4757 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4759 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4760 Remove all values except native, 8540 and 8548.
4762 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4764 * config/i386/i386.c (ix86_output_function_return): Pass
4765 INVALID_REGNUM, instead of -1, as invalid register number to
4766 indirect_thunk_name and output_indirect_thunk.
4768 2018-02-02 Julia Koval <julia.koval@intel.com>
4770 * config.gcc: Add -march=icelake.
4771 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4772 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4773 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4774 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4775 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4776 (processor_target_table): Add icelake.
4777 (ix86_option_override_internal): Handle new PTAs.
4778 (get_builtin_code_for_version): Handle icelake.
4779 (M_INTEL_COREI7_ICELAKE): New.
4780 (fold_builtin_cpu): Handle icelake.
4781 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4782 * doc/invoke.texi: Add -march=icelake.
4784 2018-02-02 Julia Koval <julia.koval@intel.com>
4786 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4787 to wide_int_bitmask.
4788 * wide-int-bitmask.h: New.
4790 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4793 * config/i386/i386.md: Replace Pmode with word_mode in
4794 builtin_setjmp_setup and builtin_longjmp to support x32.
4796 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4800 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4802 (rs6000_supported_cpu_names): New static variable.
4803 (linux_cpu_translation_table): Likewise.
4804 (elf_platform) <cpu>: Define new static variable and use it.
4805 Translate kernel AT_PLATFORM name to canonical name if needed.
4806 Error if platform name is unknown.
4808 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4811 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4813 2018-02-01 Jeff Law <law@redhat.com>
4816 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4817 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4818 the scratch if RELEASE_VIA_POP is false.
4819 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4820 If we have to save a temporary register, decrement SIZE appropriately.
4821 Pass new arguments to release_scratch_register_on_entry.
4822 (ix86_adjust_stack_and_probe): Likewise.
4823 (ix86_emit_probe_stack_range): Pass new arguments to
4824 release_scratch_register_on_entry.
4826 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4828 PR rtl-optimization/84157
4829 * combine.c (change_zero_ext): Use REG_P predicate in
4830 front of HARD_REGISTER_P predicate.
4832 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4834 * config/avr/avr.c (avr_option_override): Move disabling of
4835 -fdelete-null-pointer-checks to...
4836 * common/config/avr/avr-common.c (avr_option_optimization_table):
4839 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4841 PR tree-optimization/81635
4842 * tree-data-ref.c (split_constant_offset_1): For types that
4843 wrap on overflow, try to use range info to prove that wrapping
4846 2018-02-01 Renlin Li <renlin.li@arm.com>
4849 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4851 (aarch64_register_move_cost): Likewise.
4852 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4854 (REG_CLASS_NAMES): Likewise.
4855 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4856 TAILCALL_ADDR_REGS. Remove IP registers.
4857 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4859 2018-02-01 Richard Biener <rguenther@suse.de>
4861 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4862 for specifying RPO order and allow NULL for that.
4863 * domwalk.c (dom_walker::dom_walker): Likewise.
4864 (dom_walker::walk): Handle NULL RPO order.
4865 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4867 (rewrite_update_dom_walker): Likewise.
4868 (mark_def_dom_walker): Likewise.
4870 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4872 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4873 (aarch64_maybe_expand_sve_subreg_move): Declare.
4874 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4875 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4877 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4878 that are semantically a reverse operation.
4879 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4880 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4881 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4883 (aarch64_can_change_mode_class): For big-endian, forbid changes
4884 between two SVE modes if they have different element sizes.
4886 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4888 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4889 the TImode handling for big-endian targets.
4891 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4893 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4894 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4896 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4897 Remove BSWAP handing for big-endian targets and use the form of
4898 LD1RQ appropariate for the mode.
4900 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4902 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4903 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4906 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4909 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4910 check for operands that need to go through aarch64_sve_reload_be.
4912 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4914 PR tree-optimization/81661
4915 PR tree-optimization/84117
4916 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4917 * tree-eh.c: Include gimplify.h.
4918 (find_trapping_overflow, replace_trapping_overflow,
4919 rewrite_to_non_trapping_overflow): New functions.
4920 * tree-vect-loop.c: Include tree-eh.h.
4921 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4922 * tree-data-ref.c: Include tree-eh.h.
4923 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4925 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4927 PR rtl-optimization/84123
4928 * combine.c (change_zero_ext): Check if hard register satisfies
4929 can_change_dest_mode before calling gen_lowpart_SUBREG.
4931 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4934 * ira.c (ira_init_register_move_cost): Remove assert.
4936 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4938 PR rtl-optimization/84071
4939 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4940 * doc/tm.texi: Regenerate.
4942 2018-01-31 Richard Biener <rguenther@suse.de>
4944 PR tree-optimization/84132
4945 * tree-data-ref.c (analyze_miv_subscript): Properly
4946 check whether evolution_function_is_affine_multivariate_p
4947 before calling gcd_of_steps_may_divide_p.
4949 2018-01-31 Julia Koval <julia.koval@intel.com>
4952 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4953 * config/i386/i386.md (rdpid_rex64) New.
4954 (rdpid): Make 32bit only.
4956 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4959 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4960 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4962 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4965 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4967 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4969 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4971 PR rtl-optimization/84071
4972 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4973 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4975 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4977 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4978 (arc_attribute_table): Add 'aux' attribute.
4979 (arc_in_small_data_p): Consider aux like variables.
4980 (arc_is_aux_reg_p): New function.
4981 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4982 (arc_get_aux_arg): New function.
4983 (prepare_move_operands): Handle aux-register access.
4984 (arc_handle_aux_attribute): New function.
4985 * doc/extend.texi (ARC Variable attributes): Add subsection.
4987 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4989 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4990 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4991 (arc_attribute_table): Add 'uncached' attribute.
4992 (arc_print_operand): Print '.di' flag for uncached memory
4994 (arc_in_small_data_p): Do not consider for small data the uncached
4996 (arc_is_uncached_mem_p): New function.
4997 * config/arc/predicates.md (compact_store_memory_operand): Check
4998 for uncached memory accesses.
4999 (nonvol_nonimm_operand): Likewise.
5000 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
5002 2018-01-31 Jakub Jelinek <jakub@redhat.com>
5005 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
5006 falign-loops=): Add Optimization flag.
5008 2018-01-30 Jeff Law <law@redhat.com>
5011 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
5012 INT_REGISTERS_SAVED. Check it prior to calling
5013 get_scratch_register_on_entry.
5014 (ix86_adjust_stack_and_probe): Similarly.
5015 (ix86_emit_probe_stack_range): Similarly.
5016 (ix86_expand_prologue): Corresponding changes.
5018 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5021 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
5022 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
5024 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
5027 * lra-constraints.c (curr_insn_transform): Process AND in the
5030 2018-01-30 Jakub Jelinek <jakub@redhat.com>
5032 PR rtl-optimization/83986
5033 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
5034 dependence against last_pending_memory_flush in addition to
5037 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
5039 PR tree-optimization/81611
5040 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
5043 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
5046 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
5049 2018-01-30 Richard Biener <rguenther@suse.de>
5050 Jakub Jelinek <jakub@redhat.com>
5052 PR tree-optimization/84111
5053 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
5054 inner loops added during recursion, as they don't have up-to-date
5057 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5060 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
5061 (can_inline_edge_by_limits_p): ... here.
5062 (can_early_inline_edge_p, check_callers,
5063 update_caller_keys, update_callee_keys, recursive_inlining,
5064 add_new_edges_to_heap, speculation_useful_p,
5065 inline_small_functions,
5066 inline_small_functions, flatten_function,
5067 inline_to_all_callers_1): Update.
5069 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
5071 * profile-count.c (profile_count::combine_with_ipa_count): Handle
5074 2018-01-30 Richard Biener <rguenther@suse.de>
5076 PR tree-optimization/83008
5077 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
5078 invariant and constant vector uses in stmts when they need
5081 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5084 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
5085 * configure: Regenerate.
5087 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5089 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
5091 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
5092 Use gen_rtx_REG rather than gen_lowpart.
5094 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5096 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
5097 rather than 0 when creating partial subregs.
5099 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
5101 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
5104 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
5107 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
5108 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
5109 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
5110 flags. This restores the settings used before the 2017-07-24.
5111 Turning off pre increment/decrement/modify allows IVOPTS to
5112 optimize DF/SF loops where the index is an int.
5114 2018-01-29 Richard Biener <rguenther@suse.de>
5115 Kelvin Nilsen <kelvin@gcc.gnu.org>
5118 * tree-vect-stmts.c (vectorizable_call): Don't call
5119 targetm.vectorize_builtin_md_vectorized_function if callee is
5122 2018-01-22 Carl Love <cel@us.ibm.com>
5124 * doc/extend.tex: Fix typo in second arg in
5125 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5127 2018-01-29 Richard Biener <rguenther@suse.de>
5129 PR tree-optimization/84086
5130 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5131 (flush_ssaname_freelist): When SSA names were released reset
5132 the SCEV hash table.
5134 2018-01-29 Richard Biener <rguenther@suse.de>
5136 PR tree-optimization/84057
5137 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5138 removed paths when removing edges.
5140 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5142 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5143 -mfunction-return=@var{choice}.
5145 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5148 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5150 (layout::print_source_line): Likewise.
5151 (test_get_line_width_without_trailing_whitespace): Add test cases.
5153 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5156 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5159 2018-01-26 Jim Wilson <jimw@sifive.com>
5161 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5163 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5166 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5168 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5169 and CMP + SUB-immediate -> SUBS.
5171 2018-01-26 Martin Sebor <msebor@redhat.com>
5173 PR tree-optimization/83896
5174 * tree-ssa-strlen.c (get_string_len): Rename...
5175 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5176 Avoid assuming length is constant.
5177 (handle_char_store): Use HOST_WIDE_INT for string length.
5179 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5182 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5183 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5185 2018-01-26 Richard Biener <rguenther@suse.de>
5187 PR rtl-optimization/84003
5188 * dse.c (record_store): Only record redundant stores when
5189 the earlier store aliases at least all accesses the later one does.
5191 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5193 PR rtl-optimization/83985
5194 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5195 REG_CFA_RESTORE insns.
5196 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5197 REG_CFA_RESTORE insns here.
5200 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5201 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5203 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5205 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5206 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5207 (arc_init): Likewise.
5208 (arc_override_options): Likewise.
5209 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5211 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5213 * config/arc/arc.h (TARGET_DBNZ): Define.
5214 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5215 properly set the tune attribute.
5216 (dbnz): Use TARGET_DBNZ guard.
5217 * config/arc/arc.opt (mtune): Add core3 option.
5219 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5221 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5222 recognize new pic like addresses.
5223 (arc_delegitimize_address): Clean up.
5225 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5227 * config/arc/arc-arches.def: Option mrf16 valid for all
5229 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5230 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5231 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5232 * config/arc/arc-tables.opt: Regenerate.
5233 * config/arc/arc.c (arc_conditional_register_usage): Handle
5234 reduced register file case.
5235 (arc_file_start): Set must have build attributes.
5236 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5238 * config/arc/arc.opt (mrf16): Add new option.
5239 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5240 * config/arc/genmultilib.awk: Handle new mrf16 option.
5241 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5242 * config/arc/t-multilib: Regenerate.
5243 * doc/invoke.texi (ARC Options): Document mrf16 option.
5245 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5247 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5248 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5249 (arc_attribute_table): Add 'secure_call' attribute.
5250 (arc_print_operand): Print secure call operand.
5251 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5253 (arc_is_secure_call_p): New function. * config/arc/arc.md
5254 (call_i): Add support for sjli instruction.
5255 (call_value_i): Likewise.
5256 * config/arc/constraints.md (Csc): New constraint.
5258 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5259 John Eric Martin <John.Martin@emmicro-us.com>
5261 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5262 * config/arc/arc.c (_arc_jli_section): New struct.
5263 (arc_jli_section): New type.
5264 (rc_jli_sections): New static variable.
5265 (arc_handle_jli_attribute): New function.
5266 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5267 (arc_file_end): New function.
5268 (TARGET_ASM_FILE_END): Define.
5269 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5270 (arc_add_jli_section): New function.
5271 (jli_call_scan): Likewise.
5272 (arc_reorg): Call jli_call_scan.
5273 (arc_output_addsi): Remove 'S' from printing asm operand.
5274 (arc_is_jli_call_p): New function.
5275 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5277 (movhi_insn): Likewise.
5278 (movsi_insn): Likewise.
5279 (movsi_set_cc_insn): Likewise.
5280 (loadqi_update): Likewise.
5281 (load_zeroextendqisi_update): Likewise.
5282 (load_signextendqisi_update): Likewise.
5283 (loadhi_update): Likewise.
5284 (load_zeroextendhisi_update): Likewise.
5285 (load_signextendhisi_update): Likewise.
5286 (loadsi_update): Likewise.
5287 (loadsf_update): Likewise.
5288 (movsicc_insn): Likewise.
5289 (bset_insn): Likewise.
5290 (bxor_insn): Likewise.
5291 (bclr_insn): Likewise.
5292 (bmsk_insn): Likewise.
5293 (bicsi3_insn): Likewise.
5294 (cmpsi_cc_c_insn): Likewise.
5295 (movsi_ne): Likewise.
5296 (movsi_cond_exec): Likewise.
5297 (clrsbsi2): Likewise.
5308 (call_i): Remove 'S' asm letter, add jli instruction.
5309 (call_value_i): Likewise.
5310 * config/arc/arc.op (mjli-always): New option.
5311 * config/arc/constraints.md (Cji): New constraint.
5312 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5314 (subsf3_fpx): Likewise.
5315 (mulsf3_fpx): Likewise.
5316 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5318 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5320 * doc/invoke.texi (ARC): Document mjli-always option.
5322 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5324 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5325 avoid addition with 0 and use incw and decw where possible.
5327 2018-01-26 Richard Biener <rguenther@suse.de>
5329 PR tree-optimization/81082
5330 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5331 association if it requires casting to unsigned.
5332 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5333 from fold_plusminus_mult_expr to catch important cases late when
5334 range info is available.
5336 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5338 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5339 * configure.ac (hidden_linkonce): New test.
5340 * configure: Regenerate.
5341 * config.in: Regenerate.
5343 2018-01-26 Julia Koval <julia.koval@intel.com>
5345 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5346 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5347 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5348 _mm_mask_bitshuffle_epi64_mask): Fix type.
5349 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5350 USI_FTYPE_V4DI_V4DI_USI): Remove.
5351 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5352 __builtin_ia32_vpshufbitqmb256_mask,
5353 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5354 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5355 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5357 2018-01-26 Alan Modra <amodra@gmail.com>
5360 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5361 UNSPEC_VBPERMQ. Sort other unspecs.
5363 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5365 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5367 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5370 * predict.c (drop_profile): Do not push/pop cfun; update also
5372 (handle_missing_profiles): Fix logic looking for zero profiles.
5374 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5377 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5378 on functions with #pragma omp declare simd or functions with simd
5380 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5381 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5382 Remove trailing \n from warning_at calls.
5384 2018-01-25 Tom de Vries <tom@codesourcery.com>
5387 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5388 for neutered workers.
5390 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5393 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5394 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5396 2018-01-24 Jeff Law <law@redhat.com>
5399 * i386.c (get_probe_interval): Move to earlier point.
5400 (ix86_compute_frame_layout): If -fstack-clash-protection and
5401 the frame is larger than the probe interval, then use pushes
5402 to save registers rather than reg->mem moves.
5403 (ix86_expand_prologue): Remove conditional for int_registers_saved
5406 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5409 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5410 min/max for never referenced object.
5412 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5415 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5417 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5418 attributes from DECL_ATTRIBUTES (decl) without affecting
5419 DECL_ATTRIBUTES (current_function_decl).
5420 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5421 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5423 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5425 PR tree-optimization/83979
5426 * fold-const.c (fold_comparison): Use constant_boolean_node
5427 instead of boolean_{true,false}_node.
5429 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5431 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5434 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5436 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5437 Simplify the clause that sets the length attribute.
5438 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5439 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5440 clause that sets the length attribute.
5441 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5443 2018-01-24 Tom de Vries <tom@codesourcery.com>
5446 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5447 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5448 Add strict parameter.
5449 (prevent_branch_around_nothing): Insert dummy insn between branch to
5450 label and label with no ptx insn inbetween.
5451 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5453 2018-01-24 Tom de Vries <tom@codesourcery.com>
5456 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5457 for neutered threads in warp.
5458 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5460 2018-01-24 Richard Biener <rguenther@suse.de>
5462 PR tree-optimization/83176
5463 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5466 2018-01-24 Richard Biener <rguenther@suse.de>
5468 PR tree-optimization/82819
5469 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5470 code generating pluses that are no-ops in the target precision.
5472 2018-01-24 Richard Biener <rguenther@suse.de>
5475 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5477 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5479 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5480 to merge probabilities.
5481 * predict.c (probably_never_executed): Also mark as cold functions
5482 with global 0 profile and guessed local profile.
5483 * profile-count.c (profile_probability::combine_with_count): New
5485 * profile-count.h (profile_probability::operator*,
5486 profile_probability::operator*=, profile_probability::operator/,
5487 profile_probability::operator/=): Reduce precision to adjusted
5488 and set value to guessed on contradictory divisions.
5489 (profile_probability::combine_with_freq): Remove.
5490 (profile_probability::combine_wiht_count): Declare.
5491 (profile_count::force_nonzero):: Set to adjusted.
5492 (profile_count::probability_in):: Set quality to adjusted.
5493 * tree-ssa-tail-merge.c (replace_block_by): Use
5496 2018-01-23 Andrew Waterman <andrew@sifive.com>
5497 Jim Wilson <jimw@sifive.com>
5499 * config/riscv/riscv.c (riscv_stack_boundary): New.
5500 (riscv_option_override): Set riscv_stack_boundary. Handle
5501 riscv_preferred_stack_boundary_arg.
5502 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5503 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5504 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5505 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5506 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5507 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5509 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5512 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5513 of struct ix86_frame.
5514 (ix86_expand_epilogue): Likewise. Add a local variable for
5515 the reg_save_offset field in struct ix86_frame.
5517 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5519 PR tree-optimization/82604
5520 * tree-loop-distribution.c (enum partition_kind): New enum item
5521 PKIND_PARTIAL_MEMSET.
5522 (partition_builtin_p): Support above new enum item.
5523 (generate_code_for_partition): Ditto.
5524 (compute_access_range): Differentiate cases that equality can be
5525 proven at all loops, the innermost loops or no loops.
5526 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5527 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5528 (finalize_partitions, distribute_loop): Don't fuse partition of
5529 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5530 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5533 2018-01-23 Martin Liska <mliska@suse.cz>
5535 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5536 order to ignore the predictor.
5537 (PRED_POLYMORPHIC_CALL): Likewise.
5538 (PRED_RECURSIVE_CALL): Likewise.
5540 2018-01-23 Martin Liska <mliska@suse.cz>
5542 * tree-profile.c (tree_profiling): Print function header to
5543 aware reader which function we are working on.
5544 * value-prof.c (gimple_find_values_to_profile): Do not print
5545 not interesting value histograms.
5547 2018-01-23 Martin Liska <mliska@suse.cz>
5549 * profile-count.h (enum profile_quality): Add
5550 profile_uninitialized as the first value. Do not number values
5551 as they are zero based.
5552 (profile_count::verify): Update sanity check.
5553 (profile_probability::verify): Likewise.
5555 2018-01-23 Nathan Sidwell <nathan@acm.org>
5557 * doc/invoke.texi (ffor-scope): Deprecate.
5559 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5561 PR tree-optimization/83510
5562 * domwalk.c (set_all_edges_as_executable): New function.
5563 (dom_walker::dom_walker): Convert bool param
5564 "skip_unreachable_blocks" to enum reachability. Move setup of
5565 edge flags to set_all_edges_as_executable and only do it when
5566 reachability is REACHABLE_BLOCKS.
5567 * domwalk.h (enum dom_walker::reachability): New enum.
5568 (dom_walker::dom_walker): Convert bool param
5569 "skip_unreachable_blocks" to enum reachability.
5570 (set_all_edges_as_executable): New decl.
5571 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5572 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5574 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5575 but converting true to REACHABLE_BLOCKS.
5576 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5578 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5579 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5580 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5582 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5583 if check_all_array_refs will be called.
5585 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5587 * tree.c (selftest::test_location_wrappers): Add more test
5590 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5592 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5593 (selftest::test_bit_in_range): Likewise.
5595 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5598 * doc/sourcebuild.texi (vect_float): Say that the selector
5599 only describes the situation when -funsafe-math-optimizations is on.
5600 (vect_float_strict): Document.
5602 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5604 PR tree-optimization/83965
5605 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5606 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5607 instead of checking only for a reduction.
5608 (vect_recog_widen_sum_pattern): Likewise.
5610 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5612 * predict.c (probably_never_executed): Only use precise profile info.
5613 (compute_function_frequency): Skip after inlining hack since we now
5614 have quality checking.
5616 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5618 * profile-count.h (profile_probability::very_unlikely,
5619 profile_probability::unlikely, profile_probability::even): Set
5620 precision to guessed.
5622 2018-01-23 Richard Biener <rguenther@suse.de>
5624 PR tree-optimization/83963
5625 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5626 Properly terminate dominator walk when crossing the exit edge not
5627 when visiting its source block.
5629 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5632 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5633 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5635 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5637 PR tree-optimization/83957
5638 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5639 semicolon after for body surrounded by braces.
5641 PR tree-optimization/83081
5642 * profile-count.h (profile_probability::split): New method.
5643 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5644 Use profile_probability::split.
5645 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5646 when splitting a single conditional jump into 2.
5648 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5650 PR tree-optimization/69452
5651 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5654 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5656 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5657 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5658 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5660 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5662 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5663 * config/rl78/rl78.md (movdi): New define_expand.
5664 * config/rl78/rl78.c (rl78_split_movdi): New function.
5666 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5669 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5671 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5672 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5673 128-bit to produce an UNSPEC move to get the double word with the
5674 signbit and then a shift directly to do signbit.
5675 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5676 implementation with a new version that just does either a direct
5677 move or a regular move. Move memory interface to separate insns.
5678 Move insns so they are next to the expander.
5679 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5680 with signbit move. Split big and little endian case.
5681 (signbit<mode>2_dm_mem_le): Likewise.
5682 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5683 (signbit<mode>2_dm2): Likewise.
5685 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5687 * config/rl78/rl78.md (anddi3): New define_expand.
5689 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5691 * config/rl78/rl78.md (umindi3): New define_expand.
5693 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5695 * config/rl78/rl78.md (smindi3): New define_expand.
5697 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5699 * config/rl78/rl78.md (smaxdi3): New define_expand.
5701 2018-01-22 Carl Love <cel@us.ibm.com>
5703 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5704 LVX_V1TI): Add macro expansion.
5705 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5706 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5707 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5708 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5709 Change check to determine if the instruction is a byte reversing
5710 entry. Fix typo in comment.
5711 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5712 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5713 Add def_builtin calls for new builtins.
5714 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5715 Add define_insn expansion.
5717 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5719 * config/rl78/rl78.md (umaxdi3): New define_expand.
5721 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5723 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5724 for non-QImode registers.
5726 2018-01-22 Richard Biener <rguenther@suse.de>
5728 PR tree-optimization/83963
5729 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5730 including the loop exit block.
5731 (scop_detection::merge_sese): Likewise.
5732 (scop_detection::add_scop): Do it here instead.
5734 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5736 * doc/sourcebuild.texi (arm_softfloat): Document.
5738 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5741 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5742 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5743 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5745 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5746 David Edelsohn <dje.gcc@gmail.com>
5749 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5750 Change "crset eq" to "crset 2".
5751 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5752 (*call_indirect_aix<mode>_nospec): Likewise.
5753 (*call_value_indirect_aix<mode>_nospec): Likewise.
5754 (*call_indirect_elfv2<mode>_nospec): Likewise.
5755 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5756 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5757 change assembly output from . to $.
5758 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5759 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5760 (*tablejump<mode>_internal1_nospec): Likewise.
5762 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5765 * config/sh/sh_optimize_sett_clrt.cc:
5766 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5768 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5770 PR tree-optimization/83940
5771 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5772 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5773 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5774 use it to pass back the definition type.
5775 (vect_check_store_rhs): Likewise rhs_dt_out.
5776 (vect_build_gather_load_calls): Add a mask_dt argument and use
5777 it instead of a call to vect_is_simple_use.
5778 (vectorizable_store): Update calls to vect_check_load_store_mask
5779 and vect_check_store_rhs. Use the dt returned by the latter instead
5780 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5781 instead of calls to vect_is_simple_use. Pass the scalar rather
5782 than the vector operand to vect_is_simple_use when handling
5783 second and subsequent copies of an rhs value.
5784 (vectorizable_load): Update calls to vect_check_load_store_mask
5785 and vect_build_gather_load_calls. Use the cached mask_dt and
5786 gs_info.offset_dt instead of calls to vect_is_simple_use.
5788 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5791 * tree-emutls.c: Include gimplify.h.
5792 (lower_emutls_2): New function.
5793 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5794 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5795 it before further processing.
5798 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5799 UINTVAL (trueop1) instead of INTVAL (op1).
5801 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5805 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5806 INCOMING_FRAME_SP_OFFSET if not defined.
5807 (scan_trace): Add ENTRY argument. If true and
5808 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5809 emit a note to adjust the CFA offset.
5810 (create_cfi_notes): Adjust scan_trace callers.
5811 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5812 INCOMING_FRAME_SP_OFFSET in the CIE.
5813 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5814 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5816 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5817 * doc/tm.texi: Regenerated.
5819 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5821 PR rtl-optimization/83147
5822 * lra-constraints.c (remove_inheritance_pseudos): Use
5823 lra_substitute_pseudo_within_insn.
5825 2018-01-19 Tom de Vries <tom@codesourcery.com>
5826 Cesar Philippidis <cesar@codesourcery.com>
5829 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5831 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5834 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5835 spaces for function labels.
5837 2018-01-19 Martin Liska <mliska@suse.cz>
5839 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5840 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5841 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5842 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5843 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5844 (PRED_CONST_RETURN): Change from 69 to 65.
5845 (PRED_NULL_RETURN): Change from 91 to 71.
5846 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5847 (PRED_LOOP_GUARD): Change from 66 to 73.
5849 2018-01-19 Martin Liska <mliska@suse.cz>
5851 * predict.c (predict_insn_def): Add new assert.
5852 (struct branch_predictor): Change type to signed integer.
5853 (test_prediction_value_range): Amend test to cover
5855 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5856 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5857 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5858 (PRED_LOOP_IV_COMPARE): Likewise.
5859 * predict.h (PROB_UNINITIALIZED): Define new constant.
5861 2018-01-19 Martin Liska <mliska@suse.cz>
5863 * predict.c (dump_prediction): Add new format for
5864 analyze_brprob.py script which is enabled with -details
5866 * profile-count.h (precise_p): New function.
5868 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5870 PR tree-optimization/83922
5871 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5872 there are no statements that need masking.
5873 (vect_active_double_reduction_p): New function.
5874 (vect_analyze_loop_operations): Use it when handling phis that
5875 are not in the loop header.
5877 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5879 PR tree-optimization/83914
5880 * tree-vect-loop.c (vectorizable_induction): Don't convert
5881 init_expr or apply the peeling adjustment for inductions
5882 that are nested within the vectorized loop.
5884 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5886 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5889 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5893 * function.h (gimplify_parameters): Add gimple_seq * argument.
5894 * function.c: Include gimple.h and options.h.
5895 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5896 for the added local temporaries if needed.
5897 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5898 if there are any parameter cleanups, wrap whole body into a
5899 try/finally with the cleanups.
5901 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5904 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5905 Use GET_MODE_CLASS for scalar floating point.
5907 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5911 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5912 Fix call of call_cgraph_insertion_hooks.
5914 2018-01-18 Martin Sebor <msebor@redhat.com>
5916 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5918 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5921 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5924 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5927 * common.opt: (-ffile-prefix-map): New option.
5928 * opts.c (common_handle_option): Defer it.
5929 * opts-global.c (handle_common_deferred_options): Handle it.
5930 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5931 * file-prefix-map.h: New file.
5932 (remap_debug_filename, add_debug_prefix_map): ...here.
5933 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5934 * final.c (debug_prefix_map, add_debug_prefix_map
5935 remap_debug_filename): Move to...
5936 * file-prefix-map.c: New file.
5937 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5938 generalize, get rid of alloca(), use strrchr() instead of strchr().
5939 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5940 Implement in terms of add_prefix_map().
5941 (remap_macro_filename, remap_debug_filename): Implement in term of
5943 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5944 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5945 * dbxout.c: Include file-prefix-map.h.
5946 * varasm.c: Likewise.
5947 * vmsdbgout.c: Likewise.
5948 * xcoffout.c: Likewise.
5949 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5950 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5951 * doc/invoke.texi (-ffile-prefix-map): Document.
5952 (-fdebug-prefix-map): Update description.
5954 2018-01-18 Martin Liska <mliska@suse.cz>
5956 * config/i386/i386.c (indirect_thunk_name): Document that also
5958 (output_indirect_thunk): Document why both instructions
5959 (pause and lfence) are generated.
5961 2018-01-18 Richard Biener <rguenther@suse.de>
5963 PR tree-optimization/83887
5964 * graphite-scop-detection.c
5965 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5966 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5967 (scop_detection::merge_sese): Re-implement with a flood-fill
5968 algorithm that properly finds a SESE region if it exists.
5970 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5973 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5974 pointer_diff optimizations use view_convert instead of convert.
5976 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5978 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5979 Generate different code for -mno-speculate-indirect-jumps.
5980 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5981 (*call_indirect_aix<mode>): Disable for
5982 -mno-speculate-indirect-jumps.
5983 (*call_indirect_aix<mode>_nospec): New define_insn.
5984 (*call_value_indirect_aix<mode>): Disable for
5985 -mno-speculate-indirect-jumps.
5986 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5987 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5988 -mno-speculate-indirect-jumps.
5989 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5991 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5993 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5994 long double type, set the flags for noting the default long double
5995 type, even if we don't pass or return a long double type.
5997 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
6000 * ipa-inline.c (flatten_function): Do not overwrite final inlining
6003 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
6005 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
6006 support for merge[hl].
6007 (fold_mergehl_helper): New helper function.
6008 (tree-vector-builder.h): New #include for tree_vector_builder usage.
6009 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
6010 (altivec_vmrglw_direct): Add xxmrglw insn.
6012 2018-01-17 Andrew Waterman <andrew@sifive.com>
6014 * config/riscv/riscv.c (riscv_conditional_register_usage): If
6015 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
6017 2018-01-17 David Malcolm <dmalcolm@redhat.com>
6020 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
6021 call the lto_location_cache before reading the
6022 DECL_SOURCE_LOCATION of the types.
6024 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
6025 Richard Sandiford <richard.sandiford@linaro.org>
6027 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
6028 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
6029 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
6030 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
6031 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
6033 * config/aarch64/constraints.md (aarch64_movti_operand):
6035 * config/aarch64/predicates.md (Uti): Add new constraint.
6037 2018-01-17 Carl Love <cel@us.ibm.com>
6039 * config/rs6000/vsx.md (define_expand xl_len_r,
6040 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
6041 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
6043 (define_expand, define_insn): Move the shift left from the
6044 define_insn to the define_expand for lxvl and stxvl instructions.
6045 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
6046 and XL_LEN_R definitions to PURE.
6048 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
6050 * config/i386/i386.c (indirect_thunk_name): Declare regno
6051 as unsigned int. Compare regno with INVALID_REGNUM.
6052 (output_indirect_thunk): Ditto.
6053 (output_indirect_thunk_function): Ditto.
6054 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
6055 in the call to output_indirect_thunk_function.
6057 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
6060 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
6061 rather than the size of inner_type to determine the stack slot size
6062 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
6064 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
6067 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
6070 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
6072 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
6073 endian Linux systems to optionally enable multilibs for selecting
6074 the long double type if the user configured an explicit type.
6075 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
6076 have no long double multilibs if not defined.
6077 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
6078 warn if the user used -mabi={ieee,ibm}longdouble and we built
6079 multilibs for long double.
6080 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
6081 appropriate multilib option.
6082 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
6084 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
6085 for building long double multilibs.
6086 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
6088 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
6090 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
6093 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
6095 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
6098 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
6101 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
6104 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6106 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
6107 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
6109 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
6111 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
6112 different rtl trees depending on TARGET_64BIT.
6113 (rs6000_gen_lvx): Likewise.
6115 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
6117 * config/visium/visium.md (nop): Tweak comment.
6118 (hazard_nop): Likewise.
6120 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6122 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6123 -mspeculate-indirect-jumps.
6124 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6125 for -mno-speculate-indirect-jumps.
6126 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6127 (*call_value_indirect_elfv2<mode>): Disable for
6128 -mno-speculate-indirect-jumps.
6129 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6130 (indirect_jump): Emit different RTL for
6131 -mno-speculate-indirect-jumps.
6132 (*indirect_jump<mode>): Disable for
6133 -mno-speculate-indirect-jumps.
6134 (*indirect_jump<mode>_nospec): New define_insn.
6135 (tablejump): Emit different RTL for
6136 -mno-speculate-indirect-jumps.
6137 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6138 (tablejumpsi_nospec): New define_expand.
6139 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6140 (tablejumpdi_nospec): New define_expand.
6141 (*tablejump<mode>_internal1): Disable for
6142 -mno-speculate-indirect-jumps.
6143 (*tablejump<mode>_internal1_nospec): New define_insn.
6144 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6147 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6149 * caller-save.c (insert_save): Drop unnecessary parameter. All
6152 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6153 Richard Biener <rguenth@suse.de>
6156 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6157 return early, inline manually is_gimple_sizepos. Make sure if we
6158 call gimplify_expr we don't end up with a gimple constant.
6159 * tree.c (variably_modified_type_p): Don't return true for
6160 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6161 * gimplify.h (is_gimple_sizepos): Remove.
6163 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6165 PR tree-optimization/83857
6166 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6167 vectorizable_live_operation for pure SLP statements.
6168 (vectorizable_live_operation): Handle PHIs.
6170 2018-01-16 Richard Biener <rguenther@suse.de>
6172 PR tree-optimization/83867
6173 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6174 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6176 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6179 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6180 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6181 If off is not INTEGER_CST, issue a may not be aligned warning
6182 rather than isn't aligned. Use isn%'t rather than isn't.
6183 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6185 <case MULT_EXPR>: Improve the case when bottom and one of the
6186 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6187 operand, in that case check if the other operand is multiple of
6188 bottom divided by the INTEGER_CST operand.
6190 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6193 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6194 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6195 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6196 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6197 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6198 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6199 (pa_function_arg_size): New function.
6201 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6203 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6204 in a separate statement.
6206 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6208 PR tree-optimization/83847
6209 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6210 group gathers and scatters.
6212 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6214 PR rtl-optimization/86620
6215 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6217 PR rtl-optimization/83213
6218 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6219 to last if both are JUMP_INSNs.
6221 PR tree-optimization/83843
6222 * gimple-ssa-store-merging.c
6223 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6224 store_immediate_info for bswap/nop orig_stores.
6226 2018-01-15 Andrew Waterman <andrew@sifive.com>
6228 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6230 <UDIV>: Increase cost if !TARGET_DIV.
6232 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6234 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6235 (define_attr "cr_logical_3op"): New.
6236 (cceq_ior_compare): Adjust.
6237 (cceq_ior_compare_complement): Adjust.
6238 (*cceq_rev_compare): Adjust.
6239 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6240 (is_cracked_insn): Adjust.
6241 (insn_must_be_first_in_group): Adjust.
6242 * config/rs6000/40x.md: Adjust.
6243 * config/rs6000/440.md: Adjust.
6244 * config/rs6000/476.md: Adjust.
6245 * config/rs6000/601.md: Adjust.
6246 * config/rs6000/603.md: Adjust.
6247 * config/rs6000/6xx.md: Adjust.
6248 * config/rs6000/7450.md: Adjust.
6249 * config/rs6000/7xx.md: Adjust.
6250 * config/rs6000/8540.md: Adjust.
6251 * config/rs6000/cell.md: Adjust.
6252 * config/rs6000/e300c2c3.md: Adjust.
6253 * config/rs6000/e500mc.md: Adjust.
6254 * config/rs6000/e500mc64.md: Adjust.
6255 * config/rs6000/e5500.md: Adjust.
6256 * config/rs6000/e6500.md: Adjust.
6257 * config/rs6000/mpc.md: Adjust.
6258 * config/rs6000/power4.md: Adjust.
6259 * config/rs6000/power5.md: Adjust.
6260 * config/rs6000/power6.md: Adjust.
6261 * config/rs6000/power7.md: Adjust.
6262 * config/rs6000/power8.md: Adjust.
6263 * config/rs6000/power9.md: Adjust.
6264 * config/rs6000/rs64.md: Adjust.
6265 * config/rs6000/titan.md: Adjust.
6267 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6269 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6270 ix86_indirect_branch_register logic.
6272 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6274 * config/i386/constraints.md (Bs): Update
6275 ix86_indirect_branch_register check. Don't check
6276 ix86_indirect_branch_register with GOT_memory_operand.
6278 * config/i386/predicates.md (GOT_memory_operand): Don't check
6279 ix86_indirect_branch_register here.
6280 (GOT32_symbol_operand): Likewise.
6282 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6284 * config/i386/predicates.md (constant_call_address_operand):
6285 Rewrite ix86_indirect_branch_register logic.
6286 (sibcall_insn_operand): Likewise.
6288 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6290 * config/i386/constraints.md (Bs): Replace
6291 ix86_indirect_branch_thunk_register with
6292 ix86_indirect_branch_register.
6294 * config/i386/i386.md (indirect_jump): Likewise.
6295 (tablejump): Likewise.
6296 (*sibcall_memory): Likewise.
6297 (*sibcall_value_memory): Likewise.
6298 Peepholes of indirect call and jump via memory: Likewise.
6299 * config/i386/i386.opt: Likewise.
6300 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6301 (GOT_memory_operand): Likewise.
6302 (call_insn_operand): Likewise.
6303 (sibcall_insn_operand): Likewise.
6304 (GOT32_symbol_operand): Likewise.
6306 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6309 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6310 type rather than type addr's type points to.
6311 (expand_omp_atomic_mutex): Likewise.
6312 (expand_omp_atomic): Likewise.
6314 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6317 * config/i386/i386.c (output_indirect_thunk_function): Use
6318 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6319 for __x86_return_thunk.
6321 2018-01-15 Richard Biener <rguenther@suse.de>
6324 * expmed.c (extract_bit_field_1): Fix typo.
6326 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6329 * config/arm/iterators.md (VF): New mode iterator.
6330 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6331 Remove integer-related logic from pattern.
6332 (neon_vabd<mode>_3): Likewise.
6334 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6337 * common.opt (fstrict-overflow): No longer an alias.
6338 (fwrapv-pointer): New option.
6339 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6340 also for pointer types based on flag_wrapv_pointer.
6341 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6342 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6343 opts->x_flag_wrapv got set.
6344 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6345 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6346 POINTER_TYPE_OVERFLOW_UNDEFINED.
6347 * match.pd: Likewise in address comparison pattern.
6348 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6350 2018-01-15 Richard Biener <rguenther@suse.de>
6353 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6354 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6355 Reset type names to their identifier if their TYPE_DECL doesn't
6356 have linkage (and thus is used for ODR and devirt).
6357 (save_debug_info_for_decl): Remove.
6358 (save_debug_info_for_type): Likewise.
6359 (add_tree_to_fld_list): Adjust.
6360 * tree-pretty-print.c (dump_generic_node): Make dumping of
6361 type names more robust.
6363 2018-01-15 Richard Biener <rguenther@suse.de>
6365 * BASE-VER: Bump to 8.0.1.
6367 2018-01-14 Martin Sebor <msebor@redhat.com>
6370 * builtins.c (check_access): Avoid warning when the no-warning bit
6373 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6375 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6376 * ira-color (allocno_hard_regs_compare): Likewise.
6378 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6381 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6382 Use .pushsection/.popsection.
6384 2018-01-14 Martin Sebor <msebor@redhat.com>
6387 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6389 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6391 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6392 entry from extra_headers.
6393 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6394 extra_headers, make the list bitwise identical to the i?86-*-* one.
6396 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6398 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6399 -mcmodel=large with -mindirect-branch=thunk,
6400 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6401 -mfunction-return=thunk-extern.
6402 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6403 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6404 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6406 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6408 * config/i386/i386.c (print_reg): Print the name of the full
6409 integer register without '%'.
6410 (ix86_print_operand): Handle 'V'.
6411 * doc/extend.texi: Document 'V' modifier.
6413 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6415 * config/i386/constraints.md (Bs): Disallow memory operand for
6416 -mindirect-branch-register.
6418 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6419 (GOT_memory_operand): Likewise.
6420 (call_insn_operand): Likewise.
6421 (sibcall_insn_operand): Likewise.
6422 (GOT32_symbol_operand): Likewise.
6423 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6424 for -mindirect-branch-register.
6425 (tablejump): Likewise.
6426 (*sibcall_memory): Likewise.
6427 (*sibcall_value_memory): Likewise.
6428 Disallow peepholes of indirect call and jump via memory for
6429 -mindirect-branch-register.
6430 (*call_pop): Replace m with Bw.
6431 (*call_value_pop): Likewise.
6432 (*sibcall_pop_memory): Replace m with Bs.
6433 * config/i386/i386.opt (mindirect-branch-register): New option.
6434 * doc/invoke.texi: Document -mindirect-branch-register option.
6436 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6438 * config/i386/i386-protos.h (ix86_output_function_return): New.
6439 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6440 set function_return_type.
6441 (indirect_thunk_name): Add ret_p to indicate thunk for function
6443 (output_indirect_thunk_function): Pass false to
6444 indirect_thunk_name.
6445 (ix86_output_indirect_branch_via_reg): Likewise.
6446 (ix86_output_indirect_branch_via_push): Likewise.
6447 (output_indirect_thunk_function): Create alias for function
6448 return thunk if regno < 0.
6449 (ix86_output_function_return): New function.
6450 (ix86_handle_fndecl_attribute): Handle function_return.
6451 (ix86_attribute_table): Add function_return.
6452 * config/i386/i386.h (machine_function): Add
6453 function_return_type.
6454 * config/i386/i386.md (simple_return_internal): Use
6455 ix86_output_function_return.
6456 (simple_return_internal_long): Likewise.
6457 * config/i386/i386.opt (mfunction-return=): New option.
6458 (indirect_branch): Mention -mfunction-return=.
6459 * doc/extend.texi: Document function_return function attribute.
6460 * doc/invoke.texi: Document -mfunction-return= option.
6462 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6464 * config/i386/i386-opts.h (indirect_branch): New.
6465 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6466 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6467 with local indirect jump when converting indirect call and jump.
6468 (ix86_set_indirect_branch_type): New.
6469 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6470 (indirectlabelno): New.
6471 (indirect_thunk_needed): Likewise.
6472 (indirect_thunk_bnd_needed): Likewise.
6473 (indirect_thunks_used): Likewise.
6474 (indirect_thunks_bnd_used): Likewise.
6475 (INDIRECT_LABEL): Likewise.
6476 (indirect_thunk_name): Likewise.
6477 (output_indirect_thunk): Likewise.
6478 (output_indirect_thunk_function): Likewise.
6479 (ix86_output_indirect_branch_via_reg): Likewise.
6480 (ix86_output_indirect_branch_via_push): Likewise.
6481 (ix86_output_indirect_branch): Likewise.
6482 (ix86_output_indirect_jmp): Likewise.
6483 (ix86_code_end): Call output_indirect_thunk_function if needed.
6484 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6486 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6487 (ix86_attribute_table): Add indirect_branch.
6488 * config/i386/i386.h (machine_function): Add indirect_branch_type
6489 and has_local_indirect_jump.
6490 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6492 (tablejump): Likewise.
6493 (*indirect_jump): Use ix86_output_indirect_jmp.
6494 (*tablejump_1): Likewise.
6495 (simple_return_indirect_internal): Likewise.
6496 * config/i386/i386.opt (mindirect-branch=): New option.
6497 (indirect_branch): New.
6500 (thunk-inline): Likewise.
6501 (thunk-extern): Likewise.
6502 * doc/extend.texi: Document indirect_branch function attribute.
6503 * doc/invoke.texi: Document -mindirect-branch= option.
6505 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6508 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6510 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6512 * ipa-inline.c (want_inline_small_function_p): Return false if
6513 inlining has already failed with CIF_FINAL_ERROR.
6514 (update_caller_keys): Call want_inline_small_function_p before
6516 (update_callee_keys): Likewise.
6518 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6520 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6522 (rs6000_quadword_masked_address_p): Likewise.
6523 (quad_aligned_load_p): Likewise.
6524 (quad_aligned_store_p): Likewise.
6525 (const_load_sequence_p): Add comment to describe the outer-most loop.
6526 (mimic_memory_attributes_and_flags): New function.
6527 (rs6000_gen_stvx): Likewise.
6528 (replace_swapped_aligned_store): Likewise.
6529 (rs6000_gen_lvx): Likewise.
6530 (replace_swapped_aligned_load): Likewise.
6531 (replace_swapped_load_constant): Capitalize argument name in
6532 comment describing this function.
6533 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6534 and stores that access quad-word aligned addresses and replace
6535 with stvx or lvx instructions when appropriate.
6536 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6537 New function prototype.
6538 (rs6000_quadword_masked_address_p): Likewise.
6539 (rs6000_gen_lvx): Likewise.
6540 (rs6000_gen_stvx): Likewise.
6541 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6542 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6543 when memory address is aligned.
6544 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6545 this split to select lvx instruction when memory address is aligned.
6546 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6547 instruction when memory address is aligned.
6548 (*vsx_le_perm_load_v16qi): Likewise.
6549 (four unnamed splitters): Modify to select the stvx instruction
6550 when memory is aligned.
6552 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6554 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6555 which appears in the queue multiple times.
6557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6558 Alan Hayward <alan.hayward@arm.com>
6559 David Sherwood <david.sherwood@arm.com>
6561 * tree-vectorizer.h (vec_lower_bound): New structure.
6562 (_loop_vec_info): Add check_nonzero and lower_bounds.
6563 (LOOP_VINFO_CHECK_NONZERO): New macro.
6564 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6565 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6566 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6567 fields. Make seg_len the distance travelled, not including the
6569 (dr_direction_indicator): Declare.
6570 (dr_zero_step_indicator): Likewise.
6571 (dr_known_forward_stride_p): Likewise.
6572 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6574 (runtime_alias_check_p): Allow runtime alias checks with
6576 (operator ==): Compare access_size and align.
6577 (prune_runtime_alias_test_list): Rework for new distinction between
6578 the access_size and seg_len.
6579 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6581 (get_segment_min_max): New function.
6582 (create_intersect_range_checks): Use it.
6583 (dr_step_indicator): New function.
6584 (dr_direction_indicator): Likewise.
6585 (dr_zero_step_indicator): Likewise.
6586 (dr_known_forward_stride_p): Likewise.
6587 * tree-loop-distribution.c (data_ref_segment_size): Return
6588 DR_STEP * (niters - 1).
6589 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6591 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6592 (vect_preserves_scalar_order_p): New function, split out from...
6593 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6594 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6595 (vect_vfa_access_size): New function.
6596 (vect_vfa_align): Likewise.
6597 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6598 (dump_lower_bound): New function.
6599 (vect_check_lower_bound): Likewise.
6600 (vect_small_gap_p): Likewise.
6601 (vectorizable_with_step_bound_p): Likewise.
6602 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6603 depencies if the vectorization factor is 1. Convert the checks
6604 for nonzero steps into checks on the bounds of DR_STEP. Try using
6605 a bunds check for variable steps if the minimum required step is
6606 relatively small. Update calls to the dr_with_seg_len
6607 constructor and to vect_compile_time_alias.
6608 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6610 (vect_loop_versioning): Call it.
6611 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6613 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6615 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6616 Alan Hayward <alan.hayward@arm.com>
6617 David Sherwood <david.sherwood@arm.com>
6619 * doc/sourcebuild.texi (vect_scatter_store): Document.
6620 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6622 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6624 * genopinit.c (main): Add supports_vec_scatter_store and
6625 supports_vec_scatter_store_cached to target_optabs.
6626 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6627 IFN_MASK_SCATTER_STORE.
6628 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6630 * internal-fn.h (internal_store_fn_p): Declare.
6631 (internal_fn_stored_value_index): Likewise.
6632 * internal-fn.c (scatter_store_direct): New macro.
6633 (expand_scatter_store_optab_fn): New function.
6634 (direct_scatter_store_optab_supported_p): New macro.
6635 (internal_store_fn_p): New function.
6636 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6637 IFN_MASK_SCATTER_STORE.
6638 (internal_fn_mask_index): Likewise.
6639 (internal_fn_stored_value_index): New function.
6640 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6642 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6643 * optabs-query.c (supports_vec_scatter_store_p): New function.
6644 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6645 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6646 true for scatter stores.
6647 (vect_gather_scatter_fn_p): Handle scatter stores too.
6648 (vect_check_gather_scatter): Consider using scatter stores if
6649 supports_vec_scatter_store_p.
6650 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6652 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6653 internal_fn_stored_value_index.
6654 (check_load_store_masking): Handle scatter stores too.
6655 (vect_get_store_rhs): Make public.
6656 (vectorizable_call): Use internal_store_fn_p.
6657 (vectorizable_store): Handle scatter store internal functions.
6658 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6659 when deciding whether the end of the group has been reached.
6660 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6661 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6662 (mask_scatter_store<mode>): New insns.
6664 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6665 Alan Hayward <alan.hayward@arm.com>
6666 David Sherwood <david.sherwood@arm.com>
6668 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6669 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6670 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6672 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6673 Use vect_truncate_gather_scatter_offset if we can't treat the
6674 operation as a normal gather load or scatter store.
6675 (get_group_load_store_type): Take the gather_scatter_info
6676 as argument. Try using a gather load or scatter store for
6677 single-element groups.
6678 (get_load_store_type): Update calls to get_group_load_store_type
6679 and vect_use_strided_gather_scatters_p.
6681 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6682 Alan Hayward <alan.hayward@arm.com>
6683 David Sherwood <david.sherwood@arm.com>
6685 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6686 optional tree argument.
6687 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6689 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6690 but continue to use the current value as a fallback.
6691 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6692 to compare the updates.
6693 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6694 (get_load_store_type): Use it when handling a strided access.
6695 (vect_get_strided_load_store_ops): New function.
6696 (vect_get_data_ptr_increment): Likewise.
6697 (vectorizable_load): Handle strided gather loads. Always pass
6698 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6700 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6701 Alan Hayward <alan.hayward@arm.com>
6702 David Sherwood <david.sherwood@arm.com>
6704 * doc/md.texi (gather_load@var{m}): Document.
6705 (mask_gather_load@var{m}): Likewise.
6706 * genopinit.c (main): Add supports_vec_gather_load and
6707 supports_vec_gather_load_cached to target_optabs.
6708 * optabs-tree.c (init_tree_optimization_optabs): Use
6709 ggc_cleared_alloc to allocate target_optabs.
6710 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6711 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6713 * internal-fn.h (internal_load_fn_p): Declare.
6714 (internal_gather_scatter_fn_p): Likewise.
6715 (internal_fn_mask_index): Likewise.
6716 (internal_gather_scatter_fn_supported_p): Likewise.
6717 * internal-fn.c (gather_load_direct): New macro.
6718 (expand_gather_load_optab_fn): New function.
6719 (direct_gather_load_optab_supported_p): New macro.
6720 (direct_internal_fn_optab): New function.
6721 (internal_load_fn_p): Likewise.
6722 (internal_gather_scatter_fn_p): Likewise.
6723 (internal_fn_mask_index): Likewise.
6724 (internal_gather_scatter_fn_supported_p): Likewise.
6725 * optabs-query.c (supports_at_least_one_mode_p): New function.
6726 (supports_vec_gather_load_p): Likewise.
6727 * optabs-query.h (supports_vec_gather_load_p): Declare.
6728 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6729 and memory_type field.
6730 (NUM_PATTERNS): Bump to 15.
6731 * tree-vect-data-refs.c: Include internal-fn.h.
6732 (vect_gather_scatter_fn_p): New function.
6733 (vect_describe_gather_scatter_call): Likewise.
6734 (vect_check_gather_scatter): Try using internal functions for
6735 gather loads. Recognize existing calls to a gather load function.
6736 (vect_analyze_data_refs): Consider using gather loads if
6737 supports_vec_gather_load_p.
6738 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6739 (vect_get_gather_scatter_offset_type): Likewise.
6740 (vect_convert_mask_for_vectype): Likewise.
6741 (vect_add_conversion_to_patterm): Likewise.
6742 (vect_try_gather_scatter_pattern): Likewise.
6743 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6744 (vect_vect_recog_func_ptrs): Add it.
6745 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6746 internal_fn_mask_index and internal_gather_scatter_fn_p.
6747 (check_load_store_masking): Take the gather_scatter_info as an
6748 argument and handle gather loads.
6749 (vect_get_gather_scatter_ops): New function.
6750 (vectorizable_call): Check internal_load_fn_p.
6751 (vectorizable_load): Likewise. Handle gather load internal
6753 (vectorizable_store): Update call to check_load_store_masking.
6754 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6755 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6756 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6757 (aarch64_gather_scale_operand_d): New predicates.
6758 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6759 (mask_gather_load<mode>): New insns.
6761 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6762 Alan Hayward <alan.hayward@arm.com>
6763 David Sherwood <david.sherwood@arm.com>
6765 * optabs.def (fold_left_plus_optab): New optab.
6766 * doc/md.texi (fold_left_plus_@var{m}): Document.
6767 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6768 * internal-fn.c (fold_left_direct): Define.
6769 (expand_fold_left_optab_fn): Likewise.
6770 (direct_fold_left_optab_supported_p): Likewise.
6771 * fold-const-call.c (fold_const_fold_left): New function.
6772 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6773 * tree-parloops.c (valid_reduction_p): New function.
6774 (gather_scalar_reductions): Use it.
6775 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6776 (vect_finish_replace_stmt): Declare.
6777 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6778 (needs_fold_left_reduction_p): New function, split out from...
6779 (vect_is_simple_reduction): ...here. Accept reductions that
6780 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6781 (vect_force_simple_reduction): Also store the reduction type in
6782 the assignment's STMT_VINFO_REDUC_TYPE.
6783 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6784 (merge_with_identity): New function.
6785 (vect_expand_fold_left): Likewise.
6786 (vectorize_fold_left_reduction): Likewise.
6787 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6788 scalar phi in place for it. Check for target support and reject
6789 cases that would reassociate the operation. Defer the transform
6790 phase to vectorize_fold_left_reduction.
6791 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6792 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6793 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6795 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6797 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6798 call to ifc_temp_var.
6800 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6801 Alan Hayward <alan.hayward@arm.com>
6802 David Sherwood <david.sherwood@arm.com>
6804 * target.def (legitimize_address_displacement): Take the original
6805 offset as a poly_int.
6806 * targhooks.h (default_legitimize_address_displacement): Update
6808 * targhooks.c (default_legitimize_address_displacement): Likewise.
6809 * doc/tm.texi: Regenerate.
6810 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6811 as an argument, moving assert of ad->disp == ad->disp_term to...
6812 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6813 Try calling targetm.legitimize_address_displacement before expanding
6814 the address rather than afterwards, and adjust for the new interface.
6815 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6816 Match the new hook interface. Handle SVE addresses.
6817 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6820 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6822 * Makefile.in (OBJS): Add early-remat.o.
6823 * target.def (select_early_remat_modes): New hook.
6824 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6825 * doc/tm.texi: Regenerate.
6826 * targhooks.h (default_select_early_remat_modes): Declare.
6827 * targhooks.c (default_select_early_remat_modes): New function.
6828 * timevar.def (TV_EARLY_REMAT): New timevar.
6829 * passes.def (pass_early_remat): New pass.
6830 * tree-pass.h (make_pass_early_remat): Declare.
6831 * early-remat.c: New file.
6832 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6834 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6836 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6837 Alan Hayward <alan.hayward@arm.com>
6838 David Sherwood <david.sherwood@arm.com>
6840 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6841 vfm1 with a bound_epilog parameter.
6842 (vect_do_peeling): Update calls accordingly, and move the prologue
6843 call earlier in the function. Treat the base bound_epilog as 0 for
6844 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6845 this base when peeling for gaps.
6846 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6847 with fully-masked loops.
6848 (vect_estimate_min_profitable_iters): Handle the single peeled
6849 iteration in that case.
6851 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6852 Alan Hayward <alan.hayward@arm.com>
6853 David Sherwood <david.sherwood@arm.com>
6855 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6856 single-element interleaving even if the size is not a power of 2.
6857 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6858 accesses for single-element interleaving if the group size is
6861 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6862 Alan Hayward <alan.hayward@arm.com>
6863 David Sherwood <david.sherwood@arm.com>
6865 * doc/md.texi (fold_extract_last_@var{m}): Document.
6866 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6867 * optabs.def (fold_extract_last_optab): New optab.
6868 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6869 * internal-fn.c (fold_extract_direct): New macro.
6870 (expand_fold_extract_optab_fn): Likewise.
6871 (direct_fold_extract_optab_supported_p): Likewise.
6872 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6873 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6874 EXTRACT_LAST_REDUCTION.
6875 (get_initial_def_for_reduction): Do not create an initial vector
6876 for EXTRACT_LAST_REDUCTION reductions.
6877 (vectorizable_reduction): Leave the scalar phi in place for
6878 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6879 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6880 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6881 transform phase to vectorizable_condition.
6882 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6884 (vect_finish_stmt_generation): ...here.
6885 (vect_finish_replace_stmt): New function.
6886 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6887 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6889 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6891 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6892 Alan Hayward <alan.hayward@arm.com>
6893 David Sherwood <david.sherwood@arm.com>
6895 * doc/md.texi (extract_last_@var{m}): Document.
6896 * optabs.def (extract_last_optab): New optab.
6897 * internal-fn.def (EXTRACT_LAST): New internal function.
6898 * internal-fn.c (cond_unary_direct): New macro.
6899 (expand_cond_unary_optab_fn): Likewise.
6900 (direct_cond_unary_optab_supported_p): Likewise.
6901 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6902 loops using EXTRACT_LAST.
6903 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6904 (extract_last_<mode>): ...this optab.
6905 (vec_extract<mode><Vel>): Update accordingly.
6907 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6908 Alan Hayward <alan.hayward@arm.com>
6909 David Sherwood <david.sherwood@arm.com>
6911 * target.def (empty_mask_is_expensive): New hook.
6912 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6913 * doc/tm.texi: Regenerate.
6914 * targhooks.h (default_empty_mask_is_expensive): Declare.
6915 * targhooks.c (default_empty_mask_is_expensive): New function.
6916 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6917 if the target says that empty masks are expensive.
6918 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6920 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
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 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6927 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6928 (vect_use_loop_mask_for_alignment_p): New function.
6929 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6930 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6931 niters_skip argument. Make sure that the first niters_skip elements
6932 of the first iteration are inactive.
6933 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6934 Update call to vect_set_loop_masks_directly.
6935 (get_misalign_in_elems): New function, split out from...
6936 (vect_gen_prolog_loop_niters): ...here.
6937 (vect_update_init_of_dr): Take a code argument that specifies whether
6938 the adjustment should be added or subtracted.
6939 (vect_update_init_of_drs): Likewise.
6940 (vect_prepare_for_masked_peels): New function.
6941 (vect_do_peeling): Skip prologue peeling if we're using a mask
6942 instead. Update call to vect_update_inits_of_drs.
6943 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6945 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6946 alignment. Do not include the number of peeled iterations in
6947 the minimum threshold in that case.
6948 (vectorizable_induction): Adjust the start value down by
6949 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6950 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6951 Take the number of skipped iterations into account when calculating
6953 * tree-vect-stmts.c (vect_gen_while_not): New function.
6955 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6956 Alan Hayward <alan.hayward@arm.com>
6957 David Sherwood <david.sherwood@arm.com>
6959 * doc/sourcebuild.texi (vect_fully_masked): Document.
6960 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6962 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6964 (vect_analyze_loop_2): ...here. Don't check the vectorization
6965 factor against the number of loop iterations if the loop is
6968 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6969 Alan Hayward <alan.hayward@arm.com>
6970 David Sherwood <david.sherwood@arm.com>
6972 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6973 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6974 (dump_groups): Update accordingly.
6975 (iv_use::mem_type): New member variable.
6976 (address_p): New function.
6977 (record_use): Add a mem_type argument and initialize the new
6979 (record_group_use): Add a mem_type argument. Use address_p.
6980 Remove obsolete null checks of base_object. Update call to record_use.
6981 (find_interesting_uses_op): Update call to record_group_use.
6982 (find_interesting_uses_cond): Likewise.
6983 (find_interesting_uses_address): Likewise.
6984 (get_mem_type_for_internal_fn): New function.
6985 (find_address_like_use): Likewise.
6986 (find_interesting_uses_stmt): Try find_address_like_use before
6987 calling find_interesting_uses_op.
6988 (addr_offset_valid_p): Use the iv mem_type field as the type
6989 of the addressed memory.
6990 (add_autoinc_candidates): Likewise.
6991 (get_address_cost): Likewise.
6992 (split_small_address_groups_p): Use address_p.
6993 (split_address_groups): Likewise.
6994 (add_iv_candidate_for_use): Likewise.
6995 (autoinc_possible_for_pair): Likewise.
6996 (rewrite_groups): Likewise.
6997 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6998 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6999 (get_alias_ptr_type_for_ptr_address): New function.
7000 (rewrite_use_address): Rewrite address uses in calls that were
7001 identified by find_address_like_use.
7003 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7004 Alan Hayward <alan.hayward@arm.com>
7005 David Sherwood <david.sherwood@arm.com>
7007 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
7009 * gimple-expr.h (is_gimple_addressable: Likewise.
7010 * gimple-expr.c (is_gimple_address): Likewise.
7011 * internal-fn.c (expand_call_mem_ref): New function.
7012 (expand_mask_load_optab_fn): Use it.
7013 (expand_mask_store_optab_fn): Likewise.
7015 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7016 Alan Hayward <alan.hayward@arm.com>
7017 David Sherwood <david.sherwood@arm.com>
7019 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
7020 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
7021 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
7022 (cond_umax@var{mode}): Document.
7023 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
7024 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
7025 (cond_umin_optab, cond_umax_optab): New optabs.
7026 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
7027 (COND_IOR, COND_XOR): New internal functions.
7028 * internal-fn.h (get_conditional_internal_fn): Declare.
7029 * internal-fn.c (cond_binary_direct): New macro.
7030 (expand_cond_binary_optab_fn): Likewise.
7031 (direct_cond_binary_optab_supported_p): Likewise.
7032 (get_conditional_internal_fn): New function.
7033 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
7034 Cope with reduction statements that are vectorized as calls rather
7036 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
7037 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
7038 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
7039 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
7040 (UNSPEC_COND_EOR): New unspecs.
7041 (optab): Add mappings for them.
7042 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
7043 (sve_int_op, sve_fp_op): New int attributes.
7045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7046 Alan Hayward <alan.hayward@arm.com>
7047 David Sherwood <david.sherwood@arm.com>
7049 * optabs.def (while_ult_optab): New optab.
7050 * doc/md.texi (while_ult@var{m}@var{n}): Document.
7051 * internal-fn.def (WHILE_ULT): New internal function.
7052 * internal-fn.h (direct_internal_fn_supported_p): New override
7053 that takes two types as argument.
7054 * internal-fn.c (while_direct): New macro.
7055 (expand_while_optab_fn): New function.
7056 (convert_optab_supported_p): Likewise.
7057 (direct_while_optab_supported_p): New macro.
7058 * wide-int.h (wi::udiv_ceil): New function.
7059 * tree-vectorizer.h (rgroup_masks): New structure.
7060 (vec_loop_masks): New typedef.
7061 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
7063 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
7064 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
7065 (vect_max_vf): New function.
7066 (slpeel_make_loop_iterate_ntimes): Delete.
7067 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
7068 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
7069 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
7070 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
7071 internal-fn.h, stor-layout.h and optabs-query.h.
7072 (vect_set_loop_mask): New function.
7073 (add_preheader_seq): Likewise.
7074 (add_header_seq): Likewise.
7075 (interleave_supported_p): Likewise.
7076 (vect_maybe_permute_loop_masks): Likewise.
7077 (vect_set_loop_masks_directly): Likewise.
7078 (vect_set_loop_condition_masked): Likewise.
7079 (vect_set_loop_condition_unmasked): New function, split out from
7080 slpeel_make_loop_iterate_ntimes.
7081 (slpeel_make_loop_iterate_ntimes): Rename to..
7082 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
7083 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
7084 (vect_do_peeling): Update call accordingly.
7085 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
7087 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
7088 mask_compare_type, can_fully_mask_p and fully_masked_p.
7089 (release_vec_loop_masks): New function.
7090 (_loop_vec_info): Use it to free the loop masks.
7091 (can_produce_all_loop_masks_p): New function.
7092 (vect_get_max_nscalars_per_iter): Likewise.
7093 (vect_verify_full_masking): Likewise.
7094 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
7095 retries, and free the mask rgroups before retrying. Check loop-wide
7096 reasons for disallowing fully-masked loops. Make the final decision
7097 about whether use a fully-masked loop or not.
7098 (vect_estimate_min_profitable_iters): Do not assume that peeling
7099 for the number of iterations will be needed for fully-masked loops.
7100 (vectorizable_reduction): Disable fully-masked loops.
7101 (vectorizable_live_operation): Likewise.
7102 (vect_halve_mask_nunits): New function.
7103 (vect_double_mask_nunits): Likewise.
7104 (vect_record_loop_mask): Likewise.
7105 (vect_get_loop_mask): Likewise.
7106 (vect_transform_loop): Handle the case in which the final loop
7107 iteration might handle a partial vector. Call vect_set_loop_condition
7108 instead of slpeel_make_loop_iterate_ntimes.
7109 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
7110 (check_load_store_masking): New function.
7111 (prepare_load_store_mask): Likewise.
7112 (vectorizable_store): Handle fully-masked loops.
7113 (vectorizable_load): Likewise.
7114 (supportable_widening_operation): Use vect_halve_mask_nunits for
7116 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
7117 (vect_gen_while): New function.
7118 * config/aarch64/aarch64.md (umax<mode>3): New expander.
7119 (aarch64_uqdec<mode>): New insn.
7121 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7122 Alan Hayward <alan.hayward@arm.com>
7123 David Sherwood <david.sherwood@arm.com>
7125 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7126 (reduc_xor_scal_optab): New optabs.
7127 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7128 (reduc_xor_scal_@var{m}): Document.
7129 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7130 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7132 * fold-const-call.c (fold_const_call): Handle them.
7133 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7134 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7135 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7136 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7137 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7138 (UNSPEC_XORV): New unspecs.
7139 (optab): Add entries for them.
7140 (BITWISEV): New int iterator.
7141 (bit_reduc_op): New int attributes.
7143 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7144 Alan Hayward <alan.hayward@arm.com>
7145 David Sherwood <david.sherwood@arm.com>
7147 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7148 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7149 * optabs.def (vec_shl_insert_optab): New optab.
7150 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7151 (duplicate_and_interleave): Likewise.
7152 * tree-vect-loop.c: Include internal-fn.h.
7153 (neutral_op_for_slp_reduction): New function, split out from
7154 get_initial_defs_for_reduction.
7155 (get_initial_def_for_reduction): Handle option 2 for variable-length
7156 vectors by loading the neutral value into a vector and then shifting
7157 the initial value into element 0.
7158 (get_initial_defs_for_reduction): Replace the code argument with
7159 the neutral value calculated by neutral_op_for_slp_reduction.
7160 Use gimple_build_vector for constant-length vectors.
7161 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7162 but the first group_size elements have a neutral value.
7163 Use duplicate_and_interleave otherwise.
7164 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7165 Update call to get_initial_defs_for_reduction. Handle SLP
7166 reductions for variable-length vectors by creating one vector
7167 result for each scalar result, with the elements associated
7168 with other scalar results stubbed out with the neutral value.
7169 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7170 Require IFN_VEC_SHL_INSERT for double reductions on
7171 variable-length vectors, or SLP reductions that have
7172 a neutral value. Require can_duplicate_and_interleave_p
7173 support for variable-length unchained SLP reductions if there
7174 is no neutral value, such as for MIN/MAX reductions. Also require
7175 the number of vector elements to be a multiple of the number of
7176 SLP statements when doing variable-length unchained SLP reductions.
7177 Update call to vect_create_epilog_for_reduction.
7178 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7179 and remove initial values.
7180 (duplicate_and_interleave): Make public.
7181 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7182 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7184 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7185 Alan Hayward <alan.hayward@arm.com>
7186 David Sherwood <david.sherwood@arm.com>
7188 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7189 (can_duplicate_and_interleave_p): New function.
7190 (vect_get_and_check_slp_defs): Take the vector of statements
7191 rather than just the current one. Remove excess parentheses.
7192 Restriction rejectinon of vect_constant_def and vect_external_def
7193 for variable-length vectors to boolean types, or types for which
7194 can_duplicate_and_interleave_p is false.
7195 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7196 (duplicate_and_interleave): New function.
7197 (vect_get_constant_vectors): Use gimple_build_vector for
7198 constant-length vectors and suitable variable-length constant
7199 vectors. Use duplicate_and_interleave for other variable-length
7200 vectors. Don't defer the update when inserting new statements.
7202 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7203 Alan Hayward <alan.hayward@arm.com>
7204 David Sherwood <david.sherwood@arm.com>
7206 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7207 min_profitable_iters doesn't go negative.
7209 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7210 Alan Hayward <alan.hayward@arm.com>
7211 David Sherwood <david.sherwood@arm.com>
7213 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7214 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7215 * optabs.def (vec_mask_load_lanes_optab): New optab.
7216 (vec_mask_store_lanes_optab): Likewise.
7217 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7218 (MASK_STORE_LANES): Likewise.
7219 * internal-fn.c (mask_load_lanes_direct): New macro.
7220 (mask_store_lanes_direct): Likewise.
7221 (expand_mask_load_optab_fn): Handle masked operations.
7222 (expand_mask_load_lanes_optab_fn): New macro.
7223 (expand_mask_store_optab_fn): Handle masked operations.
7224 (expand_mask_store_lanes_optab_fn): New macro.
7225 (direct_mask_load_lanes_optab_supported_p): Likewise.
7226 (direct_mask_store_lanes_optab_supported_p): Likewise.
7227 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7229 (vect_load_lanes_supported): Likewise.
7230 * tree-vect-data-refs.c (strip_conversion): New function.
7231 (can_group_stmts_p): Likewise.
7232 (vect_analyze_data_ref_accesses): Use it instead of checking
7233 for a pair of assignments.
7234 (vect_store_lanes_supported): Take a masked_p parameter.
7235 (vect_load_lanes_supported): Likewise.
7236 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7237 vect_store_lanes_supported and vect_load_lanes_supported.
7238 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7239 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7240 parameter. Don't allow gaps for masked accesses.
7241 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7242 and vect_load_lanes_supported.
7243 (get_load_store_type): Take a masked_p parameter and update
7244 call to get_group_load_store_type.
7245 (vectorizable_store): Update call to get_load_store_type.
7246 Handle IFN_MASK_STORE_LANES.
7247 (vectorizable_load): Update call to get_load_store_type.
7248 Handle IFN_MASK_LOAD_LANES.
7250 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7251 Alan Hayward <alan.hayward@arm.com>
7252 David Sherwood <david.sherwood@arm.com>
7254 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7256 * config/aarch64/aarch64-protos.h
7257 (aarch64_sve_struct_memory_operand_p): Declare.
7258 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7259 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7260 (VPRED, vpred): Handle SVE structure modes.
7261 * config/aarch64/constraints.md (Utx): New constraint.
7262 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7263 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7264 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7265 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7266 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7267 structure modes. Split into pieces after RA.
7268 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7269 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7271 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7272 SVE structure modes.
7273 (aarch64_classify_address): Likewise.
7274 (sizetochar): Move earlier in file.
7275 (aarch64_print_operand): Handle SVE register lists.
7276 (aarch64_array_mode): New function.
7277 (aarch64_sve_struct_memory_operand_p): Likewise.
7278 (TARGET_ARRAY_MODE): Redefine.
7280 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7281 Alan Hayward <alan.hayward@arm.com>
7282 David Sherwood <david.sherwood@arm.com>
7284 * target.def (array_mode): New target hook.
7285 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7286 * doc/tm.texi: Regenerate.
7287 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7288 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7289 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7291 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7294 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7295 Alan Hayward <alan.hayward@arm.com>
7296 David Sherwood <david.sherwood@arm.com>
7298 * fold-const.c (fold_binary_loc): Check the argument types
7299 rather than the result type when testing for a vector operation.
7301 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7303 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7304 * doc/tm.texi: Regenerate.
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 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7311 (sve): Document new AArch64 extension.
7312 * doc/md.texi (w): Extend the description of the AArch64
7313 constraint to include SVE vectors.
7314 (Upl, Upa): Document new AArch64 predicate constraints.
7315 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7317 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7318 (msve-vector-bits=): New option.
7319 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7320 SVE when these are disabled.
7321 (sve): New extension.
7322 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7323 modes. Adjust their number of units based on aarch64_sve_vg.
7324 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7325 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7326 aarch64_addr_query_type.
7327 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7328 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7329 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7330 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7331 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7332 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7333 (aarch64_simd_imm_zero_p): Delete.
7334 (aarch64_check_zero_based_sve_index_immediate): Declare.
7335 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7336 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7337 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7338 (aarch64_sve_float_mul_immediate_p): Likewise.
7339 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7341 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7342 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7343 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7344 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7345 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7346 (aarch64_regmode_natural_size): Likewise.
7347 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7348 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7350 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7351 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7352 for VG and the SVE predicate registers.
7353 (V_ALIASES): Add a "z"-prefixed alias.
7354 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7355 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7356 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7357 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7358 (REG_CLASS_NAMES): Add entries for them.
7359 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7360 and the predicate registers.
7361 (aarch64_sve_vg): Declare.
7362 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7363 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7364 (REGMODE_NATURAL_SIZE): Define.
7365 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7367 * config/aarch64/aarch64.c: Include cfgrtl.h.
7368 (simd_immediate_info): Add a constructor for series vectors,
7369 and an associated step field.
7370 (aarch64_sve_vg): New variable.
7371 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7372 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7373 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7374 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7375 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7376 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7377 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7378 (aarch64_get_mask_mode): New functions.
7379 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7380 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7381 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7382 predicate modes and predicate registers. Explicitly restrict
7383 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7384 to store a vector mode if it is recognized by
7385 aarch64_classify_vector_mode.
7386 (aarch64_regmode_natural_size): New function.
7387 (aarch64_hard_regno_caller_save_mode): Return the original mode
7389 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7390 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7391 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7392 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7394 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7395 does not overlap dest if the function is frame-related. Handle
7397 (aarch64_split_add_offset): New function.
7398 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7399 them aarch64_add_offset.
7400 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7401 and update call to aarch64_sub_sp.
7402 (aarch64_add_cfa_expression): New function.
7403 (aarch64_expand_prologue): Pass extra temporary registers to the
7404 functions above. Handle the case in which we need to emit new
7405 DW_CFA_expressions for registers that were originally saved
7406 relative to the stack pointer, but now have to be expressed
7407 relative to the frame pointer.
7408 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7410 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7411 IP0 and IP1 values for SVE frames.
7412 (aarch64_expand_vec_series): New function.
7413 (aarch64_expand_sve_widened_duplicate): Likewise.
7414 (aarch64_expand_sve_const_vector): Likewise.
7415 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7416 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7417 into the register, rather than emitting a SET directly.
7418 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7419 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7420 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7421 (offset_9bit_signed_scaled_p): New functions.
7422 (aarch64_replicate_bitmask_imm): New function.
7423 (aarch64_bitmask_imm): Use it.
7424 (aarch64_cannot_force_const_mem): Reject expressions involving
7425 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7426 (aarch64_classify_index): Handle SVE indices, by requiring
7427 a plain register index with a scale that matches the element size.
7428 (aarch64_classify_address): Handle SVE addresses. Assert that
7429 the mode of the address is VOIDmode or an integer mode.
7430 Update call to aarch64_classify_symbol.
7431 (aarch64_classify_symbolic_expression): Update call to
7432 aarch64_classify_symbol.
7433 (aarch64_const_vec_all_in_range_p): New function.
7434 (aarch64_print_vector_float_operand): Likewise.
7435 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7436 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7437 and the FP immediates 1.0 and 0.5.
7438 (aarch64_print_address_internal): Handle SVE addresses.
7439 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7440 (aarch64_regno_regclass): Handle predicate registers.
7441 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7443 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7444 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7445 (aarch64_convert_sve_vector_bits): New function.
7446 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7447 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7449 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7450 Handle SVE vector and predicate modes. Accept VL-based constants
7451 that need only one temporary register, and VL offsets that require
7452 no temporary registers.
7453 (aarch64_conditional_register_usage): Mark the predicate registers
7454 as fixed if SVE isn't available.
7455 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7456 Return true for SVE vector and predicate modes.
7457 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7458 rather than an unsigned int. Handle SVE modes.
7459 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7461 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7463 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7464 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7465 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7466 (aarch64_sve_float_mul_immediate_p): New functions.
7467 (aarch64_sve_valid_immediate): New function.
7468 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7469 Explicitly reject structure modes. Check for INDEX constants.
7470 Handle PTRUE and PFALSE constants.
7471 (aarch64_check_zero_based_sve_index_immediate): New function.
7472 (aarch64_simd_imm_zero_p): Delete.
7473 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7474 vector modes. Accept constants in the range of CNT[BHWD].
7475 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7476 ask for an Advanced SIMD mode.
7477 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7478 (aarch64_simd_vector_alignment): Handle SVE predicates.
7479 (aarch64_vectorize_preferred_vector_alignment): New function.
7480 (aarch64_simd_vector_alignment_reachable): Use it instead of
7482 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7483 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7485 (MAX_VECT_LEN): Delete.
7486 (expand_vec_perm_d): Add a vec_flags field.
7487 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7488 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7489 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7491 (aarch64_evpc_rev): Rename to...
7492 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7493 (aarch64_evpc_rev_global): New function.
7494 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7495 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7497 (aarch64_evpc_sve_tbl): New function.
7498 (aarch64_expand_vec_perm_const_1): Update after rename of
7499 aarch64_evpc_rev. Handle SVE permutes too, trying
7500 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7501 than aarch64_evpc_tbl.
7502 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7503 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7504 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7505 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7506 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7507 (aarch64_expand_sve_vcond): New functions.
7508 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7509 of aarch64_vector_mode_p.
7510 (aarch64_dwarf_poly_indeterminate_value): New function.
7511 (aarch64_compute_pressure_classes): Likewise.
7512 (aarch64_can_change_mode_class): Likewise.
7513 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7514 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7515 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7516 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7517 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7518 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7519 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7520 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7522 (Dn, Dl, Dr): Accept const as well as const_vector.
7523 (Dz): Likewise. Compare against CONST0_RTX.
7524 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7525 of "vector" where appropriate.
7526 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7527 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7528 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7529 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7530 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7531 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7532 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7533 (v_int_equiv): Extend to SVE modes.
7534 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7536 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7537 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7538 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7539 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7540 (SVE_COND_FP_CMP): New int iterators.
7541 (perm_hilo): Handle the new unpack unspecs.
7542 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7544 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7545 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7546 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7547 (aarch64_equality_operator, aarch64_constant_vector_operand)
7548 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7549 (aarch64_sve_nonimmediate_operand): Likewise.
7550 (aarch64_sve_general_operand): Likewise.
7551 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7552 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7553 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7554 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7555 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7556 (aarch64_sve_float_arith_immediate): Likewise.
7557 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7558 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7559 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7560 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7561 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7562 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7563 (aarch64_sve_float_arith_operand): Likewise.
7564 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7565 (aarch64_sve_float_mul_operand): Likewise.
7566 (aarch64_sve_vec_perm_operand): Likewise.
7567 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7568 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7569 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7570 as well as const_vector.
7571 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7572 in file. Use CONST0_RTX and CONSTM1_RTX.
7573 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7574 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7575 Use aarch64_simd_imm_zero.
7576 * config/aarch64/aarch64-sve.md: New file.
7577 * config/aarch64/aarch64.md: Include it.
7578 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7579 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7580 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7581 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7582 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7583 (sve): New attribute.
7584 (enabled): Disable instructions with the sve attribute unless
7586 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7587 aarch64_expand_mov_immediate.
7588 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7589 CNT[BHSD] immediates.
7590 (movti): Split CONST_POLY_INT moves into two halves.
7591 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7592 Split additions that need a temporary here if the destination
7593 is the stack pointer.
7594 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7595 (*add<mode>3_poly_1): New instruction.
7596 (set_clobber_cc): New expander.
7598 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7600 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7601 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7602 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7603 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7604 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7605 Change innermode from fixed_mode_size to machine_mode.
7606 (simplify_subreg): Update call accordingly. Handle a constant-sized
7607 subreg of a variable-length CONST_VECTOR.
7609 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7610 Alan Hayward <alan.hayward@arm.com>
7611 David Sherwood <david.sherwood@arm.com>
7613 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7614 (add_offset_to_base): New function, split out from...
7615 (create_mem_ref): ...here. When handling a scale other than 1,
7616 check first whether the address is valid without the offset.
7617 Add it into the base if so, leaving the index and scale as-is.
7619 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7622 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7623 fold_for_warn before checking if arg2 is INTEGER_CST.
7625 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7627 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7628 (store_multiple_operation): Delete.
7629 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7630 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7631 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7632 guarded by TARGET_STRING.
7633 (rs6000_output_load_multiple): Delete.
7634 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7635 OPTION_MASK_STRING / TARGET_STRING handling.
7636 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7637 (const rs6000_opt_masks) <"string">: Change mask to 0.
7638 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7639 (MASK_STRING): Delete.
7640 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7642 (load_multiple): Delete.
7649 (store_multiple): Delete.
7656 (movmemsi_8reg): Delete.
7657 (corresponding unnamed define_insn): Delete.
7658 (movmemsi_6reg): Delete.
7659 (corresponding unnamed define_insn): Delete.
7660 (movmemsi_4reg): Delete.
7661 (corresponding unnamed define_insn): Delete.
7662 (movmemsi_2reg): Delete.
7663 (corresponding unnamed define_insn): Delete.
7664 (movmemsi_1reg): Delete.
7665 (corresponding unnamed define_insn): Delete.
7666 * config/rs6000/rs6000.opt (mno-string): New.
7667 (mstring): Replace by deprecation warning stub.
7668 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7670 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7672 * regrename.c (regrename_do_replace): If replacing the same
7673 reg multiple times, try to reuse last created gen_raw_REG.
7676 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7677 main to workaround a bug in GDB.
7679 2018-01-12 Tom de Vries <tom@codesourcery.com>
7682 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7684 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7686 PR rtl-optimization/80481
7687 * ira-color.c (get_cap_member): New function.
7688 (allocnos_conflict_by_live_ranges_p): Use it.
7689 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7690 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7692 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7695 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7696 (*saddl_se_1): Ditto.
7698 (*ssubl_se_1): Ditto.
7700 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7702 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7703 rather than wi::to_widest for DR_INITs.
7704 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7705 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7706 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7708 (vect_analyze_group_access_1): Note that here.
7710 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7712 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7713 polynomial type sizes.
7715 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7717 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7718 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7719 (gimple_add_tmp_var): Likewise.
7721 2018-01-12 Martin Liska <mliska@suse.cz>
7723 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7724 (gimple_alloc_sizes): Likewise.
7725 (dump_gimple_statistics): Use PRIu64 in printf format.
7726 * gimple.h: Change uint64_t to int.
7728 2018-01-12 Martin Liska <mliska@suse.cz>
7730 * tree-core.h: Use uint64_t instead of int.
7731 * tree.c (tree_node_counts): Likewise.
7732 (tree_node_sizes): Likewise.
7733 (dump_tree_statistics): Use PRIu64 in printf format.
7735 2018-01-12 Martin Liska <mliska@suse.cz>
7737 * Makefile.in: As qsort_chk is implemented in vec.c, add
7738 vec.o to linkage of gencfn-macros.
7739 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7740 passing the info to record_node_allocation_statistics.
7741 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7743 * ggc-common.c (struct ggc_usage): Add operator== and use
7744 it in operator< and compare function.
7745 * mem-stats.h (struct mem_usage): Likewise.
7746 * vec.c (struct vec_usage): Remove operator< and compare
7747 function. Can be simply inherited.
7749 2018-01-12 Martin Jambor <mjambor@suse.cz>
7752 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7753 * tree-ssa-math-opts.c: Include domwalk.h.
7754 (convert_mult_to_fma_1): New function.
7755 (fma_transformation_info): New type.
7756 (fma_deferring_state): Likewise.
7757 (cancel_fma_deferring): New function.
7758 (result_of_phi): Likewise.
7759 (last_fma_candidate_feeds_initial_phi): Likewise.
7760 (convert_mult_to_fma): Added deferring logic, split actual
7761 transformation to convert_mult_to_fma_1.
7762 (math_opts_dom_walker): New type.
7763 (math_opts_dom_walker::after_dom_children): New method, body moved
7764 here from pass_optimize_widening_mul::execute, added deferring logic
7766 (pass_optimize_widening_mul::execute): Moved most of code to
7767 math_opts_dom_walker::after_dom_children.
7768 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7769 * config/i386/i386.c (ix86_option_override_internal): Added
7770 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7772 2018-01-12 Richard Biener <rguenther@suse.de>
7775 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7776 inline instance vars.
7778 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7781 * config/rx/rx.c (rx_is_restricted_memory_address):
7784 2018-01-12 Richard Biener <rguenther@suse.de>
7786 PR tree-optimization/80846
7787 * target.def (split_reduction): New target hook.
7788 * targhooks.c (default_split_reduction): New function.
7789 * targhooks.h (default_split_reduction): Declare.
7790 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7791 target requests first reduce vectors by combining low and high
7793 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7794 (get_vectype_for_scalar_type_and_size): Export.
7795 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7796 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7797 * doc/tm.texi: Regenerate.
7798 * config/i386/i386.c (ix86_split_reduction): Implement
7799 TARGET_VECTORIZE_SPLIT_REDUCTION.
7801 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7804 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7805 in PIC mode except for TARGET_VXWORKS_RTP.
7806 * config/sparc/sparc.c: Include cfgrtl.h.
7807 (TARGET_INIT_PIC_REG): Define.
7808 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7809 (sparc_pic_register_p): New predicate.
7810 (sparc_legitimate_address_p): Use it.
7811 (sparc_legitimize_pic_address): Likewise.
7812 (sparc_delegitimize_address): Likewise.
7813 (sparc_mode_dependent_address_p): Likewise.
7814 (gen_load_pcrel_sym): Remove 4th parameter.
7815 (load_got_register): Adjust call to above. Remove obsolete stuff.
7816 (sparc_expand_prologue): Do not call load_got_register here.
7817 (sparc_flat_expand_prologue): Likewise.
7818 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7819 (sparc_use_pseudo_pic_reg): New function.
7820 (sparc_init_pic_reg): Likewise.
7821 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7822 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7824 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7826 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7827 Add item for branch_cost.
7829 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7831 PR rtl-optimization/83565
7832 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7833 not extend the result to a larger mode for rotate operations.
7834 (num_sign_bit_copies1): Likewise.
7836 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7839 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7841 Use values-Xc.o for -pedantic.
7842 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7844 2018-01-12 Martin Liska <mliska@suse.cz>
7847 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7849 (possible_polymorphic_call_targets): Use it.
7850 (ipa_devirt): Likewise.
7852 2018-01-12 Martin Liska <mliska@suse.cz>
7854 * profile-count.h (enum profile_quality): Use 0 as invalid
7855 enum value of profile_quality.
7857 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7859 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7860 -mext-string options.
7862 2018-01-12 Richard Biener <rguenther@suse.de>
7864 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7865 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7866 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7868 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7870 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7872 * configure.ac (--with-long-double-format): Add support for the
7873 configuration option to change the default long double format on
7875 * config.gcc (powerpc*-linux*-*): Likewise.
7876 * configure: Regenerate.
7877 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7878 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7879 used without modification.
7881 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7883 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7884 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7885 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7886 MISC_BUILTIN_SPEC_BARRIER.
7887 (rs6000_init_builtins): Likewise.
7888 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7890 (speculation_barrier): New define_insn.
7891 * doc/extend.texi: Document __builtin_speculation_barrier.
7893 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7896 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7897 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7898 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7900 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7901 integral modes instead of "ss" and "sd".
7902 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7903 vectors with 32-bit and 64-bit elements.
7904 (vecdupssescalarmodesuffix): New mode attribute.
7905 (vec_dup<mode>): Use it.
7907 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7910 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7911 frame if argument is passed on stack.
7913 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7916 * ree.c (combine_reaching_defs): Optimize also
7917 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7918 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7920 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7923 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7925 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7928 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7929 after they are computed.
7931 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7933 PR tree-optimization/83695
7934 * gimple-loop-linterchange.cc
7935 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7936 reset cached scev information after interchange.
7937 (pass_linterchange::execute): Remove call to scev_reset_htab.
7939 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7941 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7942 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7943 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7944 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7945 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7946 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7947 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7948 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7949 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7950 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7951 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7952 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7953 (V_lane_reg): Likewise.
7954 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7956 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7957 (vfmal_lane_low<mode>_intrinsic,
7958 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7959 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7960 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7961 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7962 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7963 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7965 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7967 * config/arm/arm-cpus.in (fp16fml): New feature.
7968 (ALL_SIMD): Add fp16fml.
7969 (armv8.2-a): Add fp16fml as an option.
7970 (armv8.3-a): Likewise.
7971 (armv8.4-a): Add fp16fml as part of fp16.
7972 * config/arm/arm.h (TARGET_FP16FML): Define.
7973 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7975 * config/arm/arm-modes.def (V2HF): Define.
7976 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7977 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7978 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7979 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7980 vfmsl_low, vfmsl_high): New set of builtins.
7981 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7982 (vfml_op): New code attribute.
7983 (VFMLHALVES): New int iterator.
7984 (VFML, VFMLSEL): New mode attributes.
7985 (V_reg): Define mapping for V2HF.
7986 (V_hi, V_lo): New mode attributes.
7987 (VF_constraint): Likewise.
7988 (vfml_half, vfml_half_selector): New int attributes.
7989 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7991 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7992 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7994 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7995 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7996 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7997 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7999 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
8000 Document new effective target and option set.
8002 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8004 * config/arm/arm-cpus.in (armv8_4): New feature.
8005 (ARMv8_4a): New fgroup.
8006 (armv8.4-a): New arch.
8007 * config/arm/arm-tables.opt: Regenerate.
8008 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
8009 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
8010 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
8011 Add matching rules for -march=armv8.4-a and extensions.
8012 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
8014 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
8017 * config/rx/rx.md (BW): New mode attribute.
8018 (sync_lock_test_and_setsi): Add mode suffix to insn output.
8020 2018-01-11 Richard Biener <rguenther@suse.de>
8022 PR tree-optimization/83435
8023 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
8024 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
8025 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
8027 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8028 Alan Hayward <alan.hayward@arm.com>
8029 David Sherwood <david.sherwood@arm.com>
8031 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
8033 (aarch64_classify_address): Initialize it. Track polynomial offsets.
8034 (aarch64_print_address_internal): Use it to check for a zero offset.
8036 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8037 Alan Hayward <alan.hayward@arm.com>
8038 David Sherwood <david.sherwood@arm.com>
8040 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
8041 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
8042 Return a poly_int64 rather than a HOST_WIDE_INT.
8043 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
8044 rather than a HOST_WIDE_INT.
8045 * config/aarch64/aarch64.h (aarch64_frame): Protect with
8046 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
8047 hard_fp_offset, frame_size, initial_adjust, callee_offset and
8048 final_offset from HOST_WIDE_INT to poly_int64.
8049 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
8050 to_constant when getting the number of units in an Advanced SIMD
8052 (aarch64_builtin_vectorized_function): Check for a constant number
8054 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
8056 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
8057 attribute instead of GET_MODE_NUNITS.
8058 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
8059 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
8060 GET_MODE_SIZE for fixed-size registers.
8061 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
8062 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
8063 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
8064 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
8065 (aarch64_print_operand, aarch64_print_address_internal)
8066 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
8067 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
8068 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
8069 Handle polynomial GET_MODE_SIZE.
8070 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
8071 wider than SImode without modification.
8072 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
8073 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
8074 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
8075 passing and returning SVE modes.
8076 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
8077 rather than GEN_INT.
8078 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
8079 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
8080 (aarch64_allocate_and_probe_stack_space): Likewise.
8081 (aarch64_layout_frame): Cope with polynomial offsets.
8082 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
8083 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
8085 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
8086 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
8087 poly_int64 rather than a HOST_WIDE_INT.
8088 (aarch64_get_separate_components, aarch64_process_components)
8089 (aarch64_expand_prologue, aarch64_expand_epilogue)
8090 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
8091 (aarch64_anchor_offset): New function, split out from...
8092 (aarch64_legitimize_address): ...here.
8093 (aarch64_builtin_vectorization_cost): Handle polynomial
8094 TYPE_VECTOR_SUBPARTS.
8095 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
8097 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
8098 number of elements from the PARALLEL rather than the mode.
8099 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
8100 rather than GET_MODE_BITSIZE.
8101 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
8102 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
8103 (aarch64_expand_vec_perm_const_1): Handle polynomial
8104 d->perm.length () and d->perm elements.
8105 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
8106 Apply to_constant to d->perm elements.
8107 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
8108 polynomial CONST_VECTOR_NUNITS.
8109 (aarch64_move_pointer): Take amount as a poly_int64 rather
8111 (aarch64_progress_pointer): Avoid temporary variable.
8112 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
8113 the mode attribute instead of GET_MODE.
8115 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8116 Alan Hayward <alan.hayward@arm.com>
8117 David Sherwood <david.sherwood@arm.com>
8119 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
8120 x exists before using it.
8121 (aarch64_add_constant_internal): Rename to...
8122 (aarch64_add_offset_1): ...this. Replace regnum with separate
8123 src and dest rtxes. Handle the case in which they're different,
8124 including when the offset is zero. Replace scratchreg with an rtx.
8125 Use 2 additions if there is no spare register into which we can
8126 move a 16-bit constant.
8127 (aarch64_add_constant): Delete.
8128 (aarch64_add_offset): Replace reg with separate src and dest
8129 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8130 Use aarch64_add_offset_1.
8131 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8132 an rtx rather than an int. Take the delta as a poly_int64
8133 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8134 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8135 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8136 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8137 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8139 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8140 aarch64_add_constant.
8142 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8144 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8145 Use scalar_float_mode.
8147 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8149 * config/aarch64/aarch64-simd.md
8150 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8151 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8152 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8153 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8154 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8155 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8156 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8157 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8158 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8159 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8161 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8164 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8165 targ_options->x_arm_arch_string is non NULL.
8167 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8169 * config/aarch64/aarch64.h
8170 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8172 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8175 * expmed.c (emit_store_flag_force): Swap if const op0
8176 and change VOIDmode to mode of op0.
8178 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8180 PR rtl-optimization/83761
8181 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8182 than bytes to mode_for_size.
8184 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8187 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8188 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8191 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8194 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8195 when in layout mode.
8196 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8197 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8200 2018-01-10 Michael Collison <michael.collison@arm.com>
8202 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8203 * config/aarch64/aarch64-option-extension.def: Add
8204 AARCH64_OPT_EXTENSION of 'fp16fml'.
8205 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8206 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8207 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8208 * config/aarch64/constraints.md (Ui7): New constraint.
8209 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8210 (VFMLA_SEL_W): Ditto.
8213 (VFMLA16_LOW): New int iterator.
8214 (VFMLA16_HIGH): Ditto.
8215 (UNSPEC_FMLAL): New unspec.
8216 (UNSPEC_FMLSL): Ditto.
8217 (UNSPEC_FMLAL2): Ditto.
8218 (UNSPEC_FMLSL2): Ditto.
8219 (f16mac): New code attribute.
8220 * config/aarch64/aarch64-simd-builtins.def
8221 (aarch64_fmlal_lowv2sf): Ditto.
8222 (aarch64_fmlsl_lowv2sf): Ditto.
8223 (aarch64_fmlalq_lowv4sf): Ditto.
8224 (aarch64_fmlslq_lowv4sf): Ditto.
8225 (aarch64_fmlal_highv2sf): Ditto.
8226 (aarch64_fmlsl_highv2sf): Ditto.
8227 (aarch64_fmlalq_highv4sf): Ditto.
8228 (aarch64_fmlslq_highv4sf): Ditto.
8229 (aarch64_fmlal_lane_lowv2sf): Ditto.
8230 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8231 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8232 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8233 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8234 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8235 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8236 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8237 (aarch64_fmlal_lane_highv2sf): Ditto.
8238 (aarch64_fmlsl_lane_highv2sf): Ditto.
8239 (aarch64_fmlal_laneq_highv2sf): Ditto.
8240 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8241 (aarch64_fmlalq_lane_highv4sf): Ditto.
8242 (aarch64_fmlsl_lane_highv4sf): Ditto.
8243 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8244 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8245 * config/aarch64/aarch64-simd.md:
8246 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8247 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8248 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8249 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8250 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8251 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8252 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8253 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8254 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8255 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8256 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8257 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8258 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8259 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8260 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8261 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8262 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8263 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8264 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8265 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8266 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8267 (vfmlsl_low_u32): Ditto.
8268 (vfmlalq_low_u32): Ditto.
8269 (vfmlslq_low_u32): Ditto.
8270 (vfmlal_high_u32): Ditto.
8271 (vfmlsl_high_u32): Ditto.
8272 (vfmlalq_high_u32): Ditto.
8273 (vfmlslq_high_u32): Ditto.
8274 (vfmlal_lane_low_u32): Ditto.
8275 (vfmlsl_lane_low_u32): Ditto.
8276 (vfmlal_laneq_low_u32): Ditto.
8277 (vfmlsl_laneq_low_u32): Ditto.
8278 (vfmlalq_lane_low_u32): Ditto.
8279 (vfmlslq_lane_low_u32): Ditto.
8280 (vfmlalq_laneq_low_u32): Ditto.
8281 (vfmlslq_laneq_low_u32): Ditto.
8282 (vfmlal_lane_high_u32): Ditto.
8283 (vfmlsl_lane_high_u32): Ditto.
8284 (vfmlal_laneq_high_u32): Ditto.
8285 (vfmlsl_laneq_high_u32): Ditto.
8286 (vfmlalq_lane_high_u32): Ditto.
8287 (vfmlslq_lane_high_u32): Ditto.
8288 (vfmlalq_laneq_high_u32): Ditto.
8289 (vfmlslq_laneq_high_u32): Ditto.
8290 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8291 (AARCH64_FL_FOR_ARCH8_4): New.
8292 (AARCH64_ISA_F16FML): New ISA flag.
8293 (TARGET_F16FML): New feature flag for fp16fml.
8294 (doc/invoke.texi): Document new fp16fml option.
8296 2018-01-10 Michael Collison <michael.collison@arm.com>
8298 * config/aarch64/aarch64-builtins.c:
8299 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8300 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8301 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8302 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8303 (AARCH64_ISA_SHA3): New ISA flag.
8304 (TARGET_SHA3): New feature flag for sha3.
8305 * config/aarch64/iterators.md (sha512_op): New int attribute.
8306 (CRYPTO_SHA512): New int iterator.
8307 (UNSPEC_SHA512H): New unspec.
8308 (UNSPEC_SHA512H2): Ditto.
8309 (UNSPEC_SHA512SU0): Ditto.
8310 (UNSPEC_SHA512SU1): Ditto.
8311 * config/aarch64/aarch64-simd-builtins.def
8312 (aarch64_crypto_sha512hqv2di): New builtin.
8313 (aarch64_crypto_sha512h2qv2di): Ditto.
8314 (aarch64_crypto_sha512su0qv2di): Ditto.
8315 (aarch64_crypto_sha512su1qv2di): Ditto.
8316 (aarch64_eor3qv8hi): Ditto.
8317 (aarch64_rax1qv2di): Ditto.
8318 (aarch64_xarqv2di): Ditto.
8319 (aarch64_bcaxqv8hi): Ditto.
8320 * config/aarch64/aarch64-simd.md:
8321 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8322 (aarch64_crypto_sha512su0qv2di): Ditto.
8323 (aarch64_crypto_sha512su1qv2di): Ditto.
8324 (aarch64_eor3qv8hi): Ditto.
8325 (aarch64_rax1qv2di): Ditto.
8326 (aarch64_xarqv2di): Ditto.
8327 (aarch64_bcaxqv8hi): Ditto.
8328 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8329 (vsha512h2q_u64): Ditto.
8330 (vsha512su0q_u64): Ditto.
8331 (vsha512su1q_u64): Ditto.
8332 (veor3q_u16): Ditto.
8333 (vrax1q_u64): Ditto.
8335 (vbcaxq_u16): Ditto.
8336 * config/arm/types.md (crypto_sha512): New type attribute.
8337 (crypto_sha3): Ditto.
8338 (doc/invoke.texi): Document new sha3 option.
8340 2018-01-10 Michael Collison <michael.collison@arm.com>
8342 * config/aarch64/aarch64-builtins.c:
8343 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8344 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8345 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8346 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8347 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8348 (AARCH64_ISA_SM4): New ISA flag.
8349 (TARGET_SM4): New feature flag for sm4.
8350 * config/aarch64/aarch64-simd-builtins.def
8351 (aarch64_sm3ss1qv4si): Ditto.
8352 (aarch64_sm3tt1aq4si): Ditto.
8353 (aarch64_sm3tt1bq4si): Ditto.
8354 (aarch64_sm3tt2aq4si): Ditto.
8355 (aarch64_sm3tt2bq4si): Ditto.
8356 (aarch64_sm3partw1qv4si): Ditto.
8357 (aarch64_sm3partw2qv4si): Ditto.
8358 (aarch64_sm4eqv4si): Ditto.
8359 (aarch64_sm4ekeyqv4si): Ditto.
8360 * config/aarch64/aarch64-simd.md:
8361 (aarch64_sm3ss1qv4si): Ditto.
8362 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8363 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8364 (aarch64_sm4eqv4si): Ditto.
8365 (aarch64_sm4ekeyqv4si): Ditto.
8366 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8367 (sm3part_op): Ditto.
8368 (CRYPTO_SM3TT): Ditto.
8369 (CRYPTO_SM3PART): Ditto.
8370 (UNSPEC_SM3SS1): New unspec.
8371 (UNSPEC_SM3TT1A): Ditto.
8372 (UNSPEC_SM3TT1B): Ditto.
8373 (UNSPEC_SM3TT2A): Ditto.
8374 (UNSPEC_SM3TT2B): Ditto.
8375 (UNSPEC_SM3PARTW1): Ditto.
8376 (UNSPEC_SM3PARTW2): Ditto.
8377 (UNSPEC_SM4E): Ditto.
8378 (UNSPEC_SM4EKEY): Ditto.
8379 * config/aarch64/constraints.md (Ui2): New constraint.
8380 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8381 * config/arm/types.md (crypto_sm3): New type attribute.
8382 (crypto_sm4): Ditto.
8383 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8384 (vsm3tt1aq_u32): Ditto.
8385 (vsm3tt1bq_u32): Ditto.
8386 (vsm3tt2aq_u32): Ditto.
8387 (vsm3tt2bq_u32): Ditto.
8388 (vsm3partw1q_u32): Ditto.
8389 (vsm3partw2q_u32): Ditto.
8390 (vsm4eq_u32): Ditto.
8391 (vsm4ekeyq_u32): Ditto.
8392 (doc/invoke.texi): Document new sm4 option.
8394 2018-01-10 Michael Collison <michael.collison@arm.com>
8396 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8397 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8398 (AARCH64_FL_FOR_ARCH8_4): New.
8399 (AARCH64_FL_V8_4): New flag.
8400 (doc/invoke.texi): Document new armv8.4-a option.
8402 2018-01-10 Michael Collison <michael.collison@arm.com>
8404 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8405 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8406 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8407 * config/aarch64/aarch64-option-extension.def: Add
8408 AARCH64_OPT_EXTENSION of 'sha2'.
8409 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8410 (crypto): Disable sha2 and aes if crypto disabled.
8411 (crypto): Enable aes and sha2 if enabled.
8412 (simd): Disable sha2 and aes if simd disabled.
8413 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8415 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8416 (TARGET_SHA2): New feature flag for sha2.
8417 (TARGET_AES): New feature flag for aes.
8418 * config/aarch64/aarch64-simd.md:
8419 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8420 conditional on TARGET_AES.
8421 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8422 (aarch64_crypto_sha1hsi): Make pattern conditional
8424 (aarch64_crypto_sha1hv4si): Ditto.
8425 (aarch64_be_crypto_sha1hv4si): Ditto.
8426 (aarch64_crypto_sha1su1v4si): Ditto.
8427 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8428 (aarch64_crypto_sha1su0v4si): Ditto.
8429 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8430 (aarch64_crypto_sha256su0v4si): Ditto.
8431 (aarch64_crypto_sha256su1v4si): Ditto.
8432 (doc/invoke.texi): Document new aes and sha2 options.
8434 2018-01-10 Martin Sebor <msebor@redhat.com>
8436 PR tree-optimization/83781
8437 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8440 2018-01-11 Martin Sebor <msebor@gmail.com>
8441 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8443 PR tree-optimization/83501
8444 PR tree-optimization/81703
8446 * tree-ssa-strlen.c (get_string_cst): Rename...
8447 (get_string_len): ...to this. Handle global constants.
8448 (handle_char_store): Adjust.
8450 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8451 Jim Wilson <jimw@sifive.com>
8453 * config/riscv/riscv-protos.h (riscv_output_return): New.
8454 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8455 (riscv_attribute_table, riscv_output_return),
8456 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8457 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8458 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8460 (riscv_expand_prologue): Add early return for naked function.
8461 (riscv_expand_epilogue): Likewise.
8462 (riscv_function_ok_for_sibcall): Return false for naked function.
8463 (riscv_set_current_function): New.
8464 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8465 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8466 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8467 * doc/extend.texi (RISC-V Function Attributes): New.
8469 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8471 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8472 check for 128-bit long double before checking TCmode.
8473 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8474 128-bit long doubles before checking TFmode or TCmode.
8475 (FLOAT128_IBM_P): Likewise.
8477 2018-01-10 Martin Sebor <msebor@redhat.com>
8479 PR tree-optimization/83671
8480 * builtins.c (c_strlen): Unconditionally return zero for the empty
8482 Use -Warray-bounds for warnings.
8483 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8484 for non-constant array indices with COMPONENT_REF, arrays of
8485 arrays, and pointers to arrays.
8486 (gimple_fold_builtin_strlen): Determine and set length range for
8487 non-constant character arrays.
8489 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8492 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8495 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8497 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8499 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8502 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8503 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8504 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8505 indexed_or_indirect_operand predicate.
8506 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8507 (*vsx_le_perm_load_v8hi): Likewise.
8508 (*vsx_le_perm_load_v16qi): Likewise.
8509 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8510 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8511 (*vsx_le_perm_store_v8hi): Likewise.
8512 (*vsx_le_perm_store_v16qi): Likewise.
8513 (eight unnamed splitters): Likewise.
8515 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8517 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8518 * config/rs6000/emmintrin.h: Likewise.
8519 * config/rs6000/mmintrin.h: Likewise.
8520 * config/rs6000/xmmintrin.h: Likewise.
8522 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8525 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8527 * tree.c (tree_nop_conversion): Return true for location wrapper
8529 (maybe_wrap_with_location): New function.
8530 (selftest::check_strip_nops): New function.
8531 (selftest::test_location_wrappers): New function.
8532 (selftest::tree_c_tests): Call it.
8533 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8534 (maybe_wrap_with_location): New decl.
8535 (EXPR_LOCATION_WRAPPER_P): New macro.
8536 (location_wrapper_p): New inline function.
8537 (tree_strip_any_location_wrapper): New inline function.
8539 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8542 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8543 stack_realign_offset for the largest alignment of stack slot
8545 (ix86_find_max_used_stack_alignment): New function.
8546 (ix86_finalize_stack_frame_flags): Use it. Set
8547 max_used_stack_alignment if we don't realign stack.
8548 * config/i386/i386.h (machine_function): Add
8549 max_used_stack_alignment.
8551 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8553 * config/arm/arm.opt (-mbranch-cost): New option.
8554 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8557 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8560 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8561 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8563 2018-01-10 Richard Biener <rguenther@suse.de>
8566 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8567 early out so it also covers the case where we have a non-NULL
8570 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8572 PR tree-optimization/83753
8573 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8574 for non-strided grouped accesses if the number of elements is 1.
8576 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8579 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8580 * i386.h (TARGET_USE_GATHER): Define.
8581 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8583 2018-01-10 Martin Liska <mliska@suse.cz>
8586 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8587 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8589 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8590 CLEANUP_NO_PARTITIONING is not set.
8592 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8594 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8595 for vectors, as a partial revert of r254296.
8596 * rtl.h (const_vec_p): Delete.
8597 (const_vec_duplicate_p): Don't test for vector CONSTs.
8598 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8599 * expmed.c (make_tree): Likewise.
8602 * common.md (E, F): Use CONSTANT_P instead of checking for
8604 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8605 checking for CONST_VECTOR.
8607 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8610 * predict.c (force_edge_cold): Handle in more sane way edges
8613 2018-01-09 Carl Love <cel@us.ibm.com>
8615 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8617 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8618 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8619 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8620 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8621 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8622 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8623 * config/rs6000/rs6000-protos.h: Add extern defition for
8624 rs6000_generate_float2_double_code.
8625 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8627 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8628 (float2_v2df): Add define_expand.
8630 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8633 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8634 op_mode in the force_to_mode call.
8636 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8638 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8639 instead of checking each element individually.
8640 (aarch64_evpc_uzp): Likewise.
8641 (aarch64_evpc_zip): Likewise.
8642 (aarch64_evpc_ext): Likewise.
8643 (aarch64_evpc_rev): Likewise.
8644 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8645 instead of checking each element individually. Return true without
8647 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8648 whether all selected elements come from the same input, instead of
8649 checking each element individually. Remove calls to gen_rtx_REG,
8650 start_sequence and end_sequence and instead assert that no rtl is
8653 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8655 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8656 order of HIGH and CONST checks.
8658 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8660 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8661 if the destination isn't an SSA_NAME.
8663 2018-01-09 Richard Biener <rguenther@suse.de>
8665 PR tree-optimization/83668
8666 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8668 (canonicalize_loop_form): ... here, renamed from ...
8669 (canonicalize_loop_closed_ssa_form): ... this and amended to
8670 swap successor edges for loop exit blocks to make us use
8671 the RPO order we need for initial schedule generation.
8673 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8675 PR tree-optimization/64811
8676 * match.pd: When optimizing comparisons with Inf, avoid
8677 introducing or losing exceptions from comparisons with NaN.
8679 2018-01-09 Martin Liska <mliska@suse.cz>
8682 * asan.c (shadow_mem_size): Add gcc_assert.
8684 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8686 Don't save registers in main().
8689 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8690 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8691 * config/avr/avr.c (avr_set_current_function): Don't error if
8692 naked, OS_task or OS_main are specified at the same time.
8693 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8695 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8697 * common/config/avr/avr-common.c (avr_option_optimization_table):
8698 Switch on -mmain-is-OS_task for optimizing compilations.
8700 2018-01-09 Richard Biener <rguenther@suse.de>
8702 PR tree-optimization/83572
8703 * graphite.c: Include cfganal.h.
8704 (graphite_transform_loops): Connect infinite loops to exit
8705 and remove fake edges at the end.
8707 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8709 * ipa-inline.c (edge_badness): Revert accidental checkin.
8711 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8714 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8715 symbols; not inline clones.
8717 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8720 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8721 hard registers. Formatting fixes.
8723 PR preprocessor/83722
8724 * gcc.c (try_generate_repro): Pass
8725 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8726 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8729 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8730 Kito Cheng <kito.cheng@gmail.com>
8732 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8733 (riscv_leaf_function_p): Delete.
8734 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8736 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8738 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8740 (do_ifelse): New function.
8741 (do_isel): New function.
8742 (do_sub3): New function.
8743 (do_add3): New function.
8744 (do_load_mask_compare): New function.
8745 (do_overlap_load_compare): New function.
8746 (expand_compare_loop): New function.
8747 (expand_block_compare): Call expand_compare_loop() when appropriate.
8748 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8750 (-mblock-compare-inline-loop-limit): New option.
8752 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8755 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8756 Reverse order of second and third operands in first alternative.
8757 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8758 of first and second elements in UNSPEC_VPERMR vector.
8759 (altivec_expand_vec_perm_le): Likewise.
8761 2018-01-08 Jeff Law <law@redhat.com>
8763 PR rtl-optimizatin/81308
8764 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8765 (process_switch): If group_case_labels makes a change, then set
8767 (pass_convert_switch::execute): If a switch is converted, then
8768 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8770 PR rtl-optimization/81308
8771 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8774 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8776 PR target/83663 - Revert r255946
8777 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8778 generation for cases where splatting a value is not useful.
8779 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8780 across a vec_duplicate and a paradoxical subreg forming a vector
8781 mode to a vec_concat.
8783 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8785 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8786 -march=armv8.3-a variants.
8787 * config/arm/t-multilib: Likewise.
8788 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8790 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8792 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8794 (cceq_ior_compare_complement): Give it a name so I can use it, and
8795 change boolean_or_operator predicate to boolean_operator so it can
8796 be used to generate a crand.
8797 (eqne): New code iterator.
8798 (bd/bd_neg): New code_attrs.
8799 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8800 a single define_insn.
8801 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8802 decrement (bdnzt/bdnzf/bdzt/bdzf).
8803 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8804 with the new names of the branch decrement patterns, and added the
8805 names of the branch decrement conditional patterns.
8807 2018-01-08 Richard Biener <rguenther@suse.de>
8809 PR tree-optimization/83563
8810 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8813 2018-01-08 Richard Biener <rguenther@suse.de>
8816 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8818 2018-01-08 Richard Biener <rguenther@suse.de>
8820 PR tree-optimization/83685
8821 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8822 references to abnormals.
8824 2018-01-08 Richard Biener <rguenther@suse.de>
8827 * dwarf2out.c (output_indirect_strings): Handle empty
8828 skeleton_debug_str_hash.
8829 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8831 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8833 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8834 (emit_store_direct): Likewise.
8835 (arc_trampoline_adjust_address): Likewise.
8836 (arc_asm_trampoline_template): New function.
8837 (arc_initialize_trampoline): Use asm_trampoline_template.
8838 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8839 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8840 * config/arc/arc.md (flush_icache): Delete pattern.
8842 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8844 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8845 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8848 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8851 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8852 by not USED_FOR_TARGET.
8853 (make_pass_resolve_sw_modes): Likewise.
8855 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8857 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8860 2018-01-08 Richard Biener <rguenther@suse.de>
8863 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8865 2018-01-08 Richard Biener <rguenther@suse.de>
8868 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8870 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8873 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8874 basic blocks with a small number of successors.
8875 (convert_control_dep_chain_into_preds): Improve handling of
8877 (dump_predicates): Split apart into...
8878 (dump_pred_chain): ...here...
8879 (dump_pred_info): ...and here.
8880 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8881 (can_chain_union_be_invalidated_p): Improve check for invalidation
8883 (uninit_uses_cannot_happen): Avoid unnecessary if
8884 convert_control_dep_chain_into_preds yielded nothing.
8886 2018-01-06 Martin Sebor <msebor@redhat.com>
8888 PR tree-optimization/83640
8889 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8890 subtracting negative offset from size.
8891 (builtin_access::overlap): Adjust offset bounds of the access to fall
8892 within the size of the object if possible.
8894 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8896 PR rtl-optimization/83699
8897 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8898 extract_bit_field_as_subreg to cases in which the extracted
8899 value is also a vector.
8901 * lra-constraints.c (process_alt_operands): Test for the equivalence
8902 substitutions when detecting a possible reload cycle.
8904 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8907 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8908 by default if flag_selective_schedling{,2}. Formatting fixes.
8910 PR rtl-optimization/83682
8911 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8912 if it has non-VECTOR_MODE element mode.
8913 (vec_duplicate_p): Likewise.
8916 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8917 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8919 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8922 * config/i386/i386-builtin.def
8923 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8924 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8925 Require also OPTION_MASK_ISA_AVX512F in addition to
8926 OPTION_MASK_ISA_GFNI.
8927 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8928 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8929 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8930 to OPTION_MASK_ISA_GFNI.
8931 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8932 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8933 OPTION_MASK_ISA_AVX512BW.
8934 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8935 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8936 addition to OPTION_MASK_ISA_GFNI.
8937 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8938 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8939 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8940 to OPTION_MASK_ISA_GFNI.
8941 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8942 a requirement for all ISAs rather than any of them with a few
8944 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8946 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8947 bitmasks to be enabled with 3 exceptions, instead of requiring any
8948 enabled ISA with lots of exceptions.
8949 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8950 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8951 Change avx512bw in isa attribute to avx512f.
8952 * config/i386/sgxintrin.h: Add license boilerplate.
8953 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8954 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8955 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8956 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8958 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8959 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8960 temporarily sse2 rather than sse if not enabled already.
8963 * config/i386/sse.md (VI248_VLBW): Rename to ...
8964 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8965 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8966 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8967 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8968 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8969 mode iterator instead of VI248_VLBW.
8971 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8973 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8974 (record_modified): Skip clobbers; add debug output.
8975 (param_change_prob): Use sreal frequencies.
8977 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8979 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8980 punt for user-aligned variables.
8982 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8984 * tree-chrec.c (chrec_contains_symbols): Return true for
8987 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8990 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8991 of (x|y) == x for BICS pattern.
8993 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8995 PR tree-optimization/83605
8996 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8997 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
9000 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
9002 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
9003 * config/epiphany/rtems.h: New file.
9005 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9006 Uros Bizjak <ubizjak@gmail.com>
9009 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
9010 QIreg_operand instead of register_operand predicate.
9011 * config/i386/i386.c (ix86_rop_should_change_byte_p,
9012 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
9013 comments instead of -fmitigate[-_]rop.
9015 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
9018 * cgraphunit.c (symbol_table::compile): Switch to text_section
9019 before calling assembly_start debug hook.
9020 * run-rtl-passes.c (run_rtl_passes): Likewise.
9023 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9025 * tree-vrp.c (extract_range_from_binary_expr_1): Check
9026 range_int_cst_p rather than !symbolic_range_p before calling
9027 extract_range_from_multiplicative_op_1.
9029 2018-01-04 Jeff Law <law@redhat.com>
9031 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
9032 redundant test in assertion.
9034 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9036 * doc/rtl.texi: Document machine_mode wrapper classes.
9038 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9040 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
9043 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9045 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
9046 the VEC_PERM_EXPR fold to fail.
9048 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9051 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
9052 to switched_sections.
9054 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9057 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
9060 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
9063 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
9064 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
9066 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9069 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
9070 is BLKmode and bitpos not zero or mode change is needed.
9072 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
9075 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
9078 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
9081 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
9082 instead of MULT rtx. Update all corresponding splitters.
9084 (*ssub<modesuffix>): Ditto.
9086 (*cmp_sadd_di): Update split patterns.
9087 (*cmp_sadd_si): Ditto.
9088 (*cmp_sadd_sidi): Ditto.
9089 (*cmp_ssub_di): Ditto.
9090 (*cmp_ssub_si): Ditto.
9091 (*cmp_ssub_sidi): Ditto.
9092 * config/alpha/predicates.md (const23_operand): New predicate.
9093 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
9094 Look for ASHIFT, not MULT inner operand.
9095 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
9097 2018-01-04 Martin Liska <mliska@suse.cz>
9099 PR gcov-profile/83669
9100 * gcov.c (output_intermediate_file): Add version to intermediate
9102 * doc/gcov.texi: Document new field 'version' in intermediate
9103 file format. Fix location of '-k' option of gcov command.
9105 2018-01-04 Martin Liska <mliska@suse.cz>
9108 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
9110 2018-01-04 Jakub Jelinek <jakub@redhat.com>
9112 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
9114 2018-01-03 Martin Sebor <msebor@redhat.com>
9116 PR tree-optimization/83655
9117 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
9118 checking calls with invalid arguments.
9120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9122 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9123 (vectorizable_mask_load_store): Delete.
9124 (vectorizable_call): Return false for masked loads and stores.
9125 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9126 instead of gimple_assign_rhs1.
9127 (vectorizable_load): Handle IFN_MASK_LOAD.
9128 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9130 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9132 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9134 (vectorizable_mask_load_store): ...here.
9135 (vectorizable_load): ...and here.
9137 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9139 * tree-vect-stmts.c (vect_build_all_ones_mask)
9140 (vect_build_zero_merge_argument): New functions, split out from...
9141 (vectorizable_load): ...here.
9143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9145 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9147 (vectorizable_mask_load_store): ...here.
9148 (vectorizable_store): ...and here.
9150 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9152 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9154 (vectorizable_mask_load_store): ...here.
9156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9158 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9159 (vect_model_store_cost): Take a vec_load_store_type instead of a
9161 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9162 (vect_model_store_cost): Take a vec_load_store_type instead of a
9164 (vectorizable_mask_load_store): Update accordingly.
9165 (vectorizable_store): Likewise.
9166 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9168 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9170 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9171 IFN_MASK_LOAD calls here rather than...
9172 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9175 Alan Hayward <alan.hayward@arm.com>
9176 David Sherwood <david.sherwood@arm.com>
9178 * expmed.c (extract_bit_field_1): For vector extracts,
9179 fall back to extract_bit_field_as_subreg if vec_extract
9182 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9183 Alan Hayward <alan.hayward@arm.com>
9184 David Sherwood <david.sherwood@arm.com>
9186 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9187 they are variable or constant sized.
9188 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9189 slots for constant-sized data.
9191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9192 Alan Hayward <alan.hayward@arm.com>
9193 David Sherwood <david.sherwood@arm.com>
9195 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9196 handling COND_EXPRs with boolean comparisons, try to find a better
9197 basis for the mask type than the boolean itself.
9199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9201 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9202 is calculated and how it can be overridden.
9203 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9204 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9206 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9210 Alan Hayward <alan.hayward@arm.com>
9211 David Sherwood <david.sherwood@arm.com>
9213 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9214 Remove the mode argument.
9215 (aarch64_simd_valid_immediate): Remove the mode and inverse
9217 * config/aarch64/iterators.md (bitsize): New iterator.
9218 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9219 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9220 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9221 aarch64_simd_valid_immediate.
9222 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9223 (aarch64_reg_or_bic_imm): Likewise.
9224 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9225 with an insn_type enum and msl with a modifier_type enum.
9226 Replace element_width with a scalar_mode. Change the shift
9227 to unsigned int. Add constructors for scalar_float_mode and
9228 scalar_int_mode elements.
9229 (aarch64_vect_float_const_representable_p): Delete.
9230 (aarch64_can_const_movi_rtx_p)
9231 (aarch64_simd_scalar_immediate_valid_for_move)
9232 (aarch64_simd_make_constant): Update call to
9233 aarch64_simd_valid_immediate.
9234 (aarch64_advsimd_valid_immediate_hs): New function.
9235 (aarch64_advsimd_valid_immediate): Likewise.
9236 (aarch64_simd_valid_immediate): Remove mode and inverse
9237 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9238 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9239 and aarch64_float_const_representable_p on the result.
9240 (aarch64_output_simd_mov_immediate): Remove mode argument.
9241 Update call to aarch64_simd_valid_immediate and use of
9242 simd_immediate_info.
9243 (aarch64_output_scalar_simd_mov_immediate): Update call
9246 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9247 Alan Hayward <alan.hayward@arm.com>
9248 David Sherwood <david.sherwood@arm.com>
9250 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9251 (mode_nunits): Likewise CONST_MODE_NUNITS.
9252 * machmode.def (ADJUST_NUNITS): Document.
9253 * genmodes.c (mode_data::need_nunits_adj): New field.
9254 (blank_mode): Update accordingly.
9255 (adj_nunits): New variable.
9256 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9258 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9259 listed in adj_nunits.
9260 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9261 listed in adj_nunits. Don't emit case statements for such modes.
9262 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9263 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9264 nothing if adj_nunits is nonnull.
9265 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9266 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9267 (emit_mode_fbit): Update use of print_maybe_const_decl.
9268 (emit_move_size): Likewise. Treat the array as non-const
9270 (emit_mode_adjustments): Handle adj_nunits.
9272 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9274 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9275 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9276 (VECTOR_MODES): Use it.
9277 (make_vector_modes): Take the prefix as an argument.
9279 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9280 Alan Hayward <alan.hayward@arm.com>
9281 David Sherwood <david.sherwood@arm.com>
9283 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9284 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9285 for MODE_VECTOR_BOOL.
9286 * machmode.def (VECTOR_BOOL_MODE): Document.
9287 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9288 (make_vector_bool_mode): New function.
9289 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9291 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9292 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9294 * stor-layout.c (int_mode_for_mode): Likewise.
9295 * tree.c (build_vector_type_for_mode): Likewise.
9296 * varasm.c (output_constant_pool_2): Likewise.
9297 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9298 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9299 for MODE_VECTOR_BOOL.
9300 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9301 of mode class checks.
9302 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9303 instead of a list of mode class checks.
9304 (expand_vector_scalar_condition): Likewise.
9305 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9308 Alan Hayward <alan.hayward@arm.com>
9309 David Sherwood <david.sherwood@arm.com>
9311 * machmode.h (mode_size): Change from unsigned short to
9313 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9314 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9315 or if measurement_type is not polynomial.
9316 (fixed_size_mode::includes_p): Check for constant-sized modes.
9317 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9318 return a poly_uint16 rather than an unsigned short.
9319 (emit_mode_size): Change the type of mode_size from unsigned short
9320 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9321 (emit_mode_adjustments): Cope with polynomial vector sizes.
9322 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9324 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9326 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9327 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9328 * caller-save.c (setup_save_areas): Likewise.
9329 (replace_reg_with_saved_mem): Likewise.
9330 * calls.c (emit_library_call_value_1): Likewise.
9331 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9332 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9333 (gen_lowpart_for_combine): Likewise.
9334 * convert.c (convert_to_integer_1): Likewise.
9335 * cse.c (equiv_constant, cse_insn): Likewise.
9336 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9337 (cselib_subst_to_values): Likewise.
9338 * dce.c (word_dce_process_block): Likewise.
9339 * df-problems.c (df_word_lr_mark_ref): Likewise.
9340 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9341 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9342 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9343 (rtl_for_decl_location): Likewise.
9344 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9345 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9346 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9347 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9348 (expand_expr_real_1): Likewise.
9349 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9350 (pad_below): Likewise.
9351 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9352 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9353 * ira.c (get_subreg_tracking_sizes): Likewise.
9354 * ira-build.c (ira_create_allocno_objects): Likewise.
9355 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9356 (ira_sort_regnos_for_alter_reg): Likewise.
9357 * ira-costs.c (record_operand_costs): Likewise.
9358 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9359 (resolve_simple_move): Likewise.
9360 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9361 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9362 (lra_constraints): Likewise.
9363 (CONST_POOL_OK_P): Reject variable-sized modes.
9364 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9365 (add_pseudo_to_slot, lra_spill): Likewise.
9366 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9367 * optabs-query.c (get_best_extraction_insn): Likewise.
9368 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9369 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9370 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9371 * recog.c (offsettable_address_addr_space_p): Likewise.
9372 * regcprop.c (maybe_mode_change): Likewise.
9373 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9374 * regrename.c (build_def_use): Likewise.
9375 * regstat.c (dump_reg_info): Likewise.
9376 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9377 (find_reloads, find_reloads_subreg_address): Likewise.
9378 * reload1.c (eliminate_regs_1): Likewise.
9379 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9380 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9381 (simplify_binary_operation_1, simplify_subreg): Likewise.
9382 * targhooks.c (default_function_arg_padding): Likewise.
9383 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9384 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9385 (verify_gimple_assign_ternary): Likewise.
9386 * tree-inline.c (estimate_move_cost): Likewise.
9387 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9388 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9389 (get_address_cost_ainc): Likewise.
9390 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9391 (vect_supportable_dr_alignment): Likewise.
9392 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9393 (vectorizable_reduction): Likewise.
9394 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9395 (vectorizable_operation, vectorizable_load): Likewise.
9396 * tree.c (build_same_sized_truth_vector_type): Likewise.
9397 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9398 * var-tracking.c (emit_note_insn_var_location): Likewise.
9399 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9400 (ADDR_VEC_ALIGN): Likewise.
9402 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9403 Alan Hayward <alan.hayward@arm.com>
9404 David Sherwood <david.sherwood@arm.com>
9406 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9408 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9409 or if measurement_type is polynomial.
9410 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9411 * combine.c (make_extraction): Likewise.
9412 * dse.c (find_shift_sequence): Likewise.
9413 * dwarf2out.c (mem_loc_descriptor): Likewise.
9414 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9415 (extract_bit_field, extract_low_bits): Likewise.
9416 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9417 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9418 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9419 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9420 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9421 * reload.c (find_reloads): Likewise.
9422 * reload1.c (alter_reg): Likewise.
9423 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9424 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9425 * tree-if-conv.c (predicate_mem_writes): Likewise.
9426 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9427 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9428 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9429 * valtrack.c (dead_debug_insert_temp): Likewise.
9430 * varasm.c (mergeable_constant_section): Likewise.
9431 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9434 Alan Hayward <alan.hayward@arm.com>
9435 David Sherwood <david.sherwood@arm.com>
9437 * expr.c (expand_assignment): Cope with polynomial mode sizes
9438 when assigning to a CONCAT.
9440 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9441 Alan Hayward <alan.hayward@arm.com>
9442 David Sherwood <david.sherwood@arm.com>
9444 * machmode.h (mode_precision): Change from unsigned short to
9446 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9448 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9449 or if measurement_type is not polynomial.
9450 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9451 in which the mode is already known to be a scalar_int_mode.
9452 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9453 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9455 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9456 for GET_MODE_PRECISION.
9457 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9458 for GET_MODE_PRECISION.
9459 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9461 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9462 (expand_field_assignment, make_extraction): Likewise.
9463 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9464 (get_last_value): Likewise.
9465 * convert.c (convert_to_integer_1): Likewise.
9466 * cse.c (cse_insn): Likewise.
9467 * expr.c (expand_expr_real_1): Likewise.
9468 * lra-constraints.c (simplify_operand_subreg): Likewise.
9469 * optabs-query.c (can_atomic_load_p): Likewise.
9470 * optabs.c (expand_atomic_load): Likewise.
9471 (expand_atomic_store): Likewise.
9472 * ree.c (combine_reaching_defs): Likewise.
9473 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9474 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9475 * tree.h (type_has_mode_precision_p): Likewise.
9476 * ubsan.c (instrument_si_overflow): Likewise.
9478 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9479 Alan Hayward <alan.hayward@arm.com>
9480 David Sherwood <david.sherwood@arm.com>
9482 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9483 polynomial numbers of units.
9484 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9485 (valid_vector_subparts_p): New function.
9486 (build_vector_type): Remove temporary shim and take the number
9487 of units as a poly_uint64 rather than an int.
9488 (build_opaque_vector_type): Take the number of units as a
9489 poly_uint64 rather than an int.
9490 * tree.c (build_vector_from_ctor): Handle polynomial
9491 TYPE_VECTOR_SUBPARTS.
9492 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9493 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9494 (build_vector_from_val): If the number of units is variable,
9495 use build_vec_duplicate_cst for constant operands and
9496 VEC_DUPLICATE_EXPR otherwise.
9497 (make_vector_type): Remove temporary is_constant ().
9498 (build_vector_type, build_opaque_vector_type): Take the number of
9499 units as a poly_uint64 rather than an int.
9500 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9502 * cfgexpand.c (expand_debug_expr): Likewise.
9503 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9504 (store_constructor, expand_expr_real_1): Likewise.
9505 (const_scalar_mask_from_tree): Likewise.
9506 * fold-const-call.c (fold_const_reduction): Likewise.
9507 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9508 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9509 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9510 (fold_relational_const): Likewise.
9511 (native_interpret_vector): Likewise. Change the size from an
9512 int to an unsigned int.
9513 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9514 TYPE_VECTOR_SUBPARTS.
9515 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9516 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9517 duplicating a non-constant operand into a variable-length vector.
9518 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9519 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9520 * ipa-icf.c (sem_variable::equals): Likewise.
9521 * match.pd: Likewise.
9522 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9523 * print-tree.c (print_node): Likewise.
9524 * stor-layout.c (layout_type): Likewise.
9525 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9526 * tree-cfg.c (verify_gimple_comparison): Likewise.
9527 (verify_gimple_assign_binary): Likewise.
9528 (verify_gimple_assign_ternary): Likewise.
9529 (verify_gimple_assign_single): Likewise.
9530 * tree-pretty-print.c (dump_generic_node): Likewise.
9531 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9532 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9533 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9534 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9535 (vect_shift_permute_load_chain): Likewise.
9536 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9537 (expand_vector_condition, optimize_vector_constructor): Likewise.
9538 (lower_vec_perm, get_compute_type): Likewise.
9539 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9540 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9541 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9542 (vect_recog_mask_conversion_pattern): Likewise.
9543 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9544 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9545 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9546 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9547 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9548 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9549 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9550 (supportable_widening_operation): Likewise.
9551 (supportable_narrowing_operation): Likewise.
9552 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9554 * varasm.c (output_constant): Likewise.
9556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9557 Alan Hayward <alan.hayward@arm.com>
9558 David Sherwood <david.sherwood@arm.com>
9560 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9561 so that both the length == 3 and length != 3 cases set up their
9562 own permute vectors. Add comments explaining why we know the
9563 number of elements is constant.
9564 (vect_permute_load_chain): Likewise.
9566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9567 Alan Hayward <alan.hayward@arm.com>
9568 David Sherwood <david.sherwood@arm.com>
9570 * machmode.h (mode_nunits): Change from unsigned char to
9572 (ONLY_FIXED_SIZE_MODES): New macro.
9573 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9574 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9575 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9577 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9578 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9579 or if measurement_type is not polynomial.
9580 * genmodes.c (ZERO_COEFFS): New macro.
9581 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9583 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9584 Use ZERO_COEFFS when emitting initializers.
9585 * data-streamer.h (bp_pack_poly_value): New function.
9586 (bp_unpack_poly_value): Likewise.
9587 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9588 for GET_MODE_NUNITS.
9589 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9590 for GET_MODE_NUNITS.
9591 * tree.c (make_vector_type): Remove temporary shim and make
9592 the real function take the number of units as a poly_uint64
9594 (build_vector_type_for_mode): Handle polynomial nunits.
9595 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9596 * emit-rtl.c (const_vec_series_p_1): Likewise.
9597 (gen_rtx_CONST_VECTOR): Likewise.
9598 * fold-const.c (test_vec_duplicate_folding): Likewise.
9599 * genrecog.c (validate_pattern): Likewise.
9600 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9601 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9602 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9603 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9604 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9605 * rtlanal.c (subreg_get_info): Likewise.
9606 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9607 (vect_grouped_load_supported): Likewise.
9608 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9609 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9610 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9611 (simplify_const_unary_operation, simplify_binary_operation_1)
9612 (simplify_const_binary_operation, simplify_ternary_operation)
9613 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9614 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9615 instead of CONST_VECTOR_NUNITS.
9616 * varasm.c (output_constant_pool_2): Likewise.
9617 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9618 explicit-encoded elements in the XVEC for variable-length vectors.
9620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9622 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9625 Alan Hayward <alan.hayward@arm.com>
9626 David Sherwood <david.sherwood@arm.com>
9628 * coretypes.h (fixed_size_mode): Declare.
9629 (fixed_size_mode_pod): New typedef.
9630 * builtins.h (target_builtins::x_apply_args_mode)
9631 (target_builtins::x_apply_result_mode): Change type to
9632 fixed_size_mode_pod.
9633 * builtins.c (apply_args_size, apply_result_size, result_vector)
9634 (expand_builtin_apply_args_1, expand_builtin_apply)
9635 (expand_builtin_return): Update accordingly.
9637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9639 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9640 * cselib.c (cselib_hash_rtx): Likewise.
9641 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9642 CONST_VECTOR encoding.
9644 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9645 Jeff Law <law@redhat.com>
9648 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9649 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9650 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9651 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9654 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9655 explicitly probe *sp in a noreturn function if there were any callee
9656 register saves or frame pointer is needed.
9658 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9661 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9662 BLKmode for ternary, binary or unary expressions.
9665 * var-tracking.c (delete_vta_debug_insn): New inline function.
9666 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9667 insns from get_insns () to NULL instead of each bb separately.
9668 Use delete_vta_debug_insn. No longer static.
9669 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9670 delete_vta_debug_insns callers.
9671 * rtl.h (delete_vta_debug_insns): Declare.
9672 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9673 instead of variable_tracking_main.
9675 2018-01-03 Martin Sebor <msebor@redhat.com>
9677 PR tree-optimization/83603
9678 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9679 arguments past the endof the argument list in functions declared
9680 without a prototype.
9681 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9682 Avoid checking when arguments are null.
9684 2018-01-03 Martin Sebor <msebor@redhat.com>
9687 * doc/extend.texi (attribute const): Fix a typo.
9688 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9689 issuing -Wsuggest-attribute for void functions.
9691 2018-01-03 Martin Sebor <msebor@redhat.com>
9693 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9694 offset_int::from instead of wide_int::to_shwi.
9695 (maybe_diag_overlap): Remove assertion.
9696 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9697 * gimple-ssa-sprintf.c (format_directive): Same.
9698 (parse_directive): Same.
9699 (sprintf_dom_walker::compute_format_length): Same.
9700 (try_substitute_return_value): Same.
9702 2018-01-03 Jeff Law <law@redhat.com>
9705 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9706 non-constant residual for zero at runtime and avoid probing in
9707 that case. Reorganize code for trailing problem to mirror handling
9710 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9712 PR tree-optimization/83501
9713 * tree-ssa-strlen.c (get_string_cst): New.
9714 (handle_char_store): Call get_string_cst.
9716 2018-01-03 Martin Liska <mliska@suse.cz>
9718 PR tree-optimization/83593
9719 * tree-ssa-strlen.c: Include tree-cfg.h.
9720 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9721 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9722 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9724 (strlen_dom_walker::before_dom_children): Call
9725 gimple_purge_dead_eh_edges. Dump tranformation with details
9727 (strlen_dom_walker::before_dom_children): Update call by adding
9728 new argument cleanup_eh.
9729 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9731 2018-01-03 Martin Liska <mliska@suse.cz>
9734 * cif-code.def (VARIADIC_THUNK): New enum value.
9735 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9738 2018-01-03 Jan Beulich <jbeulich@suse.com>
9740 * sse.md (mov<mode>_internal): Tighten condition for when to use
9741 vmovdqu<ssescalarsize> for TI and OI modes.
9743 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9745 Update copyright years.
9747 2018-01-03 Martin Liska <mliska@suse.cz>
9750 * ipa-visibility.c (function_and_variable_visibility): Skip
9751 functions with noipa attribure.
9753 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9755 * gcc.c (process_command): Update copyright notice dates.
9756 * gcov-dump.c (print_version): Ditto.
9757 * gcov.c (print_version): Ditto.
9758 * gcov-tool.c (print_version): Ditto.
9759 * gengtype.c (create_file): Ditto.
9760 * doc/cpp.texi: Bump @copying's copyright year.
9761 * doc/cppinternals.texi: Ditto.
9762 * doc/gcc.texi: Ditto.
9763 * doc/gccint.texi: Ditto.
9764 * doc/gcov.texi: Ditto.
9765 * doc/install.texi: Ditto.
9766 * doc/invoke.texi: Ditto.
9768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9770 * vector-builder.h (vector_builder::m_full_nelts): Change from
9771 unsigned int to poly_uint64.
9772 (vector_builder::full_nelts): Update prototype accordingly.
9773 (vector_builder::new_vector): Likewise.
9774 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9775 (vector_builder::operator ==): Likewise.
9776 (vector_builder::finalize): Likewise.
9777 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9778 Take the number of elements as a poly_uint64 rather than an
9780 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9781 from unsigned int to poly_uint64.
9782 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9783 (vec_perm_indices::new_vector): Likewise.
9784 (vec_perm_indices::length): Likewise.
9785 (vec_perm_indices::nelts_per_input): Likewise.
9786 (vec_perm_indices::input_nelts): Likewise.
9787 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9788 number of elements per input as a poly_uint64 rather than an
9789 unsigned int. Use the original encoding for variable-length
9790 vectors, rather than clamping each individual element.
9791 For the second and subsequent elements in each pattern,
9792 clamp the step and base before clamping their sum.
9793 (vec_perm_indices::series_p): Handle polynomial element counts.
9794 (vec_perm_indices::all_in_range_p): Likewise.
9795 (vec_perm_indices_to_tree): Likewise.
9796 (vec_perm_indices_to_rtx): Likewise.
9797 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9798 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9799 (tree_vector_builder::new_binary_operation): Handle polynomial
9800 element counts. Return false if we need to know the number
9801 of elements at compile time.
9802 * fold-const.c (fold_vec_perm): Punt if the number of elements
9803 isn't known at compile time.
9805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9807 * vec-perm-indices.h (vec_perm_builder): Change element type
9808 from HOST_WIDE_INT to poly_int64.
9809 (vec_perm_indices::element_type): Update accordingly.
9810 (vec_perm_indices::clamp): Handle polynomial element_types.
9811 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9812 (vec_perm_indices::all_in_range_p): Likewise.
9813 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9815 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9816 polynomial vec_perm_indices element types.
9817 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9818 * fold-const.c (fold_vec_perm): Likewise.
9819 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9820 * tree-vect-generic.c (lower_vec_perm): Likewise.
9821 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9822 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9823 element type to HOST_WIDE_INT.
9825 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9826 Alan Hayward <alan.hayward@arm.com>
9827 David Sherwood <david.sherwood@arm.com>
9829 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9830 rather than an int. Use plus_constant.
9831 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9832 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9835 Alan Hayward <alan.hayward@arm.com>
9836 David Sherwood <david.sherwood@arm.com>
9838 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9839 a HOST_WIDE_INT to a poly_int64.
9841 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9842 Alan Hayward <alan.hayward@arm.com>
9843 David Sherwood <david.sherwood@arm.com>
9845 * calls.c (load_register_parameters): Cope with polynomial
9846 mode sizes. Require a constant size for BLKmode parameters
9847 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9848 forces a parameter to be padded at the lsb end in order to
9849 fill a complete number of words, require the parameter size
9850 to be ordered wrt UNITS_PER_WORD.
9852 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9853 Alan Hayward <alan.hayward@arm.com>
9854 David Sherwood <david.sherwood@arm.com>
9856 * reload1.c (spill_stack_slot_width): Change element type
9857 from unsigned int to poly_uint64_pod.
9858 (alter_reg): Treat mode sizes as polynomial.
9860 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9861 Alan Hayward <alan.hayward@arm.com>
9862 David Sherwood <david.sherwood@arm.com>
9864 * reload.c (complex_word_subreg_p): New function.
9865 (reload_inner_reg_of_subreg, push_reload): Use it.
9867 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9868 Alan Hayward <alan.hayward@arm.com>
9869 David Sherwood <david.sherwood@arm.com>
9871 * lra-constraints.c (process_alt_operands): Reject matched
9872 operands whose sizes aren't ordered.
9873 (match_reload): Refer to this check here.
9875 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9876 Alan Hayward <alan.hayward@arm.com>
9877 David Sherwood <david.sherwood@arm.com>
9879 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9880 that the mode size is in the set {1, 2, 4, 8, 16}.
9882 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9883 Alan Hayward <alan.hayward@arm.com>
9884 David Sherwood <david.sherwood@arm.com>
9886 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9887 Use plus_constant instead of gen_rtx_PLUS.
9889 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9890 Alan Hayward <alan.hayward@arm.com>
9891 David Sherwood <david.sherwood@arm.com>
9893 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9894 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9895 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9896 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9897 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9898 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9899 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9900 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9901 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9902 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9904 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9905 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9906 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9907 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9908 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9909 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9910 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9911 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9912 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9913 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9915 * expr.c (emit_move_resolve_push): Treat the input and result
9916 of PUSH_ROUNDING as a poly_int64.
9917 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9918 (emit_push_insn): Likewise.
9919 * lra-eliminations.c (mark_not_eliminable): Likewise.
9920 * recog.c (push_operand): Likewise.
9921 * reload1.c (elimination_effects): Likewise.
9922 * rtlanal.c (nonzero_bits1): Likewise.
9923 * calls.c (store_one_arg): Likewise. Require the padding to be
9924 known at compile time.
9926 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9927 Alan Hayward <alan.hayward@arm.com>
9928 David Sherwood <david.sherwood@arm.com>
9930 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9931 Use plus_constant instead of gen_rtx_PLUS.
9933 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9934 Alan Hayward <alan.hayward@arm.com>
9935 David Sherwood <david.sherwood@arm.com>
9937 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9941 Alan Hayward <alan.hayward@arm.com>
9942 David Sherwood <david.sherwood@arm.com>
9944 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9945 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9946 via stack temporaries. Treat the mode size as polynomial too.
9948 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9949 Alan Hayward <alan.hayward@arm.com>
9950 David Sherwood <david.sherwood@arm.com>
9952 * expr.c (expand_expr_real_2): When handling conversions involving
9953 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9954 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9955 as a poly_uint64 too.
9957 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9958 Alan Hayward <alan.hayward@arm.com>
9959 David Sherwood <david.sherwood@arm.com>
9961 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
9968 REGMODE_NATURAL_SIZE.
9969 * expmed.c (store_bit_field_1): Likewise.
9970 * expr.c (store_constructor): Likewise.
9971 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9972 and polynomial REGMODE_NATURAL_SIZE.
9973 (gen_lowpart_common): Likewise.
9974 * reginfo.c (record_subregs_of_mode): Likewise.
9975 * rtlanal.c (read_modify_subreg_p): Likewise.
9977 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9978 Alan Hayward <alan.hayward@arm.com>
9979 David Sherwood <david.sherwood@arm.com>
9981 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9982 numbers of elements.
9984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9985 Alan Hayward <alan.hayward@arm.com>
9986 David Sherwood <david.sherwood@arm.com>
9988 * match.pd: Cope with polynomial numbers of vector elements.
9990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9991 Alan Hayward <alan.hayward@arm.com>
9992 David Sherwood <david.sherwood@arm.com>
9994 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9995 in a POINTER_PLUS_EXPR.
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 * omp-simd-clone.c (simd_clone_subparts): New function.
10002 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
10003 (ipa_simd_modify_function_body): Likewise.
10005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10006 Alan Hayward <alan.hayward@arm.com>
10007 David Sherwood <david.sherwood@arm.com>
10009 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
10010 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
10011 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
10012 (expand_vector_condition, vector_element): Likewise.
10013 (subparts_gt): New function.
10014 (get_compute_type): Use subparts_gt.
10015 (count_type_subparts): Delete.
10016 (expand_vector_operations_1): Use subparts_gt instead of
10017 count_type_subparts.
10019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10020 Alan Hayward <alan.hayward@arm.com>
10021 David Sherwood <david.sherwood@arm.com>
10023 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
10024 (vect_compile_time_alias): ...this new function. Do the calculation
10025 on poly_ints rather than trees.
10026 (vect_prune_runtime_alias_test_list): Update call accordingly.
10028 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10029 Alan Hayward <alan.hayward@arm.com>
10030 David Sherwood <david.sherwood@arm.com>
10032 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
10034 (vect_schedule_slp_instance): Likewise.
10036 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10037 Alan Hayward <alan.hayward@arm.com>
10038 David Sherwood <david.sherwood@arm.com>
10040 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
10041 constant and extern definitions for variable-length vectors.
10042 (vect_get_constant_vectors): Note that the number of units
10043 is known to be constant.
10045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10046 Alan Hayward <alan.hayward@arm.com>
10047 David Sherwood <david.sherwood@arm.com>
10049 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
10050 of units as polynomial. Choose between WIDE and NARROW based
10053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10054 Alan Hayward <alan.hayward@arm.com>
10055 David Sherwood <david.sherwood@arm.com>
10057 * tree-vect-stmts.c (simd_clone_subparts): New function.
10058 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
10060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10061 Alan Hayward <alan.hayward@arm.com>
10062 David Sherwood <david.sherwood@arm.com>
10064 * tree-vect-stmts.c (vectorizable_call): Treat the number of
10065 vectors as polynomial. Use build_index_vector for
10066 IFN_GOMP_SIMD_LANE.
10068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10069 Alan Hayward <alan.hayward@arm.com>
10070 David Sherwood <david.sherwood@arm.com>
10072 * tree-vect-stmts.c (get_load_store_type): Treat the number of
10073 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
10074 for variable-length vectors.
10075 (vectorizable_mask_load_store): Treat the number of units as
10076 polynomial, asserting that it is constant if the condition has
10077 already been enforced.
10078 (vectorizable_store, vectorizable_load): Likewise.
10080 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10081 Alan Hayward <alan.hayward@arm.com>
10082 David Sherwood <david.sherwood@arm.com>
10084 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
10085 of units as polynomial. Punt if we can't tell at compile time
10086 which vector contains the final result.
10088 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10089 Alan Hayward <alan.hayward@arm.com>
10090 David Sherwood <david.sherwood@arm.com>
10092 * tree-vect-loop.c (vectorizable_induction): Treat the number
10093 of units as polynomial. Punt on SLP inductions. Use an integer
10094 VEC_SERIES_EXPR for variable-length integer reductions. Use a
10095 cast of such a series for variable-length floating-point
10098 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10099 Alan Hayward <alan.hayward@arm.com>
10100 David Sherwood <david.sherwood@arm.com>
10102 * tree.h (build_index_vector): Declare.
10103 * tree.c (build_index_vector): New function.
10104 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
10105 of units as polynomial, forcibly converting it to a constant if
10106 vectorizable_reduction has already enforced the condition.
10107 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
10108 to create a {1,2,3,...} vector.
10109 (vectorizable_reduction): Treat the number of units as polynomial.
10110 Choose vectype_in based on the largest scalar element size rather
10111 than the smallest number of units. Enforce the restrictions
10114 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10115 Alan Hayward <alan.hayward@arm.com>
10116 David Sherwood <david.sherwood@arm.com>
10118 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
10119 number of units as polynomial.
10121 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10122 Alan Hayward <alan.hayward@arm.com>
10123 David Sherwood <david.sherwood@arm.com>
10125 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10126 * target.def (autovectorize_vector_sizes): Return the vector sizes
10127 by pointer, using vector_sizes rather than a bitmask.
10128 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10129 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10130 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10132 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10133 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10134 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10135 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10136 * omp-general.c (omp_max_vf): Likewise.
10137 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10138 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10139 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10140 * tree-vect-slp.c (vect_slp_bb): Likewise.
10141 * doc/tm.texi: Regenerate.
10142 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10144 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10145 the vector size as a poly_uint64 rather than an unsigned int.
10146 (current_vector_size): Change from an unsigned int to a poly_uint64.
10147 (get_vectype_for_scalar_type): Update accordingly.
10148 * tree.h (build_truth_vector_type): Take the size and number of
10149 units as a poly_uint64 rather than an unsigned int.
10150 (build_vector_type): Add a temporary overload that takes
10151 the number of units as a poly_uint64 rather than an unsigned int.
10152 * tree.c (make_vector_type): Likewise.
10153 (build_truth_vector_type): Take the number of units as a poly_uint64
10154 rather than an unsigned int.
10156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10157 Alan Hayward <alan.hayward@arm.com>
10158 David Sherwood <david.sherwood@arm.com>
10160 * target.def (get_mask_mode): Take the number of units and length
10161 as poly_uint64s rather than unsigned ints.
10162 * targhooks.h (default_get_mask_mode): Update accordingly.
10163 * targhooks.c (default_get_mask_mode): Likewise.
10164 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10165 * doc/tm.texi: Regenerate.
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 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10172 * omp-general.c (omp_max_vf): Likewise.
10173 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10174 (expand_omp_simd): Handle polynomial safelen.
10175 * omp-low.c (omplow_simd_context): Add a default constructor.
10176 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10177 (lower_rec_simd_input_clauses): Update accordingly.
10178 (lower_rec_input_clauses): Likewise.
10180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10181 Alan Hayward <alan.hayward@arm.com>
10182 David Sherwood <david.sherwood@arm.com>
10184 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10185 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10186 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10187 (vect_analyze_slp_cost): Likewise.
10188 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10189 (vect_model_load_cost): Likewise.
10191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10192 Alan Hayward <alan.hayward@arm.com>
10193 David Sherwood <david.sherwood@arm.com>
10195 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10196 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10197 from an unsigned int * to a poly_uint64_pod *.
10198 (calculate_unrolling_factor): New function.
10199 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10202 Alan Hayward <alan.hayward@arm.com>
10203 David Sherwood <david.sherwood@arm.com>
10205 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10206 from an unsigned int to a poly_uint64.
10207 (_loop_vec_info::slp_unrolling_factor): Likewise.
10208 (_loop_vec_info::vectorization_factor): Change from an int
10210 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10211 (vect_get_num_vectors): New function.
10212 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10213 (vect_get_num_copies): Use vect_get_num_vectors.
10214 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10215 to an unsigned int *.
10216 (vect_analyze_data_refs): Change min_vf from an int * to a
10218 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10219 than an unsigned HOST_WIDE_INT.
10220 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10221 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10222 to an unsigned int *.
10223 (vect_analyze_data_ref_dependences): Likewise.
10224 (vect_compute_data_ref_alignment): Handle polynomial vf.
10225 (vect_enhance_data_refs_alignment): Likewise.
10226 (vect_prune_runtime_alias_test_list): Likewise.
10227 (vect_shift_permute_load_chain): Likewise.
10228 (vect_supportable_dr_alignment): Likewise.
10229 (dependence_distance_ge_vf): Take the vectorization factor as a
10230 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10231 (vect_analyze_data_refs): Change min_vf from an int * to a
10233 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10234 vfm1 as a poly_uint64 rather than an int. Make the same change
10235 for the returned bound_scalar.
10236 (vect_gen_vector_loop_niters): Handle polynomial vf.
10237 (vect_do_peeling): Likewise. Update call to
10238 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10239 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10241 * tree-vect-loop.c (vect_determine_vectorization_factor)
10242 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10243 (vect_get_known_peeling_cost): Likewise.
10244 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10245 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10246 (vect_transform_loop): Likewise. Use the lowest possible VF when
10247 updating the upper bounds of the loop.
10248 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10249 rather than an int.
10250 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10251 polynomial unroll factors.
10252 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10253 (vect_make_slp_decision): Likewise.
10254 (vect_supported_load_permutation_p): Likewise, and polynomial
10256 (vect_analyze_slp_cost): Handle polynomial vf.
10257 (vect_slp_analyze_node_operations): Likewise.
10258 (vect_slp_analyze_bb_1): Likewise.
10259 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10260 than an unsigned HOST_WIDE_INT.
10261 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10262 (vectorizable_load): Handle polynomial vf.
10263 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10265 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10267 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10268 Alan Hayward <alan.hayward@arm.com>
10269 David Sherwood <david.sherwood@arm.com>
10271 * match.pd: Handle bit operations involving three constants
10272 and try to fold one pair.
10274 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10276 * tree-vect-loop-manip.c: Include gimple-fold.h.
10277 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10278 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10279 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10280 Add a path that uses a step of VF instead of 1, but disable it
10282 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10283 and niters_no_overflow parameters. Update calls to
10284 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10285 Create a new SSA name if the latter choses to use a ste other
10286 than zero, and return it via niters_vector_mult_vf_var.
10287 * tree-vect-loop.c (vect_transform_loop): Update calls to
10288 vect_do_peeling, vect_gen_vector_loop_niters and
10289 slpeel_make_loop_iterate_ntimes.
10290 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10291 (vect_gen_vector_loop_niters): Update declarations after above changes.
10293 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10295 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10296 128-bit round to integer instructions.
10297 (ceil<mode>2): Likewise.
10298 (btrunc<mode>2): Likewise.
10299 (round<mode>2): Likewise.
10301 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10303 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10304 unaligned VSX load/store on P8/P9.
10305 (expand_block_clear): Allow the use of unaligned VSX
10306 load/store on P8/P9.
10308 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10310 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10312 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10313 swap associated with both a load and a store.
10315 2018-01-02 Andrew Waterman <andrew@sifive.com>
10317 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10318 * config/riscv/riscv.md (clear_cache): Use it.
10320 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10322 * web.c: Remove out-of-date comment.
10324 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10326 * expr.c (fixup_args_size_notes): Check that any existing
10327 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10328 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10329 (emit_single_push_insn): ...here.
10331 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10333 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10334 (const_vector_encoded_nelts): New function.
10335 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10336 (const_vector_int_elt, const_vector_elt): Declare.
10337 * emit-rtl.c (const_vector_int_elt_1): New function.
10338 (const_vector_elt): Likewise.
10339 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10340 of CONST_VECTOR_ELT.
10342 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10344 * expr.c: Include rtx-vector-builder.h.
10345 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10346 directly on the tree encoding.
10347 (const_vector_from_tree): Likewise.
10348 * optabs.c: Include rtx-vector-builder.h.
10349 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10350 sequence of "u" values.
10351 * vec-perm-indices.c: Include rtx-vector-builder.h.
10352 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10353 directly on the vec_perm_indices encoding.
10355 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10357 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10358 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10359 * rtx-vector-builder.h: New file.
10360 * rtx-vector-builder.c: Likewise.
10361 * rtl.h (rtx_def::u2): Add a const_vector field.
10362 (CONST_VECTOR_NPATTERNS): New macro.
10363 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10364 (CONST_VECTOR_DUPLICATE_P): Likewise.
10365 (CONST_VECTOR_STEPPED_P): Likewise.
10366 (CONST_VECTOR_ENCODED_ELT): Likewise.
10367 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10368 (unwrap_const_vec_duplicate): Likewise.
10369 (const_vec_series_p): Check for a non-duplicated vector encoding.
10370 Say that the function only returns true for integer vectors.
10371 * emit-rtl.c: Include rtx-vector-builder.h.
10372 (gen_const_vec_duplicate_1): Delete.
10373 (gen_const_vector): Call gen_const_vec_duplicate instead of
10374 gen_const_vec_duplicate_1.
10375 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10376 (gen_const_vec_duplicate): Use rtx_vector_builder.
10377 (gen_const_vec_series): Likewise.
10378 (gen_rtx_CONST_VECTOR): Likewise.
10379 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10380 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10381 Build a new vector rather than modifying a CONST_VECTOR in-place.
10382 (handle_special_swappables): Update call accordingly.
10383 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10384 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10385 Build a new vector rather than modifying a CONST_VECTOR in-place.
10386 (handle_special_swappables): Update call accordingly.
10388 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10390 * simplify-rtx.c (simplify_const_binary_operation): Use
10391 CONST_VECTOR_ELT instead of XVECEXP.
10393 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10395 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10396 the selector elements to be different from the data elements
10397 if the selector is a VECTOR_CST.
10398 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10399 ssizetype for the selector.
10401 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10403 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10404 before testing each element individually.
10405 * tree-vect-generic.c (lower_vec_perm): Likewise.
10407 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10409 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10410 * selftest-run-tests.c (selftest::run_tests): Call it.
10411 * vector-builder.h (vector_builder::operator ==): New function.
10412 (vector_builder::operator !=): Likewise.
10413 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10414 (vec_perm_indices::all_from_input_p): New function.
10415 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10416 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10417 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10418 instead of reading the VECTOR_CST directly. Detect whether both
10419 vector inputs are the same before constructing the vec_perm_indices,
10420 and update the number of inputs argument accordingly. Use the
10421 utility functions added above. Only construct sel2 if we need to.
10423 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10425 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10426 the broadcast of the low byte.
10427 (expand_mult_highpart): Use an explicit encoding for the permutes.
10428 * optabs-query.c (can_mult_highpart_p): Likewise.
10429 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10430 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10431 (vectorizable_bswap): Likewise.
10432 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10433 explicit encoding for the power-of-2 permutes.
10434 (vect_permute_store_chain): Likewise.
10435 (vect_grouped_load_supported): Likewise.
10436 (vect_permute_load_chain): Likewise.
10438 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10440 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10441 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10442 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10443 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10444 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10445 (vect_gen_perm_mask_any): Likewise.
10447 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10449 * int-vector-builder.h: New file.
10450 * vec-perm-indices.h: Include int-vector-builder.h.
10451 (vec_perm_indices): Redefine as an int_vector_builder.
10452 (auto_vec_perm_indices): Delete.
10453 (vec_perm_builder): Redefine as a stand-alone class.
10454 (vec_perm_indices::vec_perm_indices): New function.
10455 (vec_perm_indices::clamp): Likewise.
10456 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10457 (vec_perm_indices::new_vector): New function.
10458 (vec_perm_indices::new_expanded_vector): Update for new
10459 vec_perm_indices class.
10460 (vec_perm_indices::rotate_inputs): New function.
10461 (vec_perm_indices::all_in_range_p): Operate directly on the
10462 encoded form, without computing elided elements.
10463 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10464 encoding. Update for new vec_perm_indices class.
10465 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10466 the given vec_perm_builder.
10467 (expand_vec_perm_var): Update vec_perm_builder constructor.
10468 (expand_mult_highpart): Use vec_perm_builder instead of
10469 auto_vec_perm_indices.
10470 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10471 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10472 or double series encoding as appropriate.
10473 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10474 vec_perm_indices instead of auto_vec_perm_indices.
10475 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10476 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10477 (vect_permute_store_chain): Likewise.
10478 (vect_grouped_load_supported): Likewise.
10479 (vect_permute_load_chain): Likewise.
10480 (vect_shift_permute_load_chain): Likewise.
10481 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10482 (vect_transform_slp_perm_load): Likewise.
10483 (vect_schedule_slp_instance): Likewise.
10484 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10485 (vectorizable_mask_load_store): Likewise.
10486 (vectorizable_bswap): Likewise.
10487 (vectorizable_store): Likewise.
10488 (vectorizable_load): Likewise.
10489 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10490 vec_perm_indices instead of auto_vec_perm_indices. Use
10491 tree_to_vec_perm_builder to read the vector from a tree.
10492 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10493 vec_perm_builder instead of a vec_perm_indices.
10494 (have_whole_vector_shift): Use vec_perm_builder and
10495 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10496 truncation to calc_vec_perm_mask_for_shift.
10497 (vect_create_epilog_for_reduction): Likewise.
10498 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10499 from auto_vec_perm_indices to vec_perm_indices.
10500 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10501 instead of changing individual elements.
10502 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10503 the vector in d.perm.
10504 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10505 from auto_vec_perm_indices to vec_perm_indices.
10506 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10507 instead of changing individual elements.
10508 (arm_vectorize_vec_perm_const): Use new_vector to install
10509 the vector in d.perm.
10510 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10511 Update vec_perm_builder constructor.
10512 (rs6000_expand_interleave): Likewise.
10513 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10514 (rs6000_expand_interleave): Likewise.
10516 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10518 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10519 to qimode could truncate the indices.
10520 * optabs.c (expand_vec_perm_var): Likewise.
10522 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10524 * Makefile.in (OBJS): Add vec-perm-indices.o.
10525 * vec-perm-indices.h: New file.
10526 * vec-perm-indices.c: Likewise.
10527 * target.h (vec_perm_indices): Replace with a forward class
10529 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10530 * optabs.h: Include vec-perm-indices.h.
10531 (expand_vec_perm): Delete.
10532 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10533 (expand_vec_perm_const): Declare.
10534 * target.def (vec_perm_const_ok): Replace with...
10535 (vec_perm_const): ...this new hook.
10536 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10537 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10538 * doc/tm.texi: Regenerate.
10539 * optabs.def (vec_perm_const): Delete.
10540 * doc/md.texi (vec_perm_const): Likewise.
10541 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10542 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10543 expand_vec_perm for constant permutation vectors. Assert that
10544 the mode of variable permutation vectors is the integer equivalent
10545 of the mode that is being permuted.
10546 * optabs-query.h (selector_fits_mode_p): Declare.
10547 * optabs-query.c: Include vec-perm-indices.h.
10548 (selector_fits_mode_p): New function.
10549 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10550 is defined, instead of checking whether the vec_perm_const_optab
10551 exists. Use targetm.vectorize.vec_perm_const instead of
10552 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10553 fit in the vector mode before using a variable permute.
10554 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10555 vec_perm_indices instead of an rtx.
10556 (expand_vec_perm): Replace with...
10557 (expand_vec_perm_const): ...this new function. Take the selector
10558 as a vec_perm_indices rather than an rtx. Also take the mode of
10559 the selector. Update call to shift_amt_for_vec_perm_mask.
10560 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10561 Use vec_perm_indices::new_expanded_vector to expand the original
10562 selector into bytes. Check whether the indices fit in the vector
10563 mode before using a variable permute.
10564 (expand_vec_perm_var): Make global.
10565 (expand_mult_highpart): Use expand_vec_perm_const.
10566 * fold-const.c: Includes vec-perm-indices.h.
10567 * tree-ssa-forwprop.c: Likewise.
10568 * tree-vect-data-refs.c: Likewise.
10569 * tree-vect-generic.c: Likewise.
10570 * tree-vect-loop.c: Likewise.
10571 * tree-vect-slp.c: Likewise.
10572 * tree-vect-stmts.c: Likewise.
10573 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10575 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10576 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10577 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10578 (aarch64_vectorize_vec_perm_const): ...this new function.
10579 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10580 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10581 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10582 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10583 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10584 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10585 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10587 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10588 check for NEON modes.
10589 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10590 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10591 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10592 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10594 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10595 the old VEC_PERM_CONST conditions.
10596 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10597 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10598 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10599 (ia64_vectorize_vec_perm_const_ok): Merge into...
10600 (ia64_vectorize_vec_perm_const): ...this new function.
10601 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10602 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10603 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10604 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10605 * config/mips/mips.c (mips_expand_vec_perm_const)
10606 (mips_vectorize_vec_perm_const_ok): Merge into...
10607 (mips_vectorize_vec_perm_const): ...this new function.
10608 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10609 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10610 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10611 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10612 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10613 (rs6000_expand_vec_perm_const): Delete.
10614 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10616 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10617 (altivec_expand_vec_perm_const_le): Take each operand individually.
10618 Operate on constant selectors rather than rtxes.
10619 (altivec_expand_vec_perm_const): Likewise. Update call to
10620 altivec_expand_vec_perm_const_le.
10621 (rs6000_expand_vec_perm_const): Delete.
10622 (rs6000_vectorize_vec_perm_const_ok): Delete.
10623 (rs6000_vectorize_vec_perm_const): New function.
10624 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10625 an element count and rtx array.
10626 (rs6000_expand_extract_even): Update call accordingly.
10627 (rs6000_expand_interleave): Likewise.
10628 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10629 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10630 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10631 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10632 (rs6000_expand_vec_perm_const): Delete.
10633 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10634 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10635 (altivec_expand_vec_perm_const_le): Take each operand individually.
10636 Operate on constant selectors rather than rtxes.
10637 (altivec_expand_vec_perm_const): Likewise. Update call to
10638 altivec_expand_vec_perm_const_le.
10639 (rs6000_expand_vec_perm_const): Delete.
10640 (rs6000_vectorize_vec_perm_const_ok): Delete.
10641 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10642 reference to the SPE evmerge intructions.
10643 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10644 an element count and rtx array.
10645 (rs6000_expand_extract_even): Update call accordingly.
10646 (rs6000_expand_interleave): Likewise.
10647 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10648 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10650 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10652 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10654 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10655 vector mode and that that mode matches the mode of the data
10657 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10658 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10659 directly using expand_vec_perm_1 when forcing selectors into
10661 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10663 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10665 * optabs-query.h (can_vec_perm_p): Delete.
10666 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10667 * optabs-query.c (can_vec_perm_p): Split into...
10668 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10669 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10670 particular selector is valid.
10671 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10672 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10673 (vect_grouped_load_supported): Likewise.
10674 (vect_shift_permute_load_chain): Likewise.
10675 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10676 (vect_transform_slp_perm_load): Likewise.
10677 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10678 (vectorizable_bswap): Likewise.
10679 (vect_gen_perm_mask_checked): Likewise.
10680 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10681 implementations of variable permutation vectors into account
10682 when deciding which selector to use.
10683 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10684 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10685 with a false third argument.
10686 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10687 to test whether the constant selector is valid and can_vec_perm_var_p
10688 to test whether a variable selector is valid.
10690 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10692 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10693 * optabs-query.c (can_vec_perm_p): Likewise.
10694 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10695 instead of vec_perm_indices.
10696 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10697 (vect_gen_perm_mask_checked): Likewise,
10698 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10699 (vect_gen_perm_mask_checked): Likewise,
10701 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10703 * optabs-query.h (qimode_for_vec_perm): Declare.
10704 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10705 (qimode_for_vec_perm): ...this new function.
10706 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10708 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10710 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10711 does not have a conditional at the top.
10713 2018-01-02 Richard Biener <rguenther@suse.de>
10715 * ipa-inline.c (big_speedup_p): Fix expression.
10717 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10720 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10723 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10727 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10728 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10729 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10730 cond_taken_branch_cost 3->4.
10732 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10734 PR tree-optimization/83581
10735 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10736 TODO_cleanup_cfg if any changes have been made.
10738 PR middle-end/83608
10739 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10740 convert_modes if target mode has the right side, but different mode
10743 PR middle-end/83609
10744 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10745 last argument when extracting from CONCAT. If either from_real or
10746 from_imag is NULL, use expansion through memory. If result is not
10747 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10748 the parts directly to inner mode, if even that fails, use expansion
10751 PR middle-end/83623
10752 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10753 check for bswap in mode rather than HImode and use that in expand_unop
10756 Copyright (C) 2018 Free Software Foundation, Inc.
10758 Copying and distribution of this file, with or without modification,
10759 are permitted in any medium without royalty provided the copyright
10760 notice and this notice are preserved.