| blob | 487d40170204ed08395b164a7a75d0d0fa1e60b5 |
1 2018-04-12 Jakub Jelinek <jakub@redhat.com>
3 PR target/85328
4 * config/i386/sse.md
5 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
7 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
8 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
9 and output is a reg, avoid creating invalid lowpart subreg, but
10 instead split into a 512-bit move. Don't split if not AVX512VL,
11 input is xmm16+ reg and output is a mem.
12 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
13 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
14 xmm16+ reg and output is a mem.
16 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
18 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
19 also for flag_dwarf2_cfi_asm.
21 2018-04-12 Jakub Jelinek <jakub@redhat.com>
23 PR rtl-optimization/85342
24 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
25 a bool scalar var inside of the loop instead. Don't try to update
26 recog_data.operand after failed apply_change_group.
28 2018-04-12 Tom de Vries <tom@codesourcery.com>
30 PR target/85296
31 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
32 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
33 array with flexible array member as array without given dimension.
34 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
35 argument for undefined param to true.
37 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
39 PR target/85321
40 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
41 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
42 from PowerPC section.
43 * config/rs6000/sysv4.opt (mcall-): Improve help text.
44 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
45 help text that is too long.
46 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
47 help text that is too long.
48 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
49 help text that is too long.
51 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
53 * config/alpha/alpha.md (stack_probe_internal): Rename
54 from "probe_stack". Update all callers.
56 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
58 PR rtl-optimization/84566
59 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
60 sched_macro_fuse_insns.
62 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
64 PR target/84301
65 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
66 (compute_block_dependences): ... from here.
68 2018-04-11 Jakub Jelinek <jakub@redhat.com>
70 PR tree-optimization/85331
71 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
72 from int to HOST_WIDE_INT.
74 2018-04-11 Martin Jambor <mjambor@suse.cz>
76 PR ipa/84149
77 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
78 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
79 not the same as the source val.
80 (cgraph_edge_brings_value_p): New parameter.
81 (gather_edges_for_value): Pass destination value to
82 cgraph_edge_brings_value_p.
83 (perhaps_add_new_callers): Likewise.
84 (get_info_about_necessary_edges): Likewise and exclude values brought
85 only by self-recursive edges.
86 (create_specialized_node): Redirect only clones of self-calling edges.
87 (+self_recursive_pass_through_p): New function.
88 (find_more_scalar_values_for_callers_subset): Use it.
89 (find_aggregate_values_for_callers_subset): Likewise.
90 (known_aggs_to_agg_replacement_list): Removed.
91 (decide_whether_version_node): Re-calculate known constants for all
92 remaining context clones.
94 2018-04-11 Richard Biener <rguenther@suse.de>
96 PR lto/85339
97 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
98 from early DWARF output.
99 (dwarf2out_early_finish): Output line info unconditionally into
100 early DWARF and add reference to it.
102 2018-04-11 Jakub Jelinek <jakub@redhat.com>
104 PR target/85281
105 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
106 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
107 other than V2DFmode using iptr mode attribute.
108 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
110 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
112 PR rtl-optimization/84659
113 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
115 2018-04-11 Jakub Jelinek <jakub@redhat.com>
117 PR debug/85302
118 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
119 SIZEP is NULL.
120 (output_loc_list): Pass address of a dummy size variable even in the
121 locview handling loop.
122 (index_location_lists): Add comment on why skip_loc_list_entry can't
123 call size_of_locs.
125 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
127 PR target/85261
128 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
129 into register.
131 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
133 PR target/85321
134 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
135 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
136 and -mstring-compare-inline-limit.
138 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
140 PR target/85287
141 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
142 for stack clash protection in a register whenever we need it to be in
143 a register.
145 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
147 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
148 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
150 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
152 PR target/85321
153 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
154 the help text.
155 (mlong-double-): Ditto.
156 * config/rs6000/sysv4.opt (msdata=): Ditto.
157 (mtls-size=): Ditto.
159 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
161 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
162 erroneous entries for
163 "vector int vec_ldl (int, long int *)", and
164 "vector unsigned int vec_ldl (int, unsigned long int *)".
165 Add comments and entries for
166 "vector bool char vec_ldl (int, bool char *)",
167 "vector bool short vec_ldl (int, bool short *)",
168 "vector bool int vec_ldl (int, bool int *)",
169 "vector bool long long vec_ldl (int, bool long long *)",
170 "vector pixel vec_ldl (int, pixel *)",
171 "vector long long vec_ldl (int, long long *)",
172 "vector unsigned long long vec_ldl (int, unsigned long long *)".
173 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
174 type tree bool_long_long_type_node and correct definition of
175 bool_V2DI_type_node to make reference to this new type tree.
176 (rs6000_mangle_type): Replace erroneous reference to
177 bool_long_type_node with bool_long_long_type_node.
178 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
179 comments to emphasize sign distinctions for char and int types and
180 replace RS6000_BTI_bool_long constant with
181 RS6000_BTI_bool_long_long constant. Also add comment to restrict
182 use of RS6000_BTI_pixel.
183 (bool_long_type_node): Remove this macro definition.
184 (bool_long_long_type_node): New macro definition
186 2018-04-10 Jakub Jelinek <jakub@redhat.com>
188 PR rtl-optimization/85300
189 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
190 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
191 simplify_unary_operation fails.
193 2018-04-10 Martin Liska <mliska@suse.cz>
195 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
196 cgraph_edge and ipa_ref.
198 2018-04-10 Jakub Jelinek <jakub@redhat.com>
200 PR target/85177
201 PR target/85255
202 * config/i386/sse.md
203 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
204 computation of the VEC_MERGE selector from mask.
205 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
206 Fix decoding of the VEC_MERGE selector into mask.
208 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
210 PR tree-optimization/85286
211 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
213 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
215 * final.c (final_1): Set insn_last_address as well as
216 insn_current_address.
218 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
220 PR target/85173
221 * explow.c (emit_stack_probe): Call validize_mem on memory location
222 before passing it to gen_probe_stack. Create address operand and
223 legitimize it for the probe_stack_address case.
225 2018-04-09 Jan Hubicka <jh@suse.cz>
227 PR lto/85078
228 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
229 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
230 * tree.c (free_lang_data_in_type): Fix handling of binfos;
231 walk basetypes.
232 (free_lang_data): Rebuild type inheritance graph.
234 2018-04-09 Martin Sebor <msebor@redhat.com>
236 * invoke.texi (-finline-small-functions): Mention other optimization
237 options.
238 (-findirect-inlining, -fpartial-inlining): Same.
239 (-finline-functions-called-once): Same.
240 (-freorder-blocks-and-partition): Same.
242 2018-04-09 Jan Hubicka <jh@suse.cz>
244 PR rtl/84058
245 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
246 jumps; choose last target that matches the criteria (i.e.
247 no partition changes for non-crossing jumps).
248 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
249 support for redirecting crossing jumps to non-crossing.
251 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
253 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
254 also for naked functions.
256 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
258 * config/arc/arc.md (add_shift): New pattern.
259 (add_shift2): Likewise.
260 (sub_shift): Likewise.
261 (sub_shift_cmp0_noout): Likewise.
262 (compare_si_ashiftsi): Likewise.
263 (xbfu_cmp0_noout): New combine pattern.
264 (xbfu_cmp0"): Likewise.
265 (movsi_set_cc_insn): Place the predicable variant first.
266 (commutative_binary_cmp0_noout): Remove clobber.
267 (commutative_binary_cmp0): New pattern.
268 (noncommutative_binary_cmp0): Likewise.
269 (noncommutative_binary_cmp0_noout): Likewise.
270 (noncommutative_binary_comparison_result_used): Removed.
271 (rsub_cmp0): New pattern.
272 (rsub_cmp0_noout): Likewise.
273 (extzvsi): Changed, keep only meaningful variants.
274 (SQH, SEZ): New iterators.
275 (SQH_postfix): New mode attribute.
276 (SEZ_prefix): New code attribute.
277 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
278 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
279 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
280 of numerical value.
281 (noncommutative_operator): Check the availability of barrel
282 shifter option.
284 2018-04-09 Richard Biener <rguenther@suse.de>
286 PR tree-optimization/85284
287 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
288 Only use the niter constraining form of simple_iv when the exit
289 is always executed.
291 2018-04-09 Tom de Vries <tom@codesourcery.com>
293 PR target/84041
294 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
295 (define_expand "*memory_barrier"): New define_expand.
296 (define_insn "memory_barrier"): New insn.
298 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
300 PR rtl-optimization/80463
301 PR rtl-optimization/83972
302 PR rtl-optimization/83480
304 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
305 correct producer for the insn.
306 (tidy_control_flow): Fixup seqnos in case of debug insns.
308 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
310 PR rtl-optimization/83913
312 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
313 different sched-times when merging exprs.
315 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
317 PR rtl-optimization/83962
319 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
320 tidy_fallthru_edge and tidy_control_flow.
322 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
324 PR rtl-optimization/83530
326 * sel-sched.c (force_next_insn): New global variable.
327 (remove_insn_for_debug): When force_next_insn is true, also leave only
328 next insn in the ready list.
329 (sel_sched_region): When the region wasn't scheduled, make another pass
330 over it with force_next_insn set to 1.
332 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
334 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
335 into tm_file.
336 * config/nds32/constants.md (unspec_volatile_element): Add enum values
337 for interrupt control.
338 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
339 functions for interrupt control.
340 * config/nds32/nds32-intrinsic.md: Likewise.
341 * config/nds32/nds32_intrinsic.h: Likewise.
342 * config/nds32/nds32.h (nds32_builtins): Likewise.
344 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
346 * config/nds32/nds32.c (nds32_init_machine_status,
347 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
348 strict_aligned_p field.
349 (nds32_expand_to_rtl_hook): New function.
350 (TARGET_EXPAND_TO_RTL_HOOK): Define.
351 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
353 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
354 Chung-Ju Wu <jasonwucj@gmail.com>
356 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
357 * config/nds32/nds32-n7.md: New file.
358 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
359 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
360 pipeline.
361 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
362 * config/nds32/nds32.md (pipeline_model): Add n7.
363 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
364 * config/nds32/pipelines.md: Include n7 settings.
366 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
367 Chung-Ju Wu <jasonwucj@gmail.com>
369 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
370 * config/nds32/nds32-e8.md: New file.
371 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
372 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
373 pipeline.
374 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
375 * config/nds32/nds32.md (pipeline_model): Add e8.
376 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
377 * config/nds32/pipelines.md: Include e8 settings.
379 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
380 Chung-Ju Wu <jasonwucj@gmail.com>
382 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
383 * config/nds32/nds32-n8.md: New file.
384 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
385 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
386 pipeline.
387 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
388 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
389 * config/nds32/nds32.md (pipeline_model): Add n8.
390 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
391 * config/nds32/pipelines.md: Include n8 settings.
393 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
394 Chung-Ju Wu <jasonwucj@gmail.com>
396 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
397 * config/nds32/nds32-n9-2r1w.md: New file.
398 * config/nds32/nds32-n9-3r2w.md: New file.
399 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
400 nds32_register_ports): New or modify for cpu n9.
401 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
402 pipeline.
403 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
404 * config/nds32/nds32-utils.c: New file.
405 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
406 TARGET_MUL_SLOW): Define.
407 * config/nds32/nds32.md (pipeline_model): New attribute.
408 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
409 New options that support cpu n9.
410 * config/nds32/pipelines.md: Include n9 settings.
411 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
413 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
415 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
416 information if necessary.
417 (output_cond_branch_compare_zero): Likewise.
418 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
419 (nds32_target_alignment): Refine for alignment.
420 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
421 (FUNCTION_BOUNDARY): Modify.
422 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
423 align case.
424 * config/nds32/nds32.opt (malways-align, malign-functions): New.
426 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
428 * config/nds32/constants.md (unspec_volatile_element): Add values for
429 TLB operation and data prefetch.
430 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
431 functions for TLB operation and data prefetch.
432 * config/nds32/nds32-intrinsic.md: Likewise.
433 * config/nds32/nds32_intrinsic.h: Likewise.
434 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
435 (nds32_print_operand): Likewise.
436 * config/nds32/nds32.h (nds32_builtins): Likewise.
438 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
439 Andrew Pinski <pinsika@gcc.gnu.org>
441 PR middle-end/82976
442 * match.pd: Use constant_boolean_node of correct type instead of
443 boolean_true_node or boolean_false_node for simplifying
444 pointer comparisons to zero.
446 2018-04-07 Jakub Jelinek <jakub@redhat.com>
448 PR tree-optimization/80021
449 * tree.c (verify_type_variant): Make error call in verify_variant_match
450 translatable and remove final full stop.
452 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
454 * config/nds32/constants.md (unspec_volatile_element): Add
455 UNSPEC_VOLATILE_EH_RETURN.
456 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
457 nds32_output_stack_pop): Support dwarf exception handling process.
458 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
459 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
460 exception handling process.
461 (nds32_compute_stack_frame): Likewise.
462 (nds32_return_addr_rtx): Likewise.
463 (nds32_initial_elimination_offset): Likewise.
464 (nds32_expand_prologue): Likewise.
465 (nds32_expand_epilogue): Likewise.
466 (nds32_dynamic_chain_address): New function.
467 * config/nds32/nds32.h (machine_function): Add fields for dwarf
468 exception handling.
469 (DYNAMIC_CHAIN_ADDRESS): Define.
470 (EH_RETURN_DATA_REGNO): Define.
471 (EH_RETURN_STACKADJ_RTX): Define.
472 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
473 patterns for dwarf exception handling.
475 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
477 * config/nds32/nds32.h: Clean up obsolete macros.
479 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
481 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
482 Add enum values for particular instructions.
483 * config/nds32/nds32-intrinsic.c: Implementation of expanding
484 particular intrinsic functions.
485 * config/nds32/nds32-intrinsic.md: Likewise.
486 * config/nds32/nds32_intrinsic.h: Likewise.
487 * config/nds32/nds32.h (nds32_builtins): Likewise.
488 * config/nds32/nds32.md (type): Add pbsad and pbsada.
489 (btst, ave): New patterns for particular instructions.
491 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
493 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
494 Add enum values for atomic load/store and memory sync.
495 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
496 and memory sync.
497 * config/nds32/nds32-intrinsic.md: Likewise.
498 * config/nds32/nds32_intrinsic.h: Likewise.
499 * config/nds32/nds32.h (nds32_builtins): Likewise.
501 2018-04-07 Jakub Jelinek <jakub@redhat.com>
503 PR tree-optimization/85257
504 * fold-const.c (native_encode_vector): If not all elts could fit
505 and off is -1, return 0 rather than offset.
506 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
507 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
508 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
509 adjust buffer in native_interpret_expr call.
511 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
513 * config/nds32/constants.md (unspec_volatile_element): Add cache
514 control enum values.
515 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
516 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
517 * config/nds32/nds32.c (nds32_cctl_names): New.
518 (nds32_print_operand): Handle cache control register names.
519 * config/nds32/nds32.h (nds32_builtins): New enum values.
520 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
521 macros.
522 * config/nds32/nds32.md (type): Add mmu.
523 * config/nds32/pipelines.md (simple_insn): Add mmu.
525 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
527 * config/nds32/nds32.md (type): Remove call.
528 * config/nds32/pipelines.md (simple_insn): Likewise.
530 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
532 * config/nds32/constants.md (unspec_volatile_element): Add
533 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
534 UNSPEC_VOLATILE_FMFCFG.
535 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
536 description for fmfcfg and fmfcsr.
537 (bdesc_1arg): Add fmtcsr.
538 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
539 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
540 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
541 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
542 unspec_fmfcfg): New patterns.
543 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
544 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
545 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
546 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
547 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
548 __nds32__fmfcfg): Define.
550 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
552 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
553 intrinsic register names.
554 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
555 intrinsic register enum values and macros.
557 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
559 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
560 for load/store addressing form.
561 (nds32_print_operand_address): Likewise.
563 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
565 PR target/85196
566 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
567 based on LABEL_REF. Remove useless assertion.
568 (pic_address_needs_scratch): Fix formatting.
569 (sparc_legitimize_pic_address): Minor tweaks.
570 (sparc_delegitimize_address): Adjust assertion accordingly.
571 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
572 into symbolic_operand.
573 (movsi_high_pic_label_ref): Likewise.
574 (movsi_lo_sum_pic_label_ref): Likewise.
575 (movdi_pic_label_ref): Likewise.
576 (movdi_high_pic_label_ref): Likewise.
577 (movdi_lo_sum_pic_label_ref): Likewise.
579 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
581 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
582 custom LIB_SPEC setup.
584 2018-04-06 Ruslan Bukin <br@bsdpad.com>
585 Kito Cheng <kito.cheng@gmail.com>
587 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
588 * config/riscv/freebsd.h: New.
590 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
592 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
593 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
594 file.
596 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
597 Kito Cheng <kito.cheng@gmail.com>
599 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
600 nds32_output_call, nds32_symbol_binds_local_p): New functions.
601 * config/nds32/nds32-protos.h (nds32_output_call,
602 nds32_output_return): Declare.
603 * config/nds32/nds32.md: Refine all the call and return patterns.
605 2018-04-06 Jakub Jelinek <jakub@redhat.com>
607 PR debug/85252
608 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
609 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
611 PR rtl-optimization/84872
612 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
613 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
614 EDGE_CROSSING edge.
616 2018-04-06 Tamar Christina <tamar.christina@arm.com>
618 * expr.c (copy_blkmode_to_reg): Revert 254862.
619 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
621 2018-04-06 Richard Biener <rguenther@suse.de>
623 PR middle-end/85244
624 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
625 after seeing a component reference with an adjacent field. Treat
626 refs to arrays at struct end of external decls similar to
627 refs to unconstrained commons.
629 2018-04-06 Jakub Jelinek <jakub@redhat.com>
631 PR sanitizer/85213
632 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
633 look through SAVE_EXPRs with non-side-effects argument. Adjust
634 recursive calls.
635 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
636 save_p here.
638 2018-04-06 Richard Biener <rguenther@suse.de>
640 PR middle-end/85180
641 * alias.c (find_base_term): New wrapper around find_base_term
642 unwinding CSELIB_VAL_PTR changes.
643 (find_base_term): Do not restore CSELIB_VAL_PTR during the
644 recursion.
646 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
648 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
649 instructions.
650 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
651 constant definitions.
652 ("nop"): lr 0,0 -> nopr r0
653 ("nop_lr0", "nop_lr1"): New insn definitions.
655 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
657 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
658 NDS32_V3PUSH_AVAILABLE_P macro.
660 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
661 Chung-Ju Wu <jasonwucj@gmail.com>
663 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
664 (nds32*-*-*): Add float and fpu_config into supported_defaults.
665 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
666 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
667 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
668 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
669 * config/nds32/constraints.md: New constraints and checking for hard
670 float configuration.
671 * config/nds32/iterators.md: New mode iterator and attribute for hard
672 float configuration.
673 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
674 patterns.
675 * config/nds32/nds32-fpu.md: New file.
676 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
677 deal with hard float code generation.
678 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
679 ARCH_V3S.
680 (abi_type, float_reg_number): New enum type.
681 * config/nds32/nds32-predicates.c: New predicates for hard float.
682 * config/nds32/nds32-protos.h: Declare functions for hard float.
683 * config/nds32/nds32.c: Implementation for hard float configuration.
684 * config/nds32/nds32.h: Definitions for hard float configuration.
685 * config/nds32/nds32.md: Include hard float machine description and
686 modify patterns for hard float configuration.
687 * config/nds32/nds32.opt: New options for hard float configuration.
688 * config/nds32/predicates.md: New predicates for hard float
689 configuration.
691 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
693 * common/config/nds32/nds32-common.c
694 (nds32_option_optimization_table): Enable -mreleax-hint by default.
696 2018-04-05 Jakub Jelinek <jakub@redhat.com>
698 PR middle-end/85195
699 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
700 CONSTRUCTOR_ELT (ctor, ...)->value.
702 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
704 PR target/85193
705 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
707 2018-04-05 Tom de Vries <tom@codesourcery.com>
709 PR target/85204
710 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
711 cond jump.
713 2018-04-05 Shiva Chen <shiva0217@gmail.com>
714 Kito Cheng <kito.cheng@gmail.com>
716 * config/nds32/constraints.md (U33): Fine-tune checking condition.
717 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
718 * config/nds32/nds32.h (nds32_16bit_address_type): Add
719 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
721 2018-04-05 Shiva Chen <shiva0217@gmail.com>
722 Kito Cheng <kito.cheng@gmail.com>
724 * config/nds32/constraints.md (Ufe): New memory constraint.
725 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
726 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
727 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
728 operands.
729 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
730 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
732 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
734 * config/nds32/nds32.md: Use optimize_size in the condition for
735 alu-shift instructions.
737 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
739 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
741 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
743 * config/nds32/nds32.md (negsi2): Refine pattern.
745 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
746 Chung-Ju Wu <jasonwucj@gmail.com>
748 * config/nds32/iterators.md (shift_rotate): New code iterator.
749 (shift): New code attribute.
750 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
751 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
752 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
753 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
754 bit-wise operations.
755 (andsi3, *andsi3): Ditto.
756 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
757 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
758 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
759 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
760 nds32_ior_operand, nds32_xor_operand): New predicates.
762 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
764 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
765 (addsi3, subsi3): ... this.
767 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
769 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
771 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
773 * config/nds32/nds32.md: Adjust indention.
775 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
777 * config/nds32/nds32.md (feature): New attribute.
779 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
781 * config/nds32/nds32.md (subtype): New attribute.
783 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
785 PR target/85203
786 * config/arm/arm-builtins.c (arm_expand_builtin): Change
787 expansion to perform a bitwise AND of the argument followed by a
788 boolean negation of the result.
790 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
792 PR rtl-optimization/84878
793 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
794 the basic block. Assert the use reference is not artificial and that
795 it has an associated insn.
797 2018-04-04 Michael Matz <matz@suse.de>
799 * builtins.c (compute_objsize): Pass correct operand
800 to array_at_struct_end_p.
802 2018-04-04 Richard Biener <rguenther@suse.de>
804 PR lto/85176
805 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
806 from contexts for DINFO_LEVEL_TERSE and below.
808 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
810 * config/nds32/nds32-doubleword.md (move_<mode>): Require
811 resiter_operand condition.
812 * config/nds32/nds32.md (*move<mode>): Ditto.
814 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
815 Monk Chiang <sh.chiang04@gmail.com>
817 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
819 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
821 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
823 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
824 Kito Cheng <kito.cheng@gmail.com>
826 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
827 nds32_cond_code_str, output_cond_branch,
828 output_cond_branch_compare_zero, nds32_expand_cbranch,
829 nds32_expand_cstore, nds32_expand_movcc,
830 nds32_output_cbranchsi4_equality_zero,
831 nds32_output_cbranchsi4_equality_reg,
832 nds32_output_cbranchsi4_equality_reg_or_const_int,
833 nds32_output_cbranchsi4_greater_less_zero: New functions.
834 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
835 nds32_expand_cstore, nds32_expand_movcc,
836 nds32_output_cbranchsi4_equality_zero,
837 nds32_output_cbranchsi4_equality_reg,
838 nds32_output_cbranchsi4_equality_reg_or_const_int,
839 nds32_output_cbranchsi4_greater_less_zero): Declare.
840 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
841 nds32_rimm11s_operand): New predicates.
842 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
843 * config/nds32/nds32.md: Rewrite all the branch and conditional move
844 patterns.
846 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
848 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
849 * config/nds32/nds32.md: Ditto.
850 * config/nds32/pipelines.md: Ditto.
852 2018-04-04 Richard Biener <rguenther@suse.de>
854 PR tree-optimization/85168
855 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
856 propagating abnormals.
858 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
860 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
862 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
863 Kito Cheng <kito.cheng@gmail.com>
865 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
866 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
867 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
868 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
869 * config/nds32/nds32.md (sibcall_internal): New.
870 (sibcall_register): Remove.
871 (sibcall_immediate): Remove.
872 (sibcall_value_internal): New.
873 (sibcall_value_register): Remove.
874 (sibcall_value_immediate): Remove.
875 * config/nds32/predicates.md (nds32_general_register_operand): New.
876 (nds32_call_address_operand): New.
878 2018-04-03 Jakub Jelinek <jakub@redhat.com>
880 PR rtl-optimization/85167
881 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
882 bb_defs if *split_p, instead preinitialize it to NULL.
884 PR tree-optimization/85156
885 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
886 evaluating the argument multiple times.
888 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
890 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
891 than vector.
892 (_mm_cvtpd_ps): Likewise.
893 (_mm_cvttpd_epi32): Likewise.
894 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
895 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
896 vector, pixel, and bool following altivec.h include.
898 2018-04-03 Martin Sebor <msebor@redhat.com>
900 * doc/extend.texi (Common Function Attributes): Clarify.
901 (const attribute): Likewise.
902 (pure attribute): Likewise.
904 2018-04-03 Jakub Jelinek <jakub@redhat.com>
906 PR target/85169
907 * config/i386/i386.c (ix86_expand_vector_set): Use
908 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
910 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
912 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
913 instructions when changing rounding bits to preserve precision bits
914 in the x87 control word.
916 2018-04-03 Martin Liska <mliska@suse.cz>
918 PR tree-optimization/82491
919 * rtl.h (strip_offset_and_add): Replace += suboffset with
920 poly_uint64 () + suboffset.
922 2018-03-29 Martin Liska <mliska@suse.cz>
923 Martin Jambor <mjambor@suse.cz>
925 PR ipa/84947
926 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
927 param_type is not an integral or pointer type.
929 2018-04-03 Richard Biener <rguenther@suse.de>
931 * sese.h (recompute_all_dominators): Remove.
933 2018-04-02 Martin Sebor <msebor@redhat.com>
935 * doc/invoke.texi (-Wrestrict): Fix typos.
937 2018-04-02 Jim Wilson <jimw@sifive.com>
939 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
940 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
941 (<optab>di3, <optab>si3_extend): Likewise.
942 (<optab>si3_mask, <optab>si3_mask_1): New.
943 (<optab>di3_mask, <optab>di3_mask_1): New.
944 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
945 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
946 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
948 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
950 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
951 example.
953 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
955 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
956 (nds32_canonicalize_comparison): New function.
958 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
959 Kito Cheng <kito.cheng@gmail.com>
960 Kuan-Lin Chen <kuanlinchentw@gmail.com>
962 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
963 * config/nds32/constants.md (unspec_volatile_element): Add
964 UNSPEC_VOLATILE_RELAX_GROUP.
965 * config/nds32/nds32-relax-opt.c: New file.
966 * config/nds32/nds32-predicates.c
967 (nds32_symbol_load_store_p): New function.
968 * config/nds32/nds32-protos.h
969 (nds32_symbol_load_store_p): Declare function.
970 (make_pass_nds32_relax_opt): Declare new rtl pass function.
971 * config/nds32/nds32.c
972 (nds32_register_pass): New function to register pass.
973 (nds32_register_passes): New function to register passes.
974 * config/nds32/nds32.md (relax_group): New pattern.
975 * config/nds32/nds32.opt (mrelax-hint): New option.
976 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
978 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
980 * config/nds32/t-nds32: Modify files dependency.
982 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
984 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
985 (PROFILE_HOOK): Define its implementation.
987 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
989 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
990 type and 32-bit size.
992 2018-04-01 Jakub Jelinek <jakub@redhat.com>
994 PR middle-end/85090
995 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
996 (V_128_256): New mode iterator.
997 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
998 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
999 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
1000 of V.
1001 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
1002 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
1004 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
1006 PR target/83315
1007 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
1008 NaN inputs correctly.
1010 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
1012 PR target/80546
1013 * config/rs6000/vsx.md (??r): New mode attribute.
1014 (*vsx_mov<mode>_64bit): Use it.
1015 (*vsx_mov<mode>_32bit): Likewise.
1017 2018-03-30 Martin Sebor <msebor@redhat.com>
1019 PR tree-optimization/84818
1020 * builtins.c (check_access): Use warning_n.
1022 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1024 PR target/83822
1025 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
1026 condition.
1027 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
1028 condition.
1030 2018-03-30 Julia Koval <julia.koval@intel.com>
1032 PR target/84413
1033 * x86-tune.def (movx, partial_reg_dependency): Enable for
1034 m_SKYLAKE_AVX512.
1036 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
1038 PR inline-asm/84985
1039 * lra-constraints.c (process_alt_operands): Move setting
1040 this_alternative_matches below.
1042 2018-03-29 Martin Liska <mliska@suse.cz>
1044 PR lto/84995.
1045 * doc/invoke.texi: Document how LTO works with debug info.
1046 Describe auto-load support of binutils. Mention 'x86-64'
1047 as valid option value of -march option.
1049 2018-03-29 Jakub Jelinek <jakub@redhat.com>
1051 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
1053 PR c/85094
1054 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
1055 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
1056 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
1057 checking.
1059 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1061 PR target/84912
1062 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
1063 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
1064 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
1065 for RS6000_BTM_POWERPC64.
1066 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
1067 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
1068 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
1069 definition.
1070 (DIVDE): Use it.
1071 (DIVDEU): Likewise.
1073 2018-03-28 Carl Love <cel@us.ibm.com>
1075 Revert
1076 2017-09-27 Carl Love <cel@us.ibm.com>
1078 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
1079 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
1080 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
1081 fctiw instruction.
1083 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
1085 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
1086 instead of __vector bool.
1087 (_mm_max_pu8): Likewise.
1088 (_mm_min_pi16): Likewise.
1090 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
1092 PR target/84912
1093 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
1094 (DIVWEUO): Likewise.
1095 (DIVDEO): Likewise.
1096 (DIVDEUO): Likewise.
1097 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
1098 DIVWEUO and DIVDEUO.
1099 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
1100 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
1101 (div_extend): Likewise.
1102 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
1103 builtin function.
1104 (__builtin_divweuo): Likewise.
1105 (__builtin_divdeo): Likewise.
1106 (__builtin_divdeuo): Likewise.
1108 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1110 PR target/85095
1111 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
1112 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
1114 PR tree-optimization/82004
1115 * gimple-match-head.c (optimize_pow_to_exp): New function.
1116 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
1117 Don't fold to exp if optimize_pow_to_exp is false.
1119 2018-03-28 Martin Liska <mliska@suse.cz>
1121 PR other/84819
1122 * calls.c (initialize_argument_information): Fix trailing space.
1123 * common.opt: Fix typo and provide better explanation for
1124 -fsanitize-coverage option.
1125 * config/i386/i386.opt: Fix typo.
1127 2018-03-28 Jakub Jelinek <jakub@redhat.com>
1128 Martin Liska <mliska@suse.cz>
1130 PR sanitizer/85081
1131 * gimplify.c (asan_poison_variable): Don't do the check for
1132 gimplify_omp_ctxp here.
1133 (gimplify_decl_expr): Do it here.
1134 (gimplify_target_expr): Likewise.
1136 2018-03-28 Martin Liska <mliska@suse.cz>
1138 PR target/84988
1139 * config/i386/i386.c (ix86_function_arg_advance): Do not call
1140 chkp_type_bounds_count if MPX is not enabled.
1142 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
1144 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
1146 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
1148 PR target/84914
1149 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
1150 function to create the function decl for complex long double
1151 multiply and divide for -mabi=ieeelongdouble.
1152 (init_float128_ieee): Call it.
1154 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
1156 PR target/85044
1157 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
1158 -fcf-protection=branch -mibt.
1159 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
1161 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1163 PR target/81863
1164 * config/arm/arm.c (arm_valid_symbolic_address): Handle arm_word_relocations
1166 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
1168 PR target/85056
1169 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
1170 extern array declarations.
1172 2018-03-27 Richard Biener <rguenther@suse.de>
1174 PR middle-end/84067
1175 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
1176 explicit single_use checks.
1178 2018-03-27 Richard Biener <rguenther@suse.de>
1180 PR tree-optimization/85082
1181 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
1182 Valueize the VUSE.
1184 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
1186 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
1187 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
1188 Turn on fasynchronous-unwind-tables and funwind-tables.
1190 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
1192 PR target/85073
1193 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
1194 (*bmi_blsr_<mode>_ccz): Ditto.
1196 2018-03-26 Tom de Vries <tom@codesourcery.com>
1198 PR tree-optimization/85063
1199 * omp-general.c (offloading_function_p): New function. Factor out
1200 of ...
1201 * omp-offload.c (pass_omp_target_link::gate): ... here.
1202 * omp-general.h (offloading_function_p): Declare.
1203 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
1204 with attribute omp declare target for offloading functions.
1206 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
1208 PR tree-optimization/84005
1209 * tree-data-ref.h (get_base_for_alignment): Declare.
1210 * tree-data-ref.c (get_base_for_alignment_1): New function.
1211 (get_base_for_alignment): Likewise.
1212 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
1213 get_base_for_alignment to find a suitable base object, instead
1214 of always using drb->base_address.
1216 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1218 PR inline-asm/85022
1219 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
1220 known size by default.
1222 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
1224 PR inline-asm/85030
1225 * lra-constraints.c (process_alt_operands): Don't match BLKmode
1226 and non BLKmode operands.
1228 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1230 PR target/85026
1231 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
1232 Clean up attributes.
1234 2018-03-23 Richard Biener <rguenther@suse.de>
1236 PR debug/85020
1237 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
1238 we are going to emit early debug for LTO.
1240 2018-03-23 Jakub Jelinek <jakub@redhat.com>
1242 PR inline-asm/85034
1243 * function.c (match_asm_constraints_1): Don't optimize if input
1244 doesn't satisfy general_operand predicate for output's mode.
1246 PR inline-asm/85022
1247 * alias.c (write_dependence_p): Don't require for x_canonicalized
1248 non-VOIDmode if x has VOIDmode.
1250 PR sanitizer/85029
1251 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
1252 just don't try to optimize it rather than assert it never happens.
1254 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1256 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
1257 macro expansions for definition of ST_INTERNAL_<mode> and
1258 LD_INTERNAL_<mode> builtins.
1259 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
1260 Remove prototype.
1261 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
1262 function.
1263 (altivec_expand_st_builtin): Likewise.
1264 (altivec_expand_builtin): Remove calls to deleted functions.
1265 (rs6000_address_for_altivec): Delete this function.
1266 * config/rs6000/vector.md: Remove expands for
1267 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
1269 2018-03-22 Sudakshina Das <sudi.das@arm.com>
1271 PR target/84826
1272 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
1273 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
1274 re-computing once computed.
1275 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
1276 (arm_init_machine_status): Initialize
1277 machine->static_chain_stack_bytes.
1279 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
1281 PR target/84760
1282 * doc/extend.texi: Add four new prototypes for vec_ld.
1283 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
1284 definitions for more logical presentation.
1285 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
1286 entries for V1TI variants of __builtin_altivec_ld builtin.
1287 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
1288 handling of V1TI variant of LVX icode pattern.
1289 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
1290 (rs6000_gimple_fold_builtin): Likewise.
1291 (altivec_init_builtins): Add code to define
1292 __builtin_altivec_lvx_v1ti function.
1294 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1296 PR inline-asm/84941
1297 * function.c (match_asm_constraints_1): Don't do the optimization
1298 if input isn't a REG, SUBREG, MEM or constant.
1300 2018-03-22 Tom de Vries <tom@codesourcery.com>
1302 PR tree-optimization/84956
1303 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
1304 bb_has_abnormal_pred.
1306 2018-03-22 Jakub Jelinek <jakub@redhat.com>
1308 PR sanitizer/85018
1309 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
1310 DECL_INITIAL (decl) to decl at the end.
1311 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
1312 adjust the comment.
1314 2018-03-21 Joseph Myers <joseph@codesourcery.com>
1316 * doc/extend.texi (__builtin_tgmath): Document when complex
1317 integer types are treated as _Complex _Float64.
1319 2018-03-21 Tom de Vries <tom@codesourcery.com>
1321 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
1323 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1325 PR tree-optimization/84960
1326 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
1327 if it is ENTRY block, move them into single succ of ENTRY in that case.
1329 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
1331 PR tree-optimization/84811
1332 * poly-int.h (poly_span_traits): Remove the T3 parameter and
1333 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
1334 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
1335 (known_subrange_p): Update accordingly. Cast each value involved
1336 in the size comparison, rather than casting the result of the
1337 subtraction.
1339 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1341 PR tree-optimization/84982
1342 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
1343 by flipping the least significant bit rather than all bits from
1344 bitpos to bitpos + bitsize - 1.
1346 2018-03-21 Nathan Sidwell <nathan@acm.org>
1348 * doc/extend.texi (Deprecated Features): Remove mention of
1349 long-deleted deprecations.
1351 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
1353 PR jit/84288
1354 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
1355 * configure: Regenerate.
1357 2018-03-21 Tom de Vries <tom@codesourcery.com>
1359 PR tree-optimization/83126
1360 * tree-parloops.c (num_phis): New function.
1361 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
1363 2018-03-21 Nathan Sidwell <nathan@acm.org>
1365 * doc/extend.texi (Deprecated Features): Update deprecated flags,
1366 mention anon-struct/union members and trailing attributes.
1368 2018-03-21 Bin Cheng <bin.cheng@arm.com>
1370 PR tree-optimization/84969
1371 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
1372 builtin memset partitions if they set different rhs values.
1374 2018-03-21 Jakub Jelinek <jakub@redhat.com>
1376 PR rtl-optimization/84989
1377 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
1378 VEC_DUPLICATE with scalar result mode.
1380 2018-03-21 Martin Liska <mliska@suse.cz>
1382 PR ipa/84963
1383 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
1384 not intended return statement.
1386 2018-03-21 Martin Liska <mliska@suse.cz>
1388 PR target/84988
1389 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
1390 (chkp_find_bound_slots_1): Limit number of iterations.
1392 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
1394 PR target/84838
1395 * Minor grammar fixes for x86 options.
1397 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1399 PR debug/84875
1400 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
1401 holding REG_CFA_RESTORE notes, instead turn them into a USE.
1403 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
1405 PR target/83789
1406 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
1407 (altivec_lvx_<mode>_1op): Likewise.
1408 (altivec_stvx_<mode>_2op): Likewise.
1409 (altivec_stvx_<mode>_1op): Likewise.
1410 (altivec_lvx_<VM2:mode>): New define_expand.
1411 (altivec_stvx_<VM2:mode>): Likewise.
1412 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
1413 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1414 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
1415 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
1416 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
1417 (rs6000_gen_lvx): Likewise.
1418 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
1419 (altivec_expand_stv_builtin): Likewise.
1420 (altivec_expand_builtin): Likewise.
1421 * config/rs6000/vector.md: Likewise.
1423 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1425 PR target/82518
1426 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
1427 BYTES_BIG_ENDIAN.
1429 2018-03-20 Richard Biener <rguenther@suse.de>
1431 PR target/84986
1432 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
1433 sign-conversions as zero, fall back to standard scalar_stmt
1434 cost for the rest.
1436 2018-03-20 Martin Liska <mliska@suse.cz>
1438 PR ipa/84825
1439 * predict.c (rebuild_frequencies): Handle case when we have
1440 PROFILE_ABSENT, but flag_guess_branch_prob is false.
1442 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1444 PR target/84990
1445 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
1446 flag_section_anchors.
1447 * varasm.c (use_blocks_for_decl_p): Remove hack for
1448 dw2_force_const_mem.
1450 PR target/84845
1451 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
1452 to ...
1453 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
1454 be created, use lowpart_subreg of operands[0] rather than operands[0]
1455 itself.
1456 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
1457 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
1458 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
1459 and n constraint instead of aarch64_shift_imm_di and Usd.
1460 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
1461 (*aarch64_<optab>_reg_minus<mode>3): ... this.
1463 2018-03-20 Sudakshina Das <sudi.das@arm.com>
1465 PR target/82989
1466 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
1467 to favor GPR over NEON registers.
1468 (<shift>di3_neon): Likewise.
1470 2018-03-20 Tom de Vries <tom@codesourcery.com>
1472 PR target/84952
1473 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
1474 (nvptx_process_pars): Emit bar.sync asap and alap.
1476 2018-03-20 Tom de Vries <tom@codesourcery.com>
1478 PR target/84954
1479 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
1480 seen_label if seen_label is already set.
1482 2018-03-20 Jakub Jelinek <jakub@redhat.com>
1484 PR target/84945
1485 * config/i386/i386.c (fold_builtin_cpu): For features above 31
1486 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
1487 Use 1U instead of 1. Formatting fixes.
1489 PR c/84953
1490 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
1491 instead of TREE_TYPE (s1) for the return value.
1493 2018-03-19 Jakub Jelinek <jakub@redhat.com>
1495 PR tree-optimization/84946
1496 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
1497 bitsize + bitsize in poly_uint64 rather than poly_int64.
1499 PR sanitizer/78651
1500 * dwarf2asm.c: Include fold-const.c.
1501 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
1502 of decl rather than decl itself.
1504 PR rtl-optimization/84643
1505 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
1507 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
1509 PR sanitizer/78651
1510 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
1511 calling assemble_variable.
1513 2018-03-19 Sudakshina Das <sudi.das@arm.com>
1515 PR target/81647
1516 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
1517 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
1519 2018-03-19 Jim Wilson <jimw@sifive.com>
1521 PR bootstrap/84856
1522 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
1523 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
1524 (riscv_first_stack_step): Likewise.
1525 (riscv_option_override): Use STACK_BOUNDARY instead of
1526 MIN_STACK_BOUNDARY.
1527 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
1528 MIN_STACK_BOUNDARY.
1529 (BIGGEST_ALIGNMENT): Set to 128.
1530 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
1531 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
1532 STACK_BOUNDARY.
1534 2018-03-19 Richard Biener <rguenther@suse.de>
1536 PR tree-optimization/84933
1537 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
1538 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
1540 2018-03-19 Richard Biener <rguenther@suse.de>
1542 PR tree-optimization/84859
1543 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
1544 (cond_if_else_store_replacement): Perform sinking operation on
1545 single-store BBs regardless of MAX_STORES_TO_SINK setting.
1546 Generalize what a BB with a single eligible store is.
1548 2018-03-19 Richard Biener <rguenther@suse.de>
1550 PR tree-optimization/84929
1551 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
1552 chrec_is_positive against non-chrec arg.
1554 2018-03-19 Tamar Christina <tamar.christina@arm.com>
1556 PR target/84711
1557 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
1559 2018-03-18 Martin Liska <mliska@suse.cz>
1561 PR rtl-optimization/84635
1562 * regrename.c (build_def_use): Use matches_mode only when
1563 matches >= 0.
1565 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
1567 PR tree-optimization/84913
1568 * tree-vect-loop.c (vectorizable_reduction): Don't try to
1569 vectorize chains of COND_EXPRs.
1571 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1573 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
1575 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1577 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
1579 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
1581 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
1583 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1584 Kito Cheng <kito.cheng@gmail.com>
1586 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
1587 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
1588 (nds32_adjust_reg_alloc_order): New function.
1589 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
1591 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
1593 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
1594 nds32_print_operand, nds32_print_operand_address): Use
1595 HOST_WIDE_INT_PRINT_DEC instead.
1597 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
1599 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
1601 2018-03-17 Jakub Jelinek <jakub@redhat.com>
1603 PR target/84902
1604 * config/i386/i386.c (initial_ix86_tune_features,
1605 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
1606 unsigned long long.
1607 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
1608 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
1609 rather than 1u << ix86_tune. Formatting fix.
1610 (ix86_option_override_internal): Change ix86_arch_mask from
1611 unsigned int to unsigned HOST_WIDE_INT, initialize to
1612 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
1613 (ix86_function_specific_restore): Likewise.
1615 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1617 PR target/84899
1618 * postreload.c (reload_combine_recognize_pattern): Perform
1619 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
1620 truncate_int_for_mode the result for the destination's mode.
1622 PR c/84909
1623 * hsa-gen.c (mem_type_for_type): Fix comment typo.
1624 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
1625 Likewise.
1626 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
1627 Likewise.
1629 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
1631 PR target/84876
1632 * lra-assigns.c (lra_split_hard_reg_for): Don't use
1633 regno_allocno_class_array and sorted_pseudos.
1634 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
1635 insns where regno is used.
1637 2018-03-16 Martin Liska <mliska@suse.cz>
1639 PR ipa/84833
1640 * multiple_target.c (create_dispatcher_calls): Redirect
1641 reference in the symbol table.
1643 2018-03-16 Martin Liska <mliska@suse.cz>
1645 PR ipa/84722
1646 * multiple_target.c (create_dispatcher_calls): Redirect also
1647 an alias.
1649 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1651 PR c++/79937
1652 PR c++/82410
1653 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
1654 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
1655 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
1657 2018-03-16 Julia Koval <julia.koval@intel.com>
1659 * doc/invoke.texi (Skylake Server): Add CLWB.
1660 Cannonlake): Remove CLWB.
1662 2018-03-16 Jakub Jelinek <jakub@redhat.com>
1664 PR tree-optimization/84841
1665 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
1666 1 << 3.
1667 (FLOAT_ONE_CONST_TYPE): Define.
1668 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
1669 (sort_by_operand_rank): Put entries with higher constant_type last
1670 rather than first to match comments.
1672 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
1674 * config/nios2/nios2.md (movsi_internal): Fix thinko in
1675 split predicate.
1677 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1679 PR c++/79085
1680 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
1681 check and use address of target always.
1683 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
1685 PR target/84574
1686 * config/i386/i386.c (indirect_thunk_needed): Update comments.
1687 (indirect_thunk_bnd_needed): Likewise.
1688 (indirect_thunks_used): Likewise.
1689 (indirect_thunks_bnd_used): Likewise.
1690 (indirect_return_needed): New.
1691 (indirect_return_bnd_needed): Likewise.
1692 (output_indirect_thunk_function): Add a bool argument for
1693 function return.
1694 (output_indirect_thunk_function): Don't generate alias for
1695 function return thunk.
1696 (ix86_code_end): Call output_indirect_thunk_function to generate
1697 function return thunks.
1698 (ix86_output_function_return): Set indirect_return_bnd_needed
1699 and indirect_return_needed instead of indirect_thunk_bnd_needed
1700 and indirect_thunk_needed.
1702 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
1704 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
1705 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
1706 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
1708 2018-03-15 David Malcolm <dmalcolm@redhat.com>
1709 Paul Hua <paul.hua.gm@gmail.com>
1711 PR c/84852
1712 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
1714 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
1716 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
1717 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
1718 resp. SFmode cases.
1720 2018-03-15 Tamar Christina <tamar.christina@arm.com>
1722 PR target/84711
1723 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
1724 instead of GET_MODE_SIZE when comparing Units.
1726 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
1728 PR target/68256
1729 * varasm.c (hash_section): Return an unchangeble hash value
1730 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
1731 Return !aarch64_can_use_per_function_literal_pools_p ().
1733 2018-03-15 Jakub Jelinek <jakub@redhat.com>
1735 PR target/84860
1736 * optabs.c (emit_conditional_move): Pass address of cmode's copy
1737 rather than address of cmode as last argument to prepare_cmp_insn.
1739 2018-03-15 Julia Koval <julia.koval@intel.com>
1741 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
1742 F_AVX512VNNI, F_AVX512BITALG): New.
1744 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
1746 PR target/83451
1747 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
1748 insn for floating-point loads and stores.
1750 2018-03-14 Carl Love <cel@us.ibm.com>
1752 * config/rs6000/rs6000-c.c: Add macro definitions for
1753 ALTIVEC_BUILTIN_VEC_PERMXOR.
1754 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
1755 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
1756 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
1757 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
1758 UNSPEC_VPERMXOR.
1759 * config/doc/extend.texi: Add prototypes for vec_permxor.
1761 2018-03-14 David Malcolm <dmalcolm@redhat.com>
1763 PR c/84852
1764 * diagnostic-show-locus.c (class layout_point): Convert m_line
1765 from int to linenum_type.
1766 (line_span::comparator): Use linenum "compare" function when
1767 comparing line numbers.
1768 (test_line_span): New function.
1769 (layout_range::contains_point): Convert param "row" from int to
1770 linenum_type.
1771 (layout_range::intersects_line_p): Likewise.
1772 (layout::will_show_line_p): Likewise.
1773 (layout::print_source_line): Likewise.
1774 (layout::should_print_annotation_line_p): Likewise.
1775 (layout::print_annotation_line): Likewise.
1776 (layout::print_leading_fixits): Likewise.
1777 (layout::annotation_line_showed_range_p): Likewise.
1778 (struct line_corrections): Likewise for field m_row.
1779 (line_corrections::line_corrections): Likewise for param "row".
1780 (layout::print_trailing_fixits): Likewise.
1781 (layout::get_state_at_point): Likewise.
1782 (layout::get_x_bound_for_row): Likewise.
1783 (layout::print_line): Likewise.
1784 (diagnostic_show_locus): Likewise for locals "last_line" and
1785 "row".
1786 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
1787 * input.c (selftest::test_linenum_comparisons): New function.
1788 (selftest::input_c_tests): Call it.
1789 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
1790 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
1791 * selftest.h (ASSERT_GT): New macro.
1792 (ASSERT_GT_AT): New macro.
1793 (ASSERT_LT): New macro.
1794 (ASSERT_LT_AT): New macro.
1796 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
1798 PR rtl-optimization/84780
1799 * combine.c (distribute_links): Don't make a link based on pc_rtx.
1801 2018-03-14 Martin Liska <mliska@suse.cz>
1803 * tree.c (record_node_allocation_statistics): Use
1804 get_stats_node_kind.
1805 (get_stats_node_kind): New function extracted from
1806 record_node_allocation_statistics.
1807 (free_node): Use get_stats_node_kind.
1809 2018-03-14 Richard Biener <rguenther@suse.de>
1811 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
1812 that the value-set of ANTIC_IN doesn't grow.
1814 Revert
1815 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
1816 member.
1817 (BB_VISITED_WITH_VISITED_SUCCS): New define.
1818 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
1820 2018-03-14 Julia Koval <julia.koval@intel.com>
1822 * config.gcc (icelake-client, icelake-server): New.
1823 (icelake): Remove.
1824 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
1825 (initial_ix86_arch_features): Ditto.
1826 (PTA_SKYLAKE): Add SGX.
1827 (PTA_ICELAKE): Remove.
1828 (PTA_ICELAKE_CLIENT): New.
1829 (PTA_ICELAKE_SERVER): New.
1830 (ix86_option_override_internal): Split up icelake on icelake client and
1831 icelake server.
1832 (get_builtin_code_for_version): Ditto.
1833 (fold_builtin_cpu): Ditto.
1834 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
1835 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
1836 * config/i386/i386.h (processor_type): Ditto.
1837 * doc/invoke.texi: Ditto.
1839 2018-03-14 Jakub Jelinek <jakub@redhat.com>
1841 PR sanitizer/83392
1842 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
1843 INTEGER_CST offset, add it together with bitpos / 8 and
1844 sign extend based on POINTER_SIZE.
1846 PR target/84844
1847 Revert
1848 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
1850 PR target/78090
1851 * config/i386/constraints.md (Yc): New register constraint.
1852 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
1853 Use Yc constraint for alternative 2 of operand 0. Remove
1854 preferred_for_speed attribute.
1856 2018-03-14 Richard Biener <rguenther@suse.de>
1858 PR tree-optimization/84830
1859 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
1860 with the old one to avoid oscillations.
1862 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
1864 PR target/83712
1865 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
1866 pseudos.
1867 (assign_by_spills): Return a flag of reload assignment failure.
1868 Do not process the reload assignment failures. Do not spill other
1869 reload pseudos if they has the same reg class. Update n if
1870 necessary.
1871 (lra_assign): Add a return arg. Set up from the result of
1872 assign_by_spills call.
1873 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
1874 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
1875 usage_insns if it is not NULL.
1876 (spill_hard_reg_in_range): New function.
1877 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
1878 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
1879 function prototypes.
1880 (lra_assign): Change prototype.
1881 * lra.c (lra): Add code to deal with fails by splitting hard reg
1882 live ranges.
1884 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
1886 * config/riscv/riscv.opt (mrelax): New option.
1887 * config/riscv/riscv.c (riscv_file_start): Emit ".option
1888 "norelax" when riscv_mrelax is disabled.
1889 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
1891 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
1893 PR target/84743
1894 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
1895 reassociation for int modes.
1897 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1899 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
1900 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
1901 for big-endian.
1902 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
1903 * config/aarch64/aarch64-sve.md
1904 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
1905 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
1906 (*extend<mode><Vwide>2): Rename to...
1907 (aarch64_sve_extend<mode><Vwide>2): ...this.
1908 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
1909 renaming the old pattern to...
1910 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
1911 unsigned packs.
1912 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
1913 define_expand, renaming the old pattern to...
1914 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
1915 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
1916 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
1917 account when deciding which SVE instruction the optab should use.
1918 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
1920 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1922 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
1923 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
1924 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
1925 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
1926 (tlsdesc_small_<mode>): Turn a define_expand and use
1927 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
1928 (tlsdesc_small_advsimd_<mode>): ...this.
1929 (tlsdesc_small_sve_<mode>): New pattern.
1931 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
1933 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
1934 (UNSPEC_UMUL_HIGHPART): New constants.
1935 (MUL_HIGHPART): New int iteraor.
1936 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
1937 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
1938 define_expand.
1939 (*<su>mul<mode>3_highpart): New define_insn.
1941 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
1943 PR lto/84805
1944 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
1945 incomplete types.
1947 2018-03-13 Martin Liska <mliska@suse.cz>
1949 PR ipa/84658.
1950 * (sem_item_optimizer::sem_item_optimizer): Initialize new
1951 vector.
1952 (sem_item_optimizer::~sem_item_optimizer): Release it.
1953 (sem_item_optimizer::merge_classes): Register variable aliases.
1954 (sem_item_optimizer::fixup_pt_set): New function.
1955 (sem_item_optimizer::fixup_points_to_sets): Likewise.
1956 * ipa-icf.h: Declare new variables and functions.
1958 2018-03-13 Jakub Jelinek <jakub@redhat.com>
1960 PR middle-end/84834
1961 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
1962 integer_pow2p@2 and test integer_pow2p in condition.
1963 (A < 0 ? C : 0): Similarly for @1.
1965 PR middle-end/84831
1966 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
1967 characters starting at p contain '\0' character, don't look beyond
1968 that.
1970 PR target/84827
1971 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
1972 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
1974 PR target/84828
1975 * reg-stack.c (change_stack): Change update_end var from int to
1976 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
1977 also call set_block_for_insn on the newly added insns and rescan.
1979 PR target/84786
1980 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
1981 on the last operand.
1983 PR c++/84704
1984 * tree.c (stabilize_reference_1): Return save_expr (e) for
1985 STATEMENT_LIST even if it doesn't have side-effects.
1987 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
1989 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
1991 2018-03-12 Renlin Li <renlin.li@arm.com>
1993 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
1994 aarch64_output_scalar_simd_mov_immediate.
1996 2018-03-12 Martin Sebor <msebor@redhat.com>
1998 PR tree-optimization/83456
1999 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
2000 for perfectly overlapping calls to memcpy.
2001 (gimple_fold_builtin_memory_chk): Same.
2002 (gimple_fold_builtin_strcpy): Handle no-warning.
2003 (gimple_fold_builtin_stxcpy_chk): Same.
2004 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
2006 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2008 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
2009 parameter. Use it for SFmode.
2010 (rs6000_function_arg_advance_1): Adjust.
2011 (rs6000_function_arg): Adjust.
2012 (rs6000_gimplify_va_arg): Pass false for that new parameter.
2014 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
2016 PR rtl-optimization/84169
2017 PR rtl-optimization/84780
2018 * combine.c (can_combine_p): Check for a 2-insn combination whether
2019 the destination register is used between the two insns, too.
2021 2018-03-12 Richard Biener <rguenther@suse.de>
2023 PR tree-optimization/84803
2024 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
2025 for refs DR analysis didn't process.
2027 2018-03-12 Richard Biener <rguenther@suse.de>
2029 PR tree-optimization/84777
2030 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
2031 force-vectorize loops ignore whether we are optimizing for size.
2033 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
2035 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
2036 (TARGET_MD_ASM_ADJUST): Define.
2038 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
2039 Kito Cheng <kito.cheng@gmail.com>
2040 Chung-Ju Wu <jasonwucj@gmail.com>
2042 * config/nds32/nds32.c (nds32_compute_stack_frame,
2043 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
2044 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
2045 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
2046 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
2047 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
2048 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
2049 * config/nds32/nds32.md (prologue, epilogue): Use macro
2050 NDS32_V3PUSH_AVAILABLE_P to do checking.
2052 2018-03-11 Jakub Jelinek <jakub@redhat.com>
2054 PR debug/58150
2055 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
2056 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
2057 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
2058 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
2059 addition of most attributes on !orig_type_die or the attribute not
2060 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
2062 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2063 Chung-Ju Wu <jasonwucj@gmail.com>
2065 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
2066 __NDS32_VH__ macro.
2067 * config/nds32/nds32.opt (mvh): New option.
2069 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2070 Chung-Ju Wu <jasonwucj@gmail.com>
2072 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
2073 function.
2074 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
2075 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
2076 definition.
2078 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
2079 Chung-Ju Wu <jasonwucj@gmail.com>
2081 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
2082 function.
2083 * config/nds32/nds32-multiple.md (strlensi): New pattern.
2084 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
2086 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
2087 Kito Cheng <kito.cheng@gmail.com>
2088 Chung-Ju Wu <jasonwucj@gmail.com>
2090 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
2091 UNSPEC_FFMISM and UNSPEC_FLMISM.
2092 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
2093 for ffb, ffmism and flmism.
2094 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
2095 (unspec_ffmism): Ditto.
2096 (unspec_flmism): Ditto.
2097 (nds32_expand_builtin_impl): Check if string extension is available.
2098 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
2099 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
2101 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
2103 Reverting patch:
2104 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2106 PR target/83712
2107 * lra-assigns.c (assign_by_spills): Return a flag of reload
2108 assignment failure. Do not process the reload assignment
2109 failures. Do not spill other reload pseudos if they has the same
2110 reg class.
2111 (lra_assign): Add a return arg. Set up from the result of
2112 assign_by_spills call.
2113 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2114 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2115 usage_insns if it is not NULL.
2116 (spill_hard_reg_in_range): New function.
2117 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2118 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2119 function prototypes.
2120 (lra_assign): Change prototype.
2121 * lra.c (lra): Add code to deal with fails by splitting hard reg
2122 live ranges.
2124 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
2126 PR target/84807
2127 * config/i386/i386.opt: Replace Enforcment with Enforcement.
2129 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
2131 PR debug/84620
2132 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
2133 (dw_val_node): Add val_symbolic_view.
2134 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
2135 (symview_upper_bound): New.
2136 (new_line_info_table): Initialize symviews_since_reset.
2137 (dwarf2out_source_line): Count symviews_since_reset and set
2138 symview_upper_bound.
2139 (dw_val_equal_p): Handle symview.
2140 (add_AT_symview): New.
2141 (print_dw_val): Handle symview.
2142 (attr_checksum, attr_checksum_ordered): Likewise.
2143 (same_dw_val_p, size_of_die): Likewise.
2144 (value_format, output_die): Likewise.
2145 (add_high_low_attributes): Use add_AT_symview for entry_view.
2146 (dwarf2out_finish): Reset symview_upper_bound, clear
2147 zero_view_p.
2149 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
2151 PR target/83969
2152 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
2153 Add strict argument and use it.
2154 (rs6000_split_multireg_move): Update for new strict argument.
2155 (mem_operand_gpr): Disallow all non-offsettable addresses.
2156 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
2158 2018-03-09 Jakub Jelinek <jakub@redhat.com>
2160 PR target/84772
2161 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
2162 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
2163 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
2165 PR c++/84767
2166 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
2167 decl, use remap_type if we want to use the type.
2169 2018-03-09 Martin Sebor <msebor@redhat.com>
2171 PR tree-optimization/84526
2172 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
2173 Remove dead code.
2174 (builtin_access::generic_overlap): Be prepared to handle non-array
2175 base objects.
2177 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
2179 PR rtl-optimization/84682
2180 * lra-constraints.c (process_address_1): Check is_address flag
2181 for address constraints.
2182 (process_alt_operands): Likewise.
2183 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
2184 preprocess_constraints.
2185 * recog.h (preprocess_constraints): Add oploc parameter.
2186 Adjust callers.
2187 * recog.c (preprocess_constraints): Test address_operand for
2188 CT_ADDRESS constraints.
2190 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
2192 PR target/83712
2193 * lra-assigns.c (assign_by_spills): Return a flag of reload
2194 assignment failure. Do not process the reload assignment
2195 failures. Do not spill other reload pseudos if they has the same
2196 reg class.
2197 (lra_assign): Add a return arg. Set up from the result of
2198 assign_by_spills call.
2199 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
2200 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
2201 usage_insns if it is not NULL.
2202 (spill_hard_reg_in_range): New function.
2203 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
2204 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
2205 function prototypes.
2206 (lra_assign): Change prototype.
2207 * lra.c (lra): Add code to deal with fails by splitting hard reg
2208 live ranges.
2210 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2212 PR target/83193
2213 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
2214 Accept complain bool parameter. Only emit errors if it is true.
2215 (arm_parse_cpu_option_name): Likewise.
2216 (arm_target_thumb_only): Adjust callers of the above.
2217 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
2218 prototype to take a default true bool parameter.
2219 (arm_parse_arch_option_name): Likewise.
2221 2018-03-09 David Malcolm <dmalcolm@redhat.com>
2222 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
2224 PR jit/64089
2225 PR jit/84288
2226 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
2227 * configure: Regenerate.
2228 * configure.ac ("linker --version-script option"): New.
2229 ("linker soname option"): New.
2231 2018-03-09 Richard Biener <rguenther@suse.de>
2233 PR tree-optimization/84775
2234 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
2235 immediate uses of predicate stmts and mark them modified.
2237 Revert
2238 PR tree-optimization/84178
2239 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2240 to caller.
2241 (version_loop_for_if_conversion): Delay update_ssa call.
2242 (tree_if_conversion): Delay update_ssa until after predicate
2243 insertion.
2245 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
2247 PR target/84763
2248 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
2249 when the function accesses prior frames.
2251 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2253 PR debug/84456
2254 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
2255 gen_llsym, otherwise call maybe_gen_llsym.
2257 PR inline-asm/84742
2258 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
2259 has ',' character inside of it.
2261 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2263 PR target/84748
2264 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
2265 as clobbering CC_REGNUM.
2267 2018-03-08 Richard Biener <rguenther@suse.de>
2269 PR middle-end/84552
2270 * tree-scalar-evolution.c: Include tree-into-ssa.h.
2271 (follow_copies_to_constant): Do not follow SSA names registered
2272 for update.
2274 2018-03-08 Richard Biener <rguenther@suse.de>
2276 PR tree-optimization/84178
2277 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
2278 to caller.
2279 (version_loop_for_if_conversion): Delay update_ssa call.
2280 (tree_if_conversion): Delay update_ssa until after predicate
2281 insertion.
2283 2018-03-08 David Malcolm <dmalcolm@redhat.com>
2285 PR tree-optimization/84178
2286 * tree-if-conv.c (release_bb_predicate): Remove the
2287 the assertion that the stmts have NULL use_ops.
2288 Discard the statements, asserting that they haven't
2289 yet been added to a BB.
2291 2018-03-08 Richard Biener <rguenther@suse.de>
2293 PR tree-optimization/84746
2294 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
2295 (phi_translate): Pass in destination ANTIC_OUT set.
2296 (phi_translate_1): Likewise. For a simplified result lookup
2297 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
2298 (phi_translate_set): Adjust.
2299 (do_pre_regular_insertion): Likewise.
2300 (do_pre_partial_partial_insertion): Likewise.
2302 2018-03-08 Martin Liska <mliska@suse.cz>
2304 PR gcov-profile/84735
2305 * doc/gcov.texi: Document usage of profile files.
2306 * gcov-io.h: Document changes in the format.
2308 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
2310 PR debug/84404
2311 PR debug/84408
2312 * dwarf2out.c (struct dw_line_info_table): Update comments for
2313 view == -1.
2314 (FORCE_RESET_NEXT_VIEW): New.
2315 (FORCE_RESETTING_VIEW_P): New.
2316 (RESETTING_VIEW_P): Check for -1 too.
2317 (ZERO_VIEW_P): Likewise.
2318 (new_line_info_table): Force-reset next view.
2319 (dwarf2out_begin_function): Likewise.
2320 (dwarf2out_source_line): Simplify zero_view_p initialization.
2321 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
2322 view directly. Omit view when omitting .loc at line 0.
2324 2018-03-08 Jakub Jelinek <jakub@redhat.com>
2326 PR tree-optimization/84740
2327 * tree-switch-conversion.c (process_switch): Call build_constructors
2328 only if info.phi_count is non-zero.
2330 PR tree-optimization/84739
2331 * tree-tailcall.c (find_tail_calls): Check call arguments against
2332 DECL_ARGUMENTS (current_function_decl) rather than
2333 DECL_ARGUMENTS (func) when checking for tail recursion.
2335 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2337 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
2338 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
2339 Volker Reichelt's entry and add entries for people that perform
2340 GCC fuzzy testing and report numerous bugs.
2342 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
2344 PR target/82411
2345 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
2346 readonly data in sdata, if that is disabled.
2347 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
2348 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
2349 -mreadonly-in-sdata option.
2351 2018-03-07 Martin Sebor <msebor@redhat.com>
2353 PR tree-optimization/84468
2354 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
2355 basic block when looking for nul assignment.
2357 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
2359 PR target/84277
2360 * except.h (output_function_exception_table): Adjust prototype.
2361 * except.c (output_function_exception_table): Remove FNNAME parameter
2362 and add SECTION parameter. Ouput one part of the table at a time.
2363 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
2364 the first part of the exception table and emit unwind directives.
2365 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
2366 (i386_pe_seh_cold_init): Likewise.
2367 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
2368 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
2369 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
2370 (ix86_output_call_insn): Emit a nop in one more case for SEH.
2371 * config/i386/winnt.c: Include except.h.
2372 (struct seh_frame_state): Add reg_offset, after_prologue and
2373 in_cold_section fields.
2374 (i386_pe_seh_end_prologue): Set seh->after_prologue.
2375 (i386_pe_seh_cold_init): New function.
2376 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
2377 to seh->in_cold_section.
2378 (seh_emit_push): Record the offset of the push.
2379 (seh_emit_save): Record the offet of the save.
2380 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
2381 Test seh->after_prologue to disregard the epilogue.
2382 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
2383 (i386_pe_end_cold_function): New function.
2385 2018-03-07 Jakub Jelinek <jakub@redhat.com>
2387 PR fortran/84565
2388 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
2389 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
2391 PR c++/84704
2392 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
2393 on tmp_var.
2394 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
2395 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
2397 PR middle-end/84723
2398 * multiple_target.c: Include tree-inline.h and intl.h.
2399 (expand_target_clones): Diagnose and fail if node->definition and
2400 !tree_versionable_function_p (node->decl).
2402 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
2404 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
2405 sprint_ul.
2406 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
2407 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
2408 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
2410 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2412 PR target/84710
2413 * combine.c (try_combine): Use reg_or_subregno instead of handling
2414 just paradoxical SUBREGs and REGs.
2416 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
2418 * config/arc/arc.c (arc_finalize_pic): Remove function.
2419 (arc_must_save_register): We use single base PIC register, remove
2420 checks to save/restore the PIC register.
2421 (arc_expand_prologue): Likewise.
2422 * config/arc/arc-protos.h (arc_set_default_type_attributes):
2423 Remove.
2424 (arc_verify_short): Likewise.
2425 (arc_attr_type): Likewise.
2426 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
2427 (walk_stores): Likewise.
2428 (arc_address_cost): Make it static.
2429 (arc_verify_short): Likewise.
2430 (branch_dest): Likewise.
2431 (arc_attr_type): Likewise.
2432 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
2433 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
2434 (arc_final_prescan_insn): Remove inserting the nops due to
2435 hardware hazards. It is done in reorg step.
2436 (insn_length_variant_t): Remove.
2437 (insn_length_parameters_t): Likewise.
2438 (arc_insn_length_parameters): Likewise.
2439 (arc_get_insn_variants): Likewise.
2440 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
2442 2018-03-06 Jakub Jelinek <jakub@redhat.com>
2444 PR inline-asm/84683
2445 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
2446 assertion failure.
2448 PR tree-optimization/84687
2449 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
2450 on new_node->decl.
2451 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
2453 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2455 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
2456 Rename to ppc_speculation_barrier.
2457 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
2458 __builtin_ppc_speculation_barrier.
2460 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2462 PR target/84700
2463 * combine.c (combine_simplify_rtx): Don't try to simplify if
2464 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
2465 are equal to x.
2467 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
2469 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
2470 to 32 bytes when compiling for POWER9.
2472 2018-03-05 Jakub Jelinek <jakub@redhat.com>
2474 PR target/84564
2475 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
2476 regparm >= 3 with no arg reg available also for calls with
2477 flag_force_indirect_call. Pass decl to ix86_function_regparm.
2479 PR target/84524
2480 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
2481 orig,vex.
2482 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
2484 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
2486 PR target/84264
2487 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
2489 2018-03-05 Richard Biener <rguenther@suse.de>
2491 PR tree-optimization/84486
2492 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
2493 When inserting a __builtin_assume_aligned call set the LHS
2494 SSA name alignment info accordingly.
2496 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
2498 PR tree-optimization/84114
2499 * config/aarch64/aarch64.c (aarch64_reassociation_width)
2500 Avoid reassociation of FLOAT_MODE addition.
2502 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
2504 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
2505 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
2506 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
2507 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
2508 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
2509 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
2510 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
2511 and -mwbnoinvd.
2512 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
2513 __builtin_ia32_wbinvd): New builtins.
2514 (SPECIAL_ARGS2): New.
2515 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
2516 (SPECIAL_ARGS2): New.
2517 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
2518 (ix86_valid_target_attribute_inner_p): Ditto.
2519 (ix86_init_mmx_sse_builtins): Add special_args2.
2520 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
2521 TARGET_WBNOINVD_P): New.
2522 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
2523 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
2524 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
2525 * config/i386/immintrin.h (_wbinvd): New intrinsic.
2526 * config/i386/pconfigintrin.h: New file.
2527 * config/i386/wbnoinvdintrin.h: Ditto.
2528 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
2529 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
2531 2018-03-05 Richard Biener <rguenther@suse.de>
2533 PR tree-optimization/84670
2534 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
2535 member.
2536 (BB_VISITED_WITH_VISITED_SUCCS): New define.
2537 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
2538 (compute_antic_aux): Only assert the number of values in ANTIC_IN
2539 doesn't grow if all successors (recursively) were visited at least
2540 once.
2542 2018-03-05 Richard Biener <rguenther@suse.de>
2544 PR tree-optimization/84650
2545 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
2546 if executed in the loop pipeline.
2548 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
2550 * doc/configfiles.texi (Configuration Files): Move info about
2551 conditionalizing $target-protos.h to...
2552 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
2553 differs from $target-protos.h.
2555 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
2556 Chung-Ju Wu <jasonwucj@gmail.com>
2558 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
2559 * config/nds32/nds32-multiple.md (setmemsi): Define.
2560 * config/nds32/nds32-memory-manipulation.c
2561 (nds32_gen_dup_4_byte_to_word_value): New.
2562 (emit_setmem_word_loop): New.
2563 (emit_setmem_byte_loop): New.
2564 (nds32_expand_setmem_loop): New.
2565 (nds32_expand_setmem_loop_v3m): New.
2566 (nds32_expand_setmem_unroll): New.
2567 (nds32_expand_setmem): New.
2569 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2570 Chung-Ju Wu <jasonwucj@gmail.com>
2572 * config/nds32/nds32-memory-manipulation.c
2573 (nds32_emit_load_store): New.
2574 (nds32_emit_post_inc_load_store): New.
2575 (nds32_emit_mem_move): New.
2576 (nds32_emit_mem_move_block): New.
2577 (nds32_expand_movmemsi_loop_unknown_size): New.
2578 (nds32_expand_movmemsi_loop_known_size): New.
2579 (nds32_expand_movmemsi_loop): New.
2580 (nds32_expand_movmemsi_unroll): New.
2581 (nds32_expand_movmemqi): Rename ...
2582 (nds32_expand_movmemsi): ... to this.
2583 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
2584 (movmemsi): ... to this.
2585 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
2586 (nds32_expand_movmemsi): ... to this.
2588 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2589 Monk Chiang <sh.chiang04@gmail.com>
2590 Chung-Ju Wu <jasonwucj@gmail.com>
2592 * config/nds32/nds32-protos.h
2593 (nds32_expand_load_multiple): New arguments.
2594 (nds32_expand_store_multiple): Ditto.
2595 (nds32_valid_multiple_load_store): Rename ...
2596 (nds32_valid_multiple_load_store_p): ... to this.
2597 * config/nds32/nds32-memory-manipulation.c
2598 (nds32_expand_load_multiple): Refine implementation.
2599 (nds32_expand_store_multiple): Ditto.
2600 * config/nds32/nds32-multiple.md
2601 (load_multiple): Update nds32_expand_load_multiple interface.
2602 (store_multiple): Update nds32_expand_store_multiple interface.
2603 * config/nds32/nds32-predicates.c
2604 (nds32_valid_multiple_load_store): Rename ...
2605 (nds32_valid_multiple_load_store_p): ... to this and refine
2606 implementation.
2607 * config/nds32/predicates.md
2608 (nds32_load_multiple_and_update_address_operation): New predicate.
2609 (nds32_store_multiple_and_update_address_operation): New predicate.
2611 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
2612 Chung-Ju Wu <jasonwucj@gmail.com>
2614 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
2615 (combo): New attribute.
2616 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
2618 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
2620 * config/nds32/nds32.opt: Change -mcmodel= default value.
2622 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
2623 Monk Chiang <sh.chiang04@gmail.com>
2624 Chung-Ju Wu <jasonwucj@gmail.com>
2626 * config/nds32/constants.md (unspec_element): New enum.
2627 * config/nds32/constraints.md (Umw): New constraint.
2628 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
2629 * config/nds32/nds32-intrinsic.md: Likewise.
2630 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
2631 (nds32_valid_smw_lwm_base_p): New.
2632 (nds32_output_smw_single_word): New.
2633 (nds32_output_lmw_single_word): New.
2634 (nds32_expand_unaligned_load): New.
2635 (nds32_expand_unaligned_store): New.
2636 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
2637 (nds32_output_smw_single_word): Declare.
2638 (nds32_output_lmw_single_word): Declare.
2639 (nds32_expand_unaligned_load): Declare.
2640 (nds32_expand_unaligned_store): Declare.
2641 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
2642 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
2643 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
2644 NDS32_BUILTIN_UASTORE_DW.
2645 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
2646 predicate.
2648 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
2649 Kito Cheng <kito.cheng@gmail.com>
2650 Chung-Ju Wu <jasonwucj@gmail.com>
2652 * config/nds32/nds32-intrinsic.c
2653 (nds32_expand_builtin_null_ftype_reg): Delete.
2654 (nds32_expand_builtin_reg_ftype_imm): Ditto.
2655 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
2656 (nds32_read_argument): New.
2657 (nds32_legitimize_target): Ditto.
2658 (nds32_legitimize_argument): Ditto.
2659 (nds32_check_constant_argument): Ditto.
2660 (nds32_expand_unop_builtin): Ditto.
2661 (nds32_expand_unopimm_builtin): Ditto.
2662 (nds32_expand_binop_builtin): Ditto.
2663 (nds32_builtin_decl_impl): Ditto.
2664 (builtin_description): Ditto.
2665 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
2666 (nds32_init_builtins_impl): Ditto.
2667 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
2668 (nds32_builtin_decl): New.
2669 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
2670 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
2672 2018-03-02 Jeff Law <law@redhat.com>
2674 * reorg.c (stop_search_p): Handle DEBUG_INSN.
2675 (redundant_insn, fill_simple_delay_slots): Likewise.
2676 (fill_slots_from_thread): Likewise.
2677 * resource.c (mark_referenced_resources): Likewise.
2678 (mark_set_resources, find_dead_or_set_registers): Likewise.
2680 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2682 * substring-locations.h (format_warning_va): Formatting fix for
2683 ATTRIBUTE_GCC_DIAG.
2684 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
2685 argument.
2686 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
2687 * substring-locations.c: Include intl.h.
2688 (format_warning_va): Turned into small wrapper around
2689 format_warning_n_va, renamed to ...
2690 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
2691 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
2692 use ngettext.
2693 (format_warning_at_substring_n): New function.
2694 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
2695 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
2696 format_warning_at_substring with just a shorter name instead of
2697 const function pointer.
2698 (fmtwarn_n): New function.
2699 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
2700 appropriate, get rid of all the fmtstr temporaries, move conditionals
2701 with G_() wrapped string literals directly into fmtwarn arguments,
2702 cast dir.len to (int), formatting fixes.
2704 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
2706 * doc/invoke.texi: Remove "Cilk Plus" references.
2708 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2709 Richard Biener <rguenther@suse.de>
2711 PR ipa/84628
2712 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
2713 for error or warning attributes if CALL_FROM_THUNK_P is set.
2714 Formatting fixes.
2716 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2718 PR target/56540
2719 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
2720 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
2722 PR target/56540
2723 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
2724 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
2726 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
2727 instead of -1U in last predictors element's probability member.
2729 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
2731 PR ipa/83983
2732 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
2733 arguments if they are comparable.
2735 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2737 PR tree-optimization/84634
2738 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
2739 masks and masked_loop_p with a single loop_masks, making sure it's
2740 null for bb vectorization.
2742 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
2744 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
2745 (vect_analyze_data_ref_access): Use loop->safe_len rather than
2746 loop->force_vectorize to check whether there is no alias.
2748 2018-03-02 Jakub Jelinek <jakub@redhat.com>
2750 PR target/84614
2751 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
2752 prototypes.
2753 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
2754 comments.
2755 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
2756 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
2757 instead of a loop around prev_real_insn.
2758 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
2759 prev_real_insn.
2761 PR inline-asm/84625
2762 * config/i386/i386.c (ix86_print_operand): Use conditional
2763 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
2764 zero vector.
2766 2018-03-02 Richard Biener <rguenther@suse.de>
2768 PR tree-optimization/84427
2769 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
2770 (bitmap_set_subtract_values): Rewrite to handle multiple
2771 exprs per value.
2772 (clean): Likewise.
2773 (prune_clobbered_mems): Likewise.
2774 (phi_translate): Take edge instead of pred/phiblock.
2775 (phi_translate_1): Likewise.
2776 (phi_translate_set): Likewise. Insert all translated
2777 exprs for a value into the set, keeping possibly multiple
2778 expressions per value.
2779 (compute_antic_aux): Adjust for phi_translate changes.
2780 When intersecting union the expressions and prune those
2781 not in the final value set, keeping possibly multiple
2782 expressions per value. Do not use value-insertion
2783 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
2784 all expressions. Add verification that the value-sets
2785 only shrink during iteration.
2786 (compute_partial_antic_aux): Adjust for the phi_translate changes.
2787 (do_pre_regular_insertion): Likewise.
2788 (do_pre_partial_partial_insertion): Likewise.
2790 2018-03-02 Richard Biener <rguenther@suse.de>
2792 PR target/82005
2793 * config/darwin.c (saved_debug_info_level): New static global.
2794 (darwin_asm_lto_start): Disable debug info generation for LTO out.
2795 (darwin_asm_lto_end): Restore debug info generation settings.
2797 2018-03-01 Martin Liska <mliska@suse.cz>
2799 PR sanitizer/82484
2800 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
2801 volatile arguments.
2803 2018-03-01 Richard Biener <rguenther@suse.de>
2805 PR debug/84645
2806 * dwarf2out.c (gen_variable_die): Properly handle late VLA
2807 type annotation with LTO when debug was disabled at compile-time.
2809 2018-03-01 Matthew Fortune <mfortune@gmail.com>
2811 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
2812 XINT with INTVAL.
2813 (mips_final_postscan_insn): Likewise.
2815 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
2817 PR rtl-optimization/84528
2818 * alias.c (init_alias_target): Add commentary.
2819 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
2820 a unique base value if the frame pointer is not eliminated
2821 to the stack pointer.
2823 2018-03-01 Tom de Vries <tom@codesourcery.com>
2825 PR rtl-optimization/83327
2826 * lra-int.h (hard_regs_spilled_into): Declare.
2827 * lra.c (hard_regs_spilled_into): Define.
2828 (init_reg_info): Init hard_regs_spilled_into.
2829 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
2830 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
2831 (process_bb_lives): Handle hard_regs_spilled_into.
2832 (lra_create_live_ranges_1): Before doing liveness propagation, clear
2833 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
2835 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
2837 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
2838 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
2839 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
2840 * config/rs6000/aix72.h: New file.
2842 2018-02-28 Jakub Jelinek <jakub@redhat.com>
2844 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
2845 instead of warning_at with conditional singular and plural messages
2846 where possible.
2848 PR target/52991
2849 * stor-layout.c (update_alignment_for_field): For
2850 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
2851 && !DECL_PACKED (field), do the alignment update, just use
2852 only desired_align instead of MAX (type_align, desired_align)
2853 as the alignment.
2854 (place_field): Don't do known_align < desired_align handling
2855 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
2856 is non-NULL, instead do it after rli->prev_field handling and
2857 only if not within a bitfield word. For DECL_PACKED (field)
2858 use type_align of BITS_PER_UNIT.
2860 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
2862 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
2863 superfluous parentheses and trailing spaces.
2865 2018-02-28 Richard Biener <rguenther@suse.de>
2867 PR tree-optimization/84584
2868 * graphite-scop-detection.c (scop_detection::add_scop): Discard
2869 SCoPs with fake exit edge.
2871 2018-02-28 Martin Liska <mliska@suse.cz>
2873 PR testsuite/84597
2874 * timevar.c (timer::print): Fix format to properly print 100%
2875 values.
2877 2018-02-28 Richard Biener <rguenther@suse.de>
2879 PR middle-end/84607
2880 * genmatch.c (capture_info::walk_match): Do not mark
2881 captured expressions without operands as expr_p given
2882 they act more like predicates and should be subject to
2883 "lost tail" side-effect preserving.
2885 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
2887 PR rtl-optimization/81611
2888 * auto-inc-dec.c (attempt_change): Move dead note from
2889 mem_insn if it's the next use of regno
2890 (find_address): Take address use of reg holding
2891 non-incremented value. Add parm to limit search to the named
2892 reg only.
2893 (merge_in_block): Attempt to use a mem insn that is the next
2894 use of the original regno.
2896 2018-02-27 Martin Sebor <msebor@redhat.com>
2898 PR c++/83871
2899 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
2900 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
2902 2018-02-27 Martin Sebor <msebor@redhat.com>
2904 PR translation/84207
2905 * diagnostic-core.h (warning_n, error_n, inform_n): Change
2906 n argument to unsigned HOST_WIDE_INT.
2907 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
2908 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
2909 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
2910 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
2912 2018-02-27 Richard Biener <rguenther@suse.de>
2914 PR tree-optimization/84512
2915 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
2916 Do not use the estimate returned from record_stmt_cost for
2917 the scalar iteration cost but sum properly using add_stmt_cost.
2919 2018-02-27 Richard Biener <rguenther@suse.de>
2921 PR tree-optimization/84466
2922 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
2923 Adjust last change to less strictly validate use operands.
2925 2018-02-27 Martin Liska <mliska@suse.cz>
2927 PR gcov-profile/84548
2928 * gcov.c (process_file): Allow partial overlap and consider it
2929 also as group functions.
2930 (output_lines): Properly calculate range of lines for a group.
2932 2018-02-27 Martin Liska <mliska@suse.cz>
2934 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
2935 'ggc' suffixes. Change first column width.
2936 (timer::print): Fix formatting of the column.
2938 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
2940 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
2941 preserve inline entry blocks for the sake of debug inline
2942 entry point markers alone.
2943 (remove_unused_locals): Suggest in comments a better place to
2944 force the preservation of inline entry blocks that are
2945 otherwise unused, but do not preserve them.
2947 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2949 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
2951 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
2953 PR target/84039
2954 * config/i386/constraints.md (Bs): Replace
2955 ix86_indirect_branch_register with
2956 TARGET_INDIRECT_BRANCH_REGISTER.
2957 (Bw): Likewise.
2958 * config/i386/i386.md (indirect_jump): Likewise.
2959 (tablejump): Likewise.
2960 (*sibcall_memory): Likewise.
2961 (*sibcall_value_memory): Likewise.
2962 Peepholes of indirect call and jump via memory: Likewise.
2963 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
2964 (*sibcall_value_GOT_32): Likewise.
2965 * config/i386/predicates.md (indirect_branch_operand): Likewise.
2966 (GOT_memory_operand): Likewise.
2967 (call_insn_operand): Likewise.
2968 (sibcall_insn_operand): Likewise.
2969 (GOT32_symbol_operand): Likewise.
2970 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
2972 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2974 PR rtl-optimization/83496
2975 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
2976 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
2977 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
2978 redundant insn, if any.
2979 (relax_delay_slots): Likewise.
2980 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
2982 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
2984 PR tree-optimization/83965
2985 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
2986 that grouped statements are part of a reduction chain. Return
2987 true if the statement is not marked as a reduction itself but
2988 is part of a group.
2989 (vect_recog_dot_prod_pattern): Don't check whether the statement
2990 is part of a group here.
2991 (vect_recog_sad_pattern): Likewise.
2992 (vect_recog_widen_sum_pattern): Likewise.
2994 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
2996 PR debug/84545
2997 * final.c (rest_of_clean_state): Also look for calls inside sequences.
2999 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
3001 PR target/84530
3002 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
3003 the bool argument.
3004 (ix86_output_indirect_function_return): New prototype.
3005 (ix86_split_simple_return_pop_internal): Likewise.
3006 * config/i386/i386.c (indirect_return_via_cx): New.
3007 (indirect_return_via_cx_bnd): Likewise.
3008 (indirect_thunk_name): Handle return va CX_REG.
3009 (output_indirect_thunk_function): Create alias for
3010 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
3011 (ix86_output_indirect_jmp): Remove the bool argument.
3012 (ix86_output_indirect_function_return): New function.
3013 (ix86_split_simple_return_pop_internal): Likewise.
3014 * config/i386/i386.md (*indirect_jump): Don't pass false
3015 to ix86_output_indirect_jmp.
3016 (*tablejump_1): Likewise.
3017 (simple_return_pop_internal): Change it to define_insn_and_split.
3018 Call ix86_split_simple_return_pop_internal to split it for
3019 -mfunction-return=.
3020 (simple_return_indirect_internal): Call
3021 ix86_output_indirect_function_return instead of
3022 ix86_output_indirect_jmp.
3024 2018-02-26 Jakub Jelinek <jakub@redhat.com>
3026 PR bootstrap/84405
3027 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
3028 memset and value initialization afterwards.
3030 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
3032 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
3034 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
3036 PR target/84521
3037 * common/config/aarch64/aarch64-common.c
3038 (aarch_option_optimization_table[]): Switch
3039 off fomit-frame-pointer
3041 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3042 Chung-Ju Wu <jasonwucj@gmail.com>
3044 * config/nds32/nds32-multiple.md (load_multiple): Disallow
3045 volatile memory.
3046 (store_multiple): Ditto.
3048 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
3050 * config.gcc: Add --with-cpu support for nds32 target.
3051 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
3052 * config/nds32/nds32.opt: Add -mcpu= option.
3054 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
3056 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
3057 isel=yes): Warn for these deprecated options.
3059 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
3061 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
3062 ISA_2_5_MASKS_EMBEDDED.
3064 2018-02-23 Jakub Jelinek <jakub@redhat.com>
3066 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
3067 p->max as pointers rather than using iterative_hash_expr.
3069 2018-02-23 Carl Love <cel@us.ibm.com>
3071 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
3072 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
3073 BU_P8V_OVERLOAD_2.
3074 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
3075 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
3076 P8V_BUILTIN_VEC_VUNSIGNED2.
3078 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
3080 PR target/81572
3081 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
3082 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
3083 LRA_UNKNOWN_ALT.
3084 * lra-constraints.c (curr_insn_transform): Set up
3085 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
3086 LRA_UNKNOWN_ALT.
3087 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
3088 * lra-eliminations.c (spill_pseudos): Ditto.
3089 (process_insn_for_elimination): Ditto.
3090 * lra-lives.c (reg_early_clobber_p): Use the new macros.
3091 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
3092 LRA_NON_CLOBBERED_ALT.
3094 2018-02-22 Martin Sebor <msebor@redhat.com>
3096 PR tree-optimization/84480
3097 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
3098 to maybe_diag_stxncpy_trunc. Call it.
3099 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
3100 from gimple_fold_builtin_strcpy. Print inlining stack.
3101 (handle_builtin_stxncpy): Print inlining stack.
3102 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
3104 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
3106 PR target/84176
3107 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
3108 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
3109 and -fcheck-pointer-bounds are used together.
3110 (indirect_thunk_prefix): New enum.
3111 (indirect_thunk_need_prefix): New function.
3112 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
3113 "_nt" instead of "_bnd" for NOTRACK prefix.
3114 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
3115 (output_indirect_thunk_function): Likewise.
3116 (): Likewise.
3117 (ix86_code_end): Update output_indirect_thunk_function calls.
3118 (ix86_output_indirect_branch_via_reg): Replace
3119 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
3120 (ix86_output_indirect_branch_via_push): Likewise.
3121 (ix86_output_function_return): Likewise.
3122 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
3123 incompatible with -fcf-protection=branch and
3124 -fcheck-pointer-bounds.
3126 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3128 PR target/83335
3129 * config/aarch64/aarch64.c (aarch64_print_address_internal):
3130 Change gcc_assert call to output_operand_lossage.
3132 2018-02-22 Steve Ellcey <sellcey@cavium.com>
3134 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
3136 2018-02-22 DJ Delorie <dj@redhat.com>
3137 Sebastian Perta <sebastian.perta@renesas.com>
3138 Oleg Endo <olegendo@gcc.gnu.org>
3140 * config/rx/rx.c (rx_rtx_costs): New function.
3141 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
3143 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
3145 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
3147 2018-02-22 Martin Liska <mliska@suse.cz>
3149 PR driver/83193
3150 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
3151 Add "native" as a possible value.
3153 2018-02-22 Martin Liska <mliska@suse.cz>
3155 PR driver/83193
3156 * config/i386/i386.c (ix86_option_override_internal):
3157 Add "native" as a possible value for -march and -mtune.
3159 2018-02-22 Jakub Jelinek <jakub@redhat.com>
3161 PR target/84502
3162 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
3163 to all type variants.
3165 PR tree-optimization/84503
3166 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
3167 width as info->bitpos + info->bitsize - start.
3168 (merged_store_group::merge_overlapping): Simplify width computation.
3169 (check_no_overlap): New function.
3170 (imm_store_chain_info::try_coalesce_bswap): Compute expected
3171 start + width and last_order of the group, fail if check_no_overlap
3172 fails.
3173 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
3174 to group if check_no_overlap fails.
3176 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3178 * config/rs6000/altivec.md: Delete contraint arguments to
3179 define_expand, define_split, and define_peephole2, and in
3180 define_insn_and_split if always unused.
3181 * config/rs6000/darwin.md: Ditto.
3182 * config/rs6000/dfp.md: Ditto.
3183 * config/rs6000/rs6000.md: Ditto.
3184 * config/rs6000/sync.md: Ditto.
3185 * config/rs6000/vector.md: Ditto.
3186 * config/rs6000/vsx.md: Ditto.
3188 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
3190 * config/rs6000/altivec.md: Write output control strings as braced
3191 blocks instead of double-quoted strings.
3192 * config/rs6000/darwin.md: Ditto.
3193 * config/rs6000/rs6000.md: Ditto.
3194 * config/rs6000/vector.md: Ditto.
3195 * config/rs6000/vsx.md: Ditto.
3197 2018-02-21 Jason Merrill <jason@redhat.com>
3199 PR c++/84314 - ICE with templates and fastcall attribute.
3200 * attribs.c (build_type_attribute_qual_variant): Remove assert.
3202 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3204 * ipa-cp.c (determine_versionability): Fix comment typos.
3206 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
3208 PR c/84229
3209 * ipa-cp.c (determine_versionability): Do not version functions caling
3210 va_arg_pack.
3212 2018-02-21 Martin Liska <mliska@suse.cz>
3214 PR driver/83193
3215 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
3216 Add "native" as a possible value.
3217 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
3218 the macro when native cpu detection is available.
3220 2018-02-21 Martin Liska <mliska@suse.cz>
3222 PR driver/83193
3223 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
3224 Add "native" as a possible value.
3225 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
3226 when native cpu detection is available.
3228 2018-02-21 Jakub Jelinek <jakub@redhat.com>
3229 Martin Sebor <msebor@redhat.com>
3231 PR tree-optimization/84478
3232 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
3233 false.
3234 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
3235 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
3236 support which is conservatively correct, for 2 only stay conservative
3237 for maxlen. Formatting and comment capitalization fixes. Add STRICT
3238 argument to the 2 argument get_range_strlen, adjust 6 arg
3239 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
3240 false.
3241 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
3242 (gimple_fold_builtin_strlen): Pass true as last argument to
3243 get_range_strlen.
3245 2018-02-20 Martin Sebor <msebor@redhat.com>
3247 PR middle-end/84095
3248 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
3249 (builtin_memref::set_base_and_offset): Same. Handle inner references.
3250 (builtin_memref::builtin_memref): Factor out parts into
3251 set_base_and_offset and call it.
3253 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3255 PR middle-end/84406
3256 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
3257 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
3258 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
3259 search at the associated MODE_INT.
3261 2018-02-20 Jeff Law <law@redhat.com>
3263 PR middle-end/82123
3264 PR tree-optimization/81592
3265 PR middle-end/79257
3266 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
3267 for range data rather than using global data.
3268 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
3269 range data rather than using global data.
3270 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
3271 pass it to children as needed.
3272 (struct directive::fmtresult): Similarly.
3273 (struct directive::set_width): Similarly.
3274 (struct directive::set_precision): Similarly.
3275 (format_integer, format_directive, parse_directive): Similarly.
3276 (format_none): Accept unnamed vr_values parameter.
3277 (format_percent, format_floating, format_character): Similarly.
3278 (format_string, format_plain): Similarly.
3279 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
3280 the EVRP range analyzer for range data rather than using global data.
3281 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
3282 gimple-ssa-evrp-analyze.h
3283 (class sprintf_dom_walker): Add after_dom_children member function.
3284 Add evrp_range_analyzer member.
3285 (sprintf_dom_walker::before_dom_children): Call into the EVRP
3286 range analyzer as needed.
3287 (sprintf_dom_walker::after_dom_children): New member function.
3288 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
3289 if not optimizing.
3290 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
3291 (evrp_range_analyzer::pop_to_marker): Likewise.
3293 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
3295 PR tree-optimization/84419
3296 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
3297 with the required type if its current type is compatible but
3298 different.
3300 2018-02-20 Jakub Jelinek <jakub@redhat.com>
3302 PR middle-end/82004
3303 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
3304 after vectorization.
3306 2018-02-20 Martin Liska <mliska@suse.cz>
3308 PR driver/83193
3309 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
3310 possible values if we don't have a hint.
3312 2018-02-20 Martin Liska <mliska@suse.cz>
3314 PR c/84310
3315 PR target/79747
3316 * final.c (shorten_branches): Build align_tab array with one
3317 more element.
3318 * opts.c (finish_options): Add alignment option limit check.
3319 (MAX_CODE_ALIGN): Likewise.
3320 (MAX_CODE_ALIGN_VALUE): Likewise.
3321 * doc/invoke.texi: Document maximum allowed option value for
3322 all -falign-* options.
3324 2018-02-19 Jakub Jelinek <jakub@redhat.com>
3326 PR target/84146
3327 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
3328 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
3329 * var-tracking.c (emit_note_insn_var_location): Remove all references
3330 to NOTE_INSN_CALL_ARG_LOCATION.
3331 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
3332 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
3333 Use copy_rtx_if_shared.
3334 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
3335 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
3336 (dwarf2out_var_location): Remove handling of
3337 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
3338 on call_insn.
3339 * final.c (final_scan_insn): Remove all references to
3340 NOTE_INSN_CALL_ARG_LOCATION.
3341 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
3342 before dumping final insns.
3343 * except.c (emit_note_eh_region_end): Remove all references to
3344 NOTE_INSN_CALL_ARG_LOCATION.
3345 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
3346 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
3347 * config/arc/arc.c (hwloop_optimize): Likewise.
3348 * config/arm/arm.c (create_fix_barrier): Likewise.
3349 * config/s390/s390.c (s390_chunkify_start): Likewise.
3350 * config/sh/sh.c (find_barrier): Likewise.
3351 * config/i386/i386.c (rest_of_insert_endbranch,
3352 ix86_seh_fixup_eh_fallthru): Likewise.
3353 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
3354 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
3355 * config/frv/frv.c (frv_function_prologue): Likewise.
3356 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
3357 reg note.
3358 (note_outside_basic_block_p): Remove all references to
3359 NOTE_INSN_CALL_ARG_LOCATION.
3360 * gengtype.c (adjust_field_rtx_def): Likewise.
3361 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
3362 Likewise.
3363 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
3364 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
3366 PR c++/84444
3367 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
3368 is ADDR_EXPR.
3370 PR tree-optimization/84452
3371 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
3372 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
3373 is NULL.
3375 2018-02-19 Martin Liska <mliska@suse.cz>
3377 PR sanitizer/82183
3378 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
3380 2018-02-19 Martin Liska <mliska@suse.cz>
3381 Richard Sandiford <richard.sandiford@linaro.org>
3383 PR tree-optimization/82491
3384 * gimple-fold.c (get_base_constructor): Make earlier bail out
3385 to prevent ubsan.
3387 2018-02-19 Carl Love <cel@us.ibm.com>
3389 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
3390 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
3391 BU_P8V_OVERLOAD_1.
3392 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
3393 P8V_BUILTIN_VEC_NEG.
3395 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
3397 * config/rl78/rl78.md (movdf): New define expand.
3399 2018-02-19 Martin Liska <mliska@suse.cz>
3401 PR other/80589
3402 * doc/invoke.texi: Fix typo.
3403 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
3405 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
3407 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
3408 handle rs6000_single_float and rs6000_double_float specially for
3409 e500 family CPUs.
3411 2018-02-16 Jeff Law <law@redhat.com>
3413 * config/rx/rx.c (add_pop_cfi_notes): New function.;
3414 (pop_regs): Use it.
3416 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3418 PR ipa/84425
3419 * ipa-inline.c (inline_small_functions): Fix a typo.
3421 2018-02-16 Nathan Sidwell <nathan@acm.org>
3423 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
3425 2018-02-16 Carl Love <cel@us.ibm.com>
3427 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
3428 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
3429 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
3430 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
3431 expansion to P8V_BUILTIN_VEC_FLOAT2.
3433 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
3435 PR rtl-optimization/70023
3436 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
3437 src_regno into account.
3439 2018-02-16 Carl Love <cel@us.ibm.com>
3441 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
3442 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
3443 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
3444 * config/rs6000/rs6000.c: Remove case statements for
3445 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
3446 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
3447 and P9V_BUILTIN_VEC_VINSERT4B.
3448 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
3449 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
3450 * config/rs6000/vsx.md:
3451 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
3452 vec_insert4b.
3454 2018-02-16 Carl Love <cel@us.ibm.com>
3456 * config/rs6000/altivec.h: Add builtin names vec_extract4b
3457 vec_insert4b.
3458 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
3459 definitions.
3460 * config/rs6000/rs6000-c.c: Add the definitions for
3461 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
3462 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
3463 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
3464 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
3465 definition for insert4b and define insn *insert3b_internal.
3466 * doc/extend.texi: Add documentation for vec_extract4b.
3468 2018-02-16 Nathan Sidwell <nathan@acm.org>
3470 * doc/extend.texi (Backwards Compatibility): Mention friend
3471 injection. Note for-scope is deprecated.
3472 * doc/invoke.texi (-ffriend-injection): Deprecate.
3474 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
3476 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
3477 that moved to I2, also allow destinations that are a paradoxical
3478 subreg (instead of a normal reg).
3480 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
3482 PR target/83831
3483 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
3484 to QImode.
3486 2018-02-16 Richard Biener <rguenther@suse.de>
3488 PR tree-optimization/84037
3489 PR tree-optimization/84016
3490 PR target/82862
3491 * config/i386/i386.c (ix86_builtin_vectorization_cost):
3492 Adjust vec_construct for the fact we need additional higher latency
3493 128bit inserts for AVX256 and AVX512 vector builds.
3494 (ix86_add_stmt_cost): Scale vector construction cost for
3495 elementwise loads.
3497 2018-02-16 Richard Biener <rguenther@suse.de>
3499 PR tree-optimization/84417
3500 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
3501 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
3502 (non_rewritable_lvalue_p): Likewise, use poly-ints.
3504 2018-02-16 Martin Liska <mliska@suse.cz>
3506 PR sanitizer/84307
3507 * internal-fn.def (ASAN_CHECK): Set proper flags.
3508 (ASAN_MARK): Likewise.
3510 2018-02-16 Julia Koval <julia.koval@intel.com>
3512 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
3513 from PTA_CANNONLAKE.
3515 2018-02-16 Jakub Jelinek <jakub@redhat.com>
3517 PR target/84272
3518 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
3519 Use ++iter rather than iter++ for std::list iterators.
3520 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
3521 defer deleting them until all nodes in the forest are processed. Do
3522 free even leaf nodes. Change to_process into auto_vec.
3524 PR bootstrap/84405
3525 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
3526 * vec.h (vec_default_construct): Use memset instead of placement new
3527 if BROKEN_VALUE_INITIALIZATION is defined.
3528 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
3529 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
3530 is defined.
3532 PR rtl-optimization/83723
3533 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
3534 * lra.c (lra_substitute_pseudo): Likewise. If true, use
3535 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
3536 recursive calls.
3537 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
3538 callers.
3539 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
3541 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
3543 PR rtl-optimization/81443
3544 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
3545 from inner REGs to paradoxical SUBREGs.
3547 2018-02-16 Richard Biener <rguenther@suse.de>
3549 PR tree-optimization/84399
3550 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
3551 For operands we can analyze at their definition make sure we can
3552 analyze them at each use as well.
3554 2018-02-16 Richard Biener <rguenther@suse.de>
3556 PR tree-optimization/84190
3557 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
3558 volatile accesses if the decl isn't volatile.
3560 2018-02-15 Jason Merrill <jason@redhat.com>
3562 PR c++/84314 - ICE with templates and fastcall attribute.
3563 * attribs.c (build_type_attribute_qual_variant): Don't clobber
3564 TYPE_CANONICAL on an existing type.
3566 2018-02-15 Jakub Jelinek <jakub@redhat.com>
3568 PR tree-optimization/84383
3569 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
3570 dstoff nor call operand_equal_p if dstbase is NULL.
3572 PR tree-optimization/84334
3573 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
3574 also a CONSTANT_CLASS_P, punt.
3576 2018-02-14 Jim Wilson <jimw@sifive.com>
3578 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
3579 first SMALL_OPERAND check. New local min_second_step. Move assert
3580 to where locals are set. Add TARGET_RVC support.
3581 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
3583 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
3585 * doc/invoke.texi: Correct -Wformat-overflow code sample.
3587 2018-02-14 Martin Sebor <msebor@redhat.com>
3589 PR tree-optimization/83698
3590 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
3591 arrays constrain the offset range to their bounds.
3592 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
3593 (builtin_access::overlap): Avoid setting the size of overlap if it's
3594 already been set.
3595 (maybe_diag_overlap): Also consider arrays when deciding what values
3596 of offsets to include in diagnostics.
3598 2018-02-14 Martin Sebor <msebor@redhat.com>
3600 PR c/84108
3601 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
3602 that correspond to the kind of a declaration.
3604 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
3606 PR target/83984
3607 * config/pa/pa.md: Load address of PIC label using the linkage table
3608 if the label is nonlocal.
3610 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
3612 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
3613 warning message if user requests -maltivec=be.
3614 * doc/invoke.texi: Document deprecation of -maltivec=be.
3616 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
3618 PR target/84220
3619 * config/rs6000/rs6000-c.c: Update definitions for
3620 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
3621 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
3623 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
3625 PR target/84239
3626 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
3627 add _get_ssp intrinsics. Remove argument from
3628 __builtin_ia32_rdssp[d|q].
3629 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
3630 * config/i386/i386-builtin.def: Remove argument from
3631 __builtin_ia32_rdssp[d|q].
3632 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
3633 ix86_expand_special_args_builtin for _rdssp[d|q].
3634 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
3635 Clear register before usage.
3636 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
3637 Add documentation for new _get_ssp and _inc_ssp intrinsics.
3639 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
3641 PR tree-optimization/84357
3642 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
3643 operand 1 of an ARRAY_REF too.
3645 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
3647 PR target/83831
3648 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
3649 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3650 declarations.
3651 (set_of_reg): New struct.
3652 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
3653 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
3654 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
3655 functions.
3656 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
3657 Split into bitclr, bitset, bitinvert patterns if appropriate.
3658 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
3659 use rx_fuse_in_memory_bitop.
3660 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
3661 to named insn, correct maximum insn length.
3663 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
3665 PR target/79242
3666 * machmode.def: Define a complex mode for PARTIAL_INT.
3667 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
3668 MODE_PARTIAL_INT.
3669 * doc/rtl.texi: Document CSPImode.
3670 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
3671 handling.
3672 (msp430_hard_regno_nregs_with_padding): Likewise.
3674 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
3676 PR target/84279
3677 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
3679 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
3681 PR rtl-optimization/84169
3682 * combine.c (try_combine): New variable split_i2i3. Set it to true if
3683 we generated a parallel as new i3 and we split that to new i2 and i3
3684 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
3685 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
3686 those to i2, not i1. Partially rewrite this scan code.
3688 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3690 PR c/82210
3691 * stor-layout.c (place_field): For variable length fields, adjust
3692 offset_align afterwards not just based on the field's alignment,
3693 but also on the size.
3695 PR middle-end/84309
3696 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
3697 of exps and logs in the use_exp2 case.
3699 2018-02-13 Jeff Law <law@redhat.com>
3701 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
3702 entry for "vector".
3704 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
3705 ARGS as unused.
3707 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
3709 PR debug/84342
3710 PR debug/84319
3711 * common.opt (gas-loc-support, gas-locview-support): New.
3712 (ginline-points, ginternal-reset-location-views): New.
3713 * doc/invoke.texi: Document them. Use @itemx where intended.
3714 (gvariable-location-views): Adjust.
3715 * target.def (reset_location_view): New.
3716 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
3717 (TARGET_RESET_LOCATION_VIEW): New.
3718 * doc/tm.texi: Rebuilt.
3719 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
3720 (dwarf2out_default_as_locview_support): New.
3721 (output_asm_line_debug_info): Use option variables.
3722 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
3723 (output_loc_list): Likewise.
3724 (add_high_low_attributes): Check option variables.
3725 Don't output entry view attribute in strict mode.
3726 (gen_inlined_subroutine_die): Check option variables.
3727 (dwarf2out_inline_entry): Likewise.
3728 (init_sections_and_labels): Likewise.
3729 (dwarf2out_early_finish): Likewise.
3730 (maybe_reset_location_view): New, from...
3731 (dwarf2out_var_location): ... here. Call it.
3732 * debug.h (dwarf2out_default_as_loc_support): Declare.
3733 (dwarf2out_default_as_locview_support): Declare.
3734 * hooks.c (hook_int_rtx_insn_0): New.
3735 * hooks.h (hook_int_rtx_insn_0): Declare.
3736 * toplev.c (process_options): Take -gas-loc-support and
3737 -gas-locview-support from dwarf2out. Enable
3738 -gvariable-location-views by default only with locview
3739 assembler support. Enable -ginternal-reset-location-views by
3740 default only if the target defines the corresponding hook.
3741 Enable -ginline-points by default if location views are
3742 enabled; force it disabled if statement frontiers are
3743 disabled.
3744 * tree-inline.c (expand_call_inline): Check option variables.
3745 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
3747 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
3749 PR tree-optimization/84321
3750 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
3751 handling. Also check whether the anti-range contains any values
3752 that satisfy the mask; switch to a VR_RANGE if not.
3754 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
3756 PR sanitizer/84340
3757 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
3759 2018-02-13 Martin Jambor <mjambor@suse.cz>
3761 PR c++/83990
3762 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
3763 of call statements, also set location of a load to a temporary.
3765 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
3767 * config/rl78/rl78.c (add_vector_labels): New function.
3768 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
3769 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
3770 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
3771 which checks that no arguments are passed.
3772 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
3773 * doc/extend.texi: Documentation for the new attribute.
3775 2018-02-13 Andreas Schwab <schwab@suse.de>
3777 * config/riscv/linux.h (CPP_SPEC): Define.
3779 2018-02-13 Jakub Jelinek <jakub@redhat.com>
3781 PR target/84335
3782 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
3783 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
3784 OPTION_MASK_ISA_AES as first argument to def_builtin_const
3785 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
3786 instead of OPTION_MASK_ISA_PCLMUL as first argument to
3787 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
3788 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
3789 temporarily for AES and PCLMUL builtins.
3791 PR tree-optimization/84339
3792 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
3793 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
3794 Formatting fixes.
3796 PR middle-end/84309
3797 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
3798 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
3799 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
3800 inline function.
3801 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
3802 inline function.
3803 * omp-simd-clone.h: New file.
3804 * omp-simd-clone.c: Include omp-simd-clone.h.
3805 (expand_simd_clones): No longer static.
3806 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
3807 cgraph.h and omp-simd-clone.h.
3808 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
3809 (vect_recog_widen_shift_pattern): Formatting fix.
3810 (vect_pattern_recog_1): Don't check optab for calls.
3812 PR target/84336
3813 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
3814 operands[2] into a REG before using gen_lowpart on it.
3816 2018-02-12 Jeff Law <law@redhat.com>
3818 PR target/83760
3819 * config/sh/sh.c (find_barrier): Consider a sibling call
3820 a barrier as well.
3822 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
3823 successfully back substituting a reg.
3825 2018-02-12 Richard Biener <rguenther@suse.de>
3827 PR tree-optimization/84037
3828 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
3829 parameter, move visited init to caller.
3830 (vect_slp_analyze_operations): Separate cost from validity
3831 check, initialize visited once for all instances.
3832 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
3833 for all instances.
3834 * tree-vect-stmts.c (vect_model_simple_cost): Make early
3835 out an assert.
3836 (vect_model_promotion_demotion_cost): Likewise.
3837 (vectorizable_bswap): Guard cost modeling with !slp_node
3838 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
3839 SLP stmts.
3840 (vectorizable_call): Likewise.
3841 (vectorizable_conversion): Likewise.
3842 (vectorizable_assignment): Likewise.
3843 (vectorizable_shift): Likewise.
3844 (vectorizable_operation): Likewise.
3845 (vectorizable_store): Likewise.
3846 (vectorizable_load): Likewise.
3847 (vectorizable_condition): Likewise.
3848 (vectorizable_comparison): Likewise.
3850 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
3852 PR sanitizer/84307
3853 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
3854 (ASAN_MARK): Fix fnspec to account for return value, change pointer
3855 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
3857 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
3859 PR middle-end/83665
3860 * params.def (inline-min-speedup): Increase from 8 to 15.
3861 (max-inline-insns-auto): Decrease from 40 to 30.
3862 * ipa-split.c (consider_split): Add some buffer for function to
3863 be considered inlining candidate.
3864 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
3865 default values.
3867 2018-02-12 Richard Biener <rguenther@suse.de>
3869 PR tree-optimization/84037
3870 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
3871 matched stmts if we cannot swap the non-matched ones.
3873 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3875 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
3876 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
3877 _mm_maskz_scalef_round_ss): New intrinsics.
3878 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
3879 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
3880 __builtin_ia32_scalefss_round): Remove.
3881 (__builtin_ia32_scalefsd_mask_round,
3882 __builtin_ia32_scalefss_mask_round): New intrinsics.
3883 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
3884 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
3885 ((match_operand:VF_128 2 "<round_nimm_predicate>"
3886 "<round_constraint>")): Changed to ...
3887 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
3888 "<round_scalar_constraint>")): ... this.
3889 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
3890 %0, %1, %2<round_op3>}"): Changed to ...
3891 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
3892 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
3893 %2<round_scalar_mask_op3>}"): ... this.
3894 * config/i386/subst.md (round_scalar_nimm_predicate): New.
3896 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
3898 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
3899 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
3900 (_mm_maskz_sqrt_round_ss): New intrinsics.
3901 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
3902 (__builtin_ia32_sqrtsd_mask_round)
3903 (__builtin_ia32_sqrtss_mask_round): New builtins.
3904 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
3905 (__builtin_ia32_sqrtss_round): Remove.
3906 (__builtin_ia32_sqrtsd_mask_round)
3907 (__builtin_ia32_sqrtss_mask_round): New builtins.
3908 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
3909 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
3910 ((match_operand:VF_128 1 "vector_operand"
3911 "xBm,<round_constraint>")): Changed to ...
3912 ((match_operand:VF_128 1 "vector_operand"
3913 "xBm,<round_scalar_constraint>")): ... this.
3914 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
3915 %0, %2, %<iptr>1<round_op3>}): Changed to ...
3916 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
3917 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
3918 %<iptr>1<round_scalar_mask_op3>}): ... this.
3919 ((set_attr "prefix" "<round_prefix>")): Changed to ...
3920 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
3922 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
3924 PR target/84266
3925 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
3926 Cast vec_cmpeq result to correct type.
3927 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
3928 Cast vec_cmpgt result to correct type.
3930 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
3932 * final.c (final_scan_insn_1): Renamed from...
3933 (final_scan_insn): ... this. New wrapper, to recover
3934 seen from the outermost call in recursive ones.
3935 * config/sparc/sparc.c (output_return): Drop seen from call.
3936 (output_sibcall): Likewise.
3937 * config/visium/visium.c (output_branch): Likewise.
3939 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
3941 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
3942 function label.
3944 2018-02-10 Alan Modra <amodra@gmail.com>
3946 PR target/84300
3947 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
3948 Specify LR as an input.
3950 2018-02-10 Jakub Jelinek <jakub@redhat.com>
3952 PR sanitizer/83987
3953 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
3954 remove_member_access_dummy_vars): New functions.
3955 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
3956 lower_omp_1, execute_lower_omp): Use them.
3958 PR rtl-optimization/84308
3959 * shrink-wrap.c (spread_components): Release todo vector.
3961 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
3963 PR rtl-optimization/57193
3964 * ira-color.c (struct allocno_color_data): Add member
3965 conflict_allocno_hard_prefs.
3966 (update_conflict_allocno_hard_prefs): New.
3967 (bucket_allocno_compare_func): Add a preference based on
3968 conflict_allocno_hard_prefs.
3969 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
3970 (color_allocnos): Remove a dead code. Initiate
3971 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
3973 2018-02-09 Jakub Jelinek <jakub@redhat.com>
3975 PR target/84226
3976 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
3977 constraint from =wa to wa. Avoid a subreg on the output operand,
3978 instead use a pseudo and subreg it in a move.
3979 (p9_xxbrd_<mode>): Changed to ...
3980 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
3981 (p9_xxbrd_v2df): New expander.
3982 (p9_xxbrw_<mode>): Changed to ...
3983 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
3984 (p9_xxbrw_v4sf): New expander.
3986 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
3988 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
3990 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
3992 PR target/83926
3993 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
3994 multiply in 32-bit mode.
3995 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
3996 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
3997 mode.
3999 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
4001 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
4002 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
4003 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
4004 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
4006 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
4008 PR lto/84213
4009 * dwarf2out.c (is_trivial_indirect_ref): New function.
4010 (dwarf2out_late_global_decl): Do not generate a location
4011 attribute for variables that have a non-trivial DECL_VALUE_EXPR
4012 and that are not defined in the current unit.
4014 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4016 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
4017 instead of a libcall for UNORDERED.
4019 2018-02-09 Tamar Christina <tamar.christina@arm.com>
4021 PR target/82641
4022 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
4023 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
4025 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4027 PR target/PR84295
4028 * config/s390/s390.c (s390_set_current_function): Invoke
4029 s390_indirect_branch_settings also if fndecl didn't change.
4031 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4033 * config/rs6000/rs6000.md (blockage): Set length to zero.
4035 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
4037 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
4039 2018-02-09 Jakub Jelinek <jakub@redhat.com>
4041 PR sanitizer/84285
4042 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
4043 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
4044 -static-lib*san.
4046 PR debug/84252
4047 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
4048 PARALLEL incoming that failed vt_get_decl_and_offset check.
4050 PR middle-end/84237
4051 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
4052 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
4053 TREE_READONLY bit.
4054 (get_variable_section): For decls in named .bss* sections pass true as
4055 second argument to bss_initializer_p.
4057 2018-02-09 Marek Polacek <polacek@redhat.com>
4058 Jakub Jelinek <jakub@redhat.com>
4060 PR c++/83659
4061 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
4062 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
4063 Sync some changes from cxx_fold_indirect_ref.
4065 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
4067 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
4068 markers.
4069 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
4070 (BLOCK_INLINE_ENTRY_LABEL): New.
4071 (dwarf2out_var_location): Disregard inline entry markers.
4072 (inline_entry_data): New struct.
4073 (inline_entry_data_hasher): New hashtable type.
4074 (inline_entry_data_hasher::hash): New.
4075 (inline_entry_data_hasher::equal): New.
4076 (inline_entry_data_table): New variable.
4077 (add_high_low_attributes): Add DW_AT_entry_pc and
4078 DW_AT_GNU_entry_view attributes if a pending entry is found
4079 in inline_entry_data_table. Add old entry_pc attribute only
4080 if debug nonbinding markers are disabled.
4081 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
4082 markers are enabled.
4083 (block_within_block_p, dwarf2out_inline_entry): New.
4084 (dwarf2out_finish): Check that no entries remained in
4085 inline_entry_data_table.
4086 * final.c (reemit_insn_block_notes): Handle inline entry notes.
4087 (final_scan_insn, notice_source_line): Likewise.
4088 (rest_of_clean_state): Skip inline entry markers.
4089 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
4090 markers.
4091 * gimple.c (gimple_build_debug_inline_entry): New.
4092 * gimple.h (enum gimple_debug_subcode): Add
4093 GIMPLE_DEBUG_INLINE_ENTRY.
4094 (gimple_build_debug_inline_entry): Declare.
4095 (gimple_debug_inline_entry_p): New.
4096 (gimple_debug_nonbind_marker_p): Adjust.
4097 * insn-notes.def (INLINE_ENTRY): New.
4098 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
4099 inline entry marker notes.
4100 (print_insn): Likewise.
4101 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
4102 (INSN_DEBUG_MARKER_KIND): Likewise.
4103 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
4104 * tree-inline.c (expand_call_inline): Build and insert
4105 debug_inline_entry stmt.
4106 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
4107 inline entry blocks early, if nonbind markers are enabled.
4108 (dump_scope_block): Dump fragment info.
4109 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
4110 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
4111 (gimple_build_debug_inline_entry): New.
4112 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
4113 Enable/disable inline entry points too.
4114 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
4115 (DEBUG_INSN): Describe inline entry markers.
4117 * common.opt (gvariable-location-views): New.
4118 (gvariable-location-views=incompat5): New.
4119 * config.in: Rebuilt.
4120 * configure: Rebuilt.
4121 * configure.ac: Test assembler for view support.
4122 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
4123 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
4124 * dwarf2out.c (var_loc_view): New typedef.
4125 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
4126 (dwarf2out_locviews_in_attribute): New.
4127 (dwarf2out_locviews_in_loclist): New.
4128 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
4129 (enum dw_line_info_opcode): Add LI_adv_address.
4130 (struct dw_line_info_table): Add view.
4131 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
4132 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
4133 (zero_view_p): New variable.
4134 (ZERO_VIEW_P): New macro.
4135 (output_asm_line_debug_info): New.
4136 (struct var_loc_node): Add view.
4137 (add_AT_view_list, AT_loc_list): New.
4138 (add_var_loc_to_decl): Add view param. Test it against last.
4139 (new_loc_list): Add view params. Record them.
4140 (AT_loc_list_ptr): Handle loc and view lists.
4141 (view_list_to_loc_list_val_node): New.
4142 (print_dw_val): Handle dw_val_class_view_list.
4143 (size_of_die): Likewise.
4144 (value_format): Likewise.
4145 (loc_list_has_views): New.
4146 (gen_llsym): Set vl_symbol too.
4147 (maybe_gen_llsym, skip_loc_list_entry): New.
4148 (dwarf2out_maybe_output_loclist_view_pair): New.
4149 (output_loc_list): Output view list or entries too.
4150 (output_view_list_offset): New.
4151 (output_die): Handle dw_val_class_view_list.
4152 (output_dwarf_version): New.
4153 (output_compilation_unit_header): Use it.
4154 (output_skeleton_debug_sections): Likewise.
4155 (output_rnglists, output_line_info): Likewise.
4156 (output_pubnames, output_aranges): Update version comments.
4157 (output_one_line_info_table): Output view numbers in asm comments.
4158 (dw_loc_list): Determine current endview, pass it to new_loc_list.
4159 Call maybe_gen_llsym.
4160 (loc_list_from_tree_1): Adjust.
4161 (add_AT_location_description): Create view list attribute if
4162 needed, check it's absent otherwise.
4163 (convert_cfa_to_fb_loc_list): Adjust.
4164 (maybe_emit_file): Call output_asm_line_debug_info for test.
4165 (dwarf2out_var_location): Reset views as needed. Precompute
4166 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
4167 attribute. Set view.
4168 (new_line_info_table): Reset next view.
4169 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
4170 (dwarf2out_source_line): Likewise. Output view resets and labels to
4171 the assembler, or select appropriate line info opcodes.
4172 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
4173 (optimize_string_length): Catch it. Adjust.
4174 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
4175 dw_val_class_view_list, and remove it if no longer needed.
4176 (hash_loc_list): Hash view numbers.
4177 (loc_list_hasher::equal): Compare them.
4178 (optimize_location_lists): Check whether a view list symbol is
4179 needed, and whether the locview attribute is present, and
4180 whether they match. Remove the locview attribute if no longer
4181 needed.
4182 (index_location_lists): Call skip_loc_list_entry for test.
4183 (dwarf2out_finish): Call output_asm_line_debug_info for test.
4184 Use output_dwarf_version.
4185 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
4186 (struct dw_val_node): Add val_view_list.
4187 * final.c (SEEN_NEXT_VIEW): New.
4188 (set_next_view_needed): New.
4189 (clear_next_view_needed): New.
4190 (maybe_output_next_view): New.
4191 (final_start_function): Rename to...
4192 (final_start_function_1): ... this. Take pointer to FIRST,
4193 add SEEN parameter. Emit param bindings in the initial view.
4194 (final_start_function): Reintroduce SEEN-less interface.
4195 (final): Rename to...
4196 (final_1): ... this. Take SEEN parameter. Output final pending
4197 next view at the end.
4198 (final): Reintroduce seen-less interface.
4199 (final_scan_insn): Output pending next view before switching
4200 sections or ending a block. Mark the next view as needed when
4201 outputting variable locations. Notify debug backend of section
4202 changes, and of location view changes.
4203 (rest_of_handle_final): Adjust.
4204 * toplev.c (process_options): Autodetect value for debug variable
4205 location views option. Warn on incompat5 without -gdwarf-5.
4206 * doc/invoke.texi (gvariable-location-views): New.
4207 (gvariable-location-views=incompat5): New.
4208 (gno-variable-location-views): New.
4210 2018-02-08 David Malcolm <dmalcolm@redhat.com>
4212 PR tree-optimization/84136
4213 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
4214 that the result of find_edge is non-NULL.
4216 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4218 PR target/83008
4219 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
4220 storing integer register in SImode. Fix cost of 256 and 512
4221 byte aligned SSE register store.
4223 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
4225 * config/i386/i386.c (ix86_multiplication_cost): Fix
4226 multiplication cost for TARGET_AVX512DQ.
4228 2018-02-08 Marek Polacek <polacek@redhat.com>
4230 PR tree-optimization/84238
4231 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
4232 get_range_strlen.
4234 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4236 PR tree-optimization/84265
4237 * tree-vect-stmts.c (vectorizable_store): Don't treat
4238 VMAT_CONTIGUOUS accesses as grouped.
4239 (vectorizable_load): Likewise.
4241 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4243 PR tree-optimization/81635
4244 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
4245 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
4246 (test_round_for_mask): New functions.
4247 (wide_int_cc_tests): Call test_round_for_mask.
4248 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
4249 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
4250 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
4251 range returned by get_range_info.
4253 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
4255 PR ipa/81360
4256 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
4257 * symtab.c: Include builtins.h
4258 (symtab_node::output_to_lto_symbol_table_p): Move here
4259 from lto-streamer-out.c:output_symbol_p.
4260 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
4261 (output_symbol_p): Move all logic to symtab.c
4262 (produce_symtab): Update.
4264 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
4266 * config/s390/s390-opts.h (enum indirect_branch): Define.
4267 * config/s390/s390-protos.h (s390_return_addr_from_memory)
4268 (s390_indirect_branch_via_thunk)
4269 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
4270 (enum s390_indirect_branch_type): Define.
4271 * config/s390/s390.c (struct s390_frame_layout, struct
4272 machine_function): Remove.
4273 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
4274 (indirect_branch_table_label_no, indirect_branch_table_name):
4275 Define variables.
4276 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
4277 (enum s390_indirect_branch_option): Define.
4278 (s390_return_addr_from_memory): New function.
4279 (s390_handle_string_attribute): New function.
4280 (s390_attribute_table): Add new attribute handler.
4281 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
4282 (s390_indirect_branch_via_thunk): New function.
4283 (s390_indirect_branch_via_inline_thunk): New function.
4284 (s390_function_ok_for_sibcall): When jumping via thunk disallow
4285 sibling call optimization for non z10 compiles.
4286 (s390_emit_call): Force indirect branch target to be a single
4287 register. Add r1 clobber for non-z10 compiles.
4288 (s390_emit_epilogue): Emit return jump via return_use expander.
4289 (s390_reorg): Handle JUMP_INSNs as execute targets.
4290 (s390_option_override_internal): Perform validity checks for the
4291 new command line options.
4292 (s390_indirect_branch_attrvalue): New function.
4293 (s390_indirect_branch_settings): New function.
4294 (s390_set_current_function): Invoke s390_indirect_branch_settings.
4295 (s390_output_indirect_thunk_function): New function.
4296 (s390_code_end): Implement target hook.
4297 (s390_case_values_threshold): Implement target hook.
4298 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
4299 macros.
4300 * config/s390/s390.h (struct s390_frame_layout)
4301 (struct machine_function): Move here from s390.c.
4302 (TARGET_INDIRECT_BRANCH_NOBP_RET)
4303 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
4304 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
4305 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
4306 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
4307 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
4308 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
4309 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
4310 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
4311 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
4312 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
4313 (mnemonic attribute): Add values which aren't recognized
4314 automatically.
4315 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
4316 pattern for branch conversion. Fix mnemonic attribute.
4317 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
4318 indirect branch via thunk if requested.
4319 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
4320 ("*indirect_jump"): Disable for branch conversion using out of
4321 line thunks.
4322 ("indirect_jump_via_thunk<mode>_z10")
4323 ("indirect_jump_via_thunk<mode>")
4324 ("indirect_jump_via_inlinethunk<mode>_z10")
4325 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
4326 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
4327 ("casesi_jump_via_inlinethunk<mode>_z10")
4328 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
4329 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
4330 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
4331 ("*indirect2_jump"): Disable for branch conversion.
4332 ("casesi_jump"): Turn into expander and expand patterns for branch
4333 conversion.
4334 ("return_use"): New expander.
4335 ("*return"): Emit return via thunk and rename it to ...
4336 ("*return<mode>"): ... this one.
4337 * config/s390/s390.opt: Add new options and and enum for the
4338 option values.
4340 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
4342 * lra-constraints.c (match_reload): Unconditionally use
4343 gen_lowpart_SUBREG, rather than selecting between that
4344 and equivalent gen_rtx_SUBREG code.
4346 2018-02-08 Richard Biener <rguenther@suse.de>
4348 PR tree-optimization/84233
4349 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
4350 changed flag instead of boguously re-using phi_inserted.
4352 2018-02-08 Martin Jambor <mjambor@suse.cz>
4354 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
4355 static local variables.
4357 2018-02-08 Richard Biener <rguenther@suse.de>
4359 PR tree-optimization/84278
4360 * tree-vect-stmts.c (vectorizable_store): When looking for
4361 smaller vector types to perform grouped strided loads/stores
4362 make sure the mode is supported by the target.
4363 (vectorizable_load): Likewise.
4365 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4367 * config/aarch64/aarch64.c (aarch64_components_for_bb):
4368 Increase LDP/STP opportunities by adding adjacent callee-saves.
4370 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
4372 PR rtl-optimization/84068
4373 PR rtl-optimization/83459
4374 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
4376 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
4378 PR tree-optimization/84224
4379 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
4380 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
4381 non-zero arguments.
4383 2018-02-07 Iain Sandoe <iain@codesourcery.com>
4385 PR target/84113
4386 * config/rs6000/altivec.md (*restore_world): Remove LR use.
4387 * config/rs6000/predicates.md (restore_world_operation): Adjust op
4388 count, remove one USE.
4390 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
4392 * doc/install.texi (Configuration): Document the
4393 --with-long-double-format={ibm,ieee} PowerPC configuration
4394 options.
4396 PR target/84154
4397 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
4398 Convert from define_expand to be define_insn_and_split. Rework
4399 float/double/_Float128 conversions to QI/HI/SImode to work with
4400 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
4401 conversions to QI/HImode types did a store and then a load to
4402 truncate the value. For conversions to VSX registers, don't split
4403 the insn, instead emit the code directly. Use the code iterator
4404 any_fix to combine signed and unsigned conversions.
4405 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
4406 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
4407 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4408 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
4409 (fix_<mode>di2_hw): Likewise.
4410 (fixuns_<mode>di2_hw): Likewise.
4411 (fix_<mode>si2_hw): Likewise.
4412 (fixuns_<mode>si2_hw): Likewise.
4413 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
4414 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
4415 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
4416 fix<uns>_trunc<SFDF:mode>si2_p8.
4417 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
4418 used.
4419 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
4420 (fix<uns>_<mode>_mem): Likewise.
4421 (fctiw<u>z_<mode>_mem): Likewise.
4422 (fix<uns>_<mode>_mem): Likewise.
4423 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
4424 the register allocator from doing a direct move to the GPRs to do
4425 a store, and instead use the ISA 3.0 store byte/half-word from
4426 vector register instruction. For IEEE 128-bit floating point,
4427 also optimize stores of 32-bit ints.
4428 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
4430 2018-02-07 Alan Hayward <alan.hayward@arm.com>
4432 * genextract.c (push_pathstr_operand): New function to support
4433 [a-zA-Z].
4434 (walk_rtx): Call push_pathstr_operand.
4435 (print_path): Support [a-zA-Z].
4437 2018-02-07 Richard Biener <rguenther@suse.de>
4439 PR tree-optimization/84037
4440 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
4441 (cse_and_gimplify_to_preheader): Declare.
4442 (vect_get_place_in_interleaving_chain): Likewise.
4443 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4444 ivexpr_map.
4445 (_loop_vec_info::~_loop_vec_info): Delete it.
4446 (cse_and_gimplify_to_preheader): New function.
4447 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
4448 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
4449 (vectorizable_load): Likewise. For grouped stores always base
4450 the IV on the first element.
4451 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
4452 condition before gimplifying.
4454 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4456 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
4457 *DIV_EXPR and *MOD_EXPR.
4459 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
4461 PR target/84248
4462 * config/i386/i386.c (ix86_option_override_internal): Mask out
4463 the CF_SET bit when checking -fcf-protection.
4465 2018-02-07 Tom de Vries <tom@codesourcery.com>
4467 PR libgomp/84217
4468 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
4469 enough.
4471 2018-02-07 Richard Biener <rguenther@suse.de>
4473 PR tree-optimization/84204
4474 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4475 this place.
4477 PR tree-optimization/84205
4478 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
4479 special-case isl_ast_op_zdiv_r.
4481 PR tree-optimization/84223
4482 * graphite-scop-detection.c (gather_bbs::before_dom_children):
4483 Only add conditions from within the region.
4484 (gather_bbs::after_dom_children): Adjust.
4486 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
4488 PR target/84209
4489 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
4490 * config/avr/avr.md: Only post-reload split REG-REG moves if
4491 either register is GENERAL_REG_P.
4493 2018-02-07 Jakub Jelinek <jakub@redhat.com>
4495 PR tree-optimization/84235
4496 * tree-ssa-scopedtables.c
4497 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
4498 if the subtraction is performed in floating point type where NaNs are
4499 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
4500 build 1. Formatting fix.
4502 2018-02-06 Jakub Jelinek <jakub@redhat.com>
4504 PR target/84146
4505 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
4506 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
4507 and skip it regardless of bb boundaries. Use CALL_P macro,
4508 don't test INSN_P (insn) together with CALL_P or JUMP_P check
4509 unnecessarily, formatting fix.
4511 2018-02-06 Michael Collison <michael.collison@arm.com>
4513 * config/arm/thumb2.md:
4514 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
4515 (*thumb_mov_notscc): Ditto.
4517 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
4519 PR target/84154
4520 * config/rs6000/rs6000.md (su code attribute): Use "u" for
4521 unsigned_fix, not "s".
4523 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4525 * configure.ac (gcc_fn_eh_frame_ro): New function.
4526 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
4527 correct .eh_frame permissions.
4528 * configure: Regenerate.
4530 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
4532 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
4533 irrelevant options.
4535 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
4537 * config/rs6000/rs6000.c (rs6000_option_override_internal):
4538 Display warning message for -mno-speculate-indirect-jumps.
4540 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
4542 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
4543 Undocumented.
4544 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
4546 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
4548 PR tree-optimization/84225
4549 * tree-eh.c (find_trapping_overflow): Only call
4550 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
4552 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4554 PR target/84145
4555 * config/i386/i386.c: Reimplement the check of possible options
4556 -mibt/-mshstk conbination. Change error messages.
4557 * doc/invoke.texi: Fix a typo: remove extra '='.
4559 2018-02-06 Marek Polacek <polacek@redhat.com>
4561 PR tree-optimization/84228
4562 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
4564 2018-02-06 Tamar Christina <tamar.christina@arm.com>
4566 PR target/82641
4567 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
4568 emitted arch directives.
4569 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
4570 __ARM_FEATURE_COPROC before changing architectures.
4572 2018-02-06 Richard Biener <rguenther@suse.de>
4574 * config/i386/i386.c (print_reg): Fix typo.
4575 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
4577 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
4579 * configure: Regenerate.
4581 2018-02-05 Martin Sebor <msebor@redhat.com>
4583 PR tree-optimization/83369
4584 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
4585 inlining context.
4587 2018-02-05 Martin Liska <mliska@suse.cz>
4589 * doc/invoke.texi: Cherry-pick upstream r323995.
4591 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
4593 * ira.c (ira_init_register_move_cost): Adjust comment.
4595 2018-02-05 Martin Liska <mliska@suse.cz>
4597 PR gcov-profile/84137
4598 * doc/gcov.texi: Fix typo in documentation.
4600 2018-02-05 Martin Liska <mliska@suse.cz>
4602 PR gcov-profile/83879
4603 * doc/gcov.texi: Document necessity of --dynamic-list-data when
4604 using dlopen functionality.
4606 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
4608 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
4609 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
4610 _mm_maskz_range_ss, _mm_mask_range_round_ss,
4611 _mm_maskz_range_round_ss): New intrinsics.
4612 (__builtin_ia32_rangesd128_round)
4613 (__builtin_ia32_rangess128_round): Remove.
4614 (__builtin_ia32_rangesd128_mask_round,
4615 __builtin_ia32_rangess128_mask_round): New builtins.
4616 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
4617 __builtin_ia32_rangess128_round): Remove.
4618 (__builtin_ia32_rangesd128_mask_round,
4619 __builtin_ia32_rangess128_mask_round): New builtins.
4620 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
4621 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
4622 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
4623 "<round_saeonly_constraint>")): Changed to ...
4624 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
4625 "<round_saeonly_scalar_constraint>")): ... this.
4626 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
4627 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
4628 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
4629 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
4630 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
4632 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
4634 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
4635 options.
4636 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
4637 Remove all values except native, 8540 and 8548.
4639 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
4641 * config/i386/i386.c (ix86_output_function_return): Pass
4642 INVALID_REGNUM, instead of -1, as invalid register number to
4643 indirect_thunk_name and output_indirect_thunk.
4645 2018-02-02 Julia Koval <julia.koval@intel.com>
4647 * config.gcc: Add -march=icelake.
4648 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
4649 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
4650 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
4651 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
4652 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
4653 (processor_target_table): Add icelake.
4654 (ix86_option_override_internal): Handle new PTAs.
4655 (get_builtin_code_for_version): Handle icelake.
4656 (M_INTEL_COREI7_ICELAKE): New.
4657 (fold_builtin_cpu): Handle icelake.
4658 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
4659 * doc/invoke.texi: Add -march=icelake.
4661 2018-02-02 Julia Koval <julia.koval@intel.com>
4663 * config/i386/i386.c (ix86_option_override_internal): Change flags type
4664 to wide_int_bitmask.
4665 * wide-int-bitmask.h: New.
4667 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
4669 PR target/84066
4670 * config/i386/i386.md: Replace Pmode with word_mode in
4671 builtin_setjmp_setup and builtin_longjmp to support x32.
4673 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
4675 PR target/56010
4676 PR target/83743
4677 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
4678 #include "opts.h".
4679 (rs6000_supported_cpu_names): New static variable.
4680 (linux_cpu_translation_table): Likewise.
4681 (elf_platform) <cpu>: Define new static variable and use it.
4682 Translate kernel AT_PLATFORM name to canonical name if needed.
4683 Error if platform name is unknown.
4685 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
4687 PR target/84089
4688 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
4690 2018-02-01 Jeff Law <law@redhat.com>
4692 PR target/84128
4693 * config/i386/i386.c (release_scratch_register_on_entry): Add new
4694 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
4695 the scratch if RELEASE_VIA_POP is false.
4696 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
4697 If we have to save a temporary register, decrement SIZE appropriately.
4698 Pass new arguments to release_scratch_register_on_entry.
4699 (ix86_adjust_stack_and_probe): Likewise.
4700 (ix86_emit_probe_stack_range): Pass new arguments to
4701 release_scratch_register_on_entry.
4703 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
4705 PR rtl-optimization/84157
4706 * combine.c (change_zero_ext): Use REG_P predicate in
4707 front of HARD_REGISTER_P predicate.
4709 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
4711 * config/avr/avr.c (avr_option_override): Move disabling of
4712 -fdelete-null-pointer-checks to...
4713 * common/config/avr/avr-common.c (avr_option_optimization_table):
4714 ...here.
4716 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4718 PR tree-optimization/81635
4719 * tree-data-ref.c (split_constant_offset_1): For types that
4720 wrap on overflow, try to use range info to prove that wrapping
4721 cannot occur.
4723 2018-02-01 Renlin Li <renlin.li@arm.com>
4725 PR target/83370
4726 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
4727 TAILCALL_ADDR_REGS.
4728 (aarch64_register_move_cost): Likewise.
4729 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
4730 TAILCALL_ADDR_REGS.
4731 (REG_CLASS_NAMES): Likewise.
4732 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
4733 TAILCALL_ADDR_REGS. Remove IP registers.
4734 * config/aarch64/aarch64.md (Ucs): Update register constraint.
4736 2018-02-01 Richard Biener <rguenther@suse.de>
4738 * domwalk.h (dom_walker::dom_walker): Add additional constructor
4739 for specifying RPO order and allow NULL for that.
4740 * domwalk.c (dom_walker::dom_walker): Likewise.
4741 (dom_walker::walk): Handle NULL RPO order.
4742 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
4743 in RPO order.
4744 (rewrite_update_dom_walker): Likewise.
4745 (mark_def_dom_walker): Likewise.
4747 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4749 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
4750 (aarch64_maybe_expand_sve_subreg_move): Declare.
4751 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
4752 * config/aarch64/predicates.md (aarch64_any_register_operand): New
4753 predicate.
4754 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
4755 that are semantically a reverse operation.
4756 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
4757 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
4758 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
4759 functions.
4760 (aarch64_can_change_mode_class): For big-endian, forbid changes
4761 between two SVE modes if they have different element sizes.
4763 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4765 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
4766 the TImode handling for big-endian targets.
4768 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4770 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
4771 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
4772 not just bytes.
4773 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
4774 Remove BSWAP handing for big-endian targets and use the form of
4775 LD1RQ appropariate for the mode.
4777 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4779 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
4780 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
4781 duplicated element.
4783 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
4785 PR tearget/83845
4786 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
4787 check for operands that need to go through aarch64_sve_reload_be.
4789 2018-02-01 Jakub Jelinek <jakub@redhat.com>
4791 PR tree-optimization/81661
4792 PR tree-optimization/84117
4793 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
4794 * tree-eh.c: Include gimplify.h.
4795 (find_trapping_overflow, replace_trapping_overflow,
4796 rewrite_to_non_trapping_overflow): New functions.
4797 * tree-vect-loop.c: Include tree-eh.h.
4798 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
4799 * tree-data-ref.c: Include tree-eh.h.
4800 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
4802 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
4804 PR rtl-optimization/84123
4805 * combine.c (change_zero_ext): Check if hard register satisfies
4806 can_change_dest_mode before calling gen_lowpart_SUBREG.
4808 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
4810 PR target/82444
4811 * ira.c (ira_init_register_move_cost): Remove assert.
4813 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4815 PR rtl-optimization/84071
4816 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
4817 * doc/tm.texi: Regenerate.
4819 2018-01-31 Richard Biener <rguenther@suse.de>
4821 PR tree-optimization/84132
4822 * tree-data-ref.c (analyze_miv_subscript): Properly
4823 check whether evolution_function_is_affine_multivariate_p
4824 before calling gcd_of_steps_may_divide_p.
4826 2018-01-31 Julia Koval <julia.koval@intel.com>
4828 PR target/83618
4829 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
4830 * config/i386/i386.md (rdpid_rex64) New.
4831 (rdpid): Make 32bit only.
4833 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
4835 PR lto/84105
4836 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
4837 an IDENTIFIER_NODE for FUNCTION_TYPE's.
4839 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4841 Revert
4842 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
4844 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
4846 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
4848 PR rtl-optimization/84071
4849 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
4850 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
4852 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4854 * config/arc/arc.c (arc_handle_aux_attribute): New function.
4855 (arc_attribute_table): Add 'aux' attribute.
4856 (arc_in_small_data_p): Consider aux like variables.
4857 (arc_is_aux_reg_p): New function.
4858 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
4859 (arc_get_aux_arg): New function.
4860 (prepare_move_operands): Handle aux-register access.
4861 (arc_handle_aux_attribute): New function.
4862 * doc/extend.texi (ARC Variable attributes): Add subsection.
4864 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
4866 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
4867 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
4868 (arc_attribute_table): Add 'uncached' attribute.
4869 (arc_print_operand): Print '.di' flag for uncached memory
4870 accesses.
4871 (arc_in_small_data_p): Do not consider for small data the uncached
4872 types.
4873 (arc_is_uncached_mem_p): New function.
4874 * config/arc/predicates.md (compact_store_memory_operand): Check
4875 for uncached memory accesses.
4876 (nonvol_nonimm_operand): Likewise.
4877 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
4879 2018-01-31 Jakub Jelinek <jakub@redhat.com>
4881 PR c/84100
4882 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
4883 falign-loops=): Add Optimization flag.
4885 2018-01-30 Jeff Law <law@redhat.com>
4887 PR target/84064
4888 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
4889 INT_REGISTERS_SAVED. Check it prior to calling
4890 get_scratch_register_on_entry.
4891 (ix86_adjust_stack_and_probe): Similarly.
4892 (ix86_emit_probe_stack_range): Similarly.
4893 (ix86_expand_prologue): Corresponding changes.
4895 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4897 PR target/40411
4898 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
4899 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
4901 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
4903 PR target/84112
4904 * lra-constraints.c (curr_insn_transform): Process AND in the
4905 address.
4907 2018-01-30 Jakub Jelinek <jakub@redhat.com>
4909 PR rtl-optimization/83986
4910 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
4911 dependence against last_pending_memory_flush in addition to
4912 pending_jump_insns.
4914 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
4916 PR tree-optimization/81611
4917 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
4918 copies.
4920 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
4922 PR target/83758
4923 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
4924 a reg rtx.
4926 2018-01-30 Richard Biener <rguenther@suse.de>
4927 Jakub Jelinek <jakub@redhat.com>
4929 PR tree-optimization/84111
4930 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
4931 inner loops added during recursion, as they don't have up-to-date
4932 SSA form.
4934 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4936 PR ipa/81360
4937 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
4938 (can_inline_edge_by_limits_p): ... here.
4939 (can_early_inline_edge_p, check_callers,
4940 update_caller_keys, update_callee_keys, recursive_inlining,
4941 add_new_edges_to_heap, speculation_useful_p,
4942 inline_small_functions,
4943 inline_small_functions, flatten_function,
4944 inline_to_all_callers_1): Update.
4946 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
4948 * profile-count.c (profile_count::combine_with_ipa_count): Handle
4949 zeros correctly.
4951 2018-01-30 Richard Biener <rguenther@suse.de>
4953 PR tree-optimization/83008
4954 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
4955 invariant and constant vector uses in stmts when they need
4956 more than one stmt.
4958 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4960 PR bootstrap/84017
4961 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
4962 * configure: Regenerate.
4964 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4966 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
4967 pattern.
4968 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
4969 Use gen_rtx_REG rather than gen_lowpart.
4971 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4973 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
4974 rather than 0 when creating partial subregs.
4976 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
4978 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
4979 of usage.
4981 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
4983 PR target/81550
4984 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
4985 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
4986 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
4987 flags. This restores the settings used before the 2017-07-24.
4988 Turning off pre increment/decrement/modify allows IVOPTS to
4989 optimize DF/SF loops where the index is an int.
4991 2018-01-29 Richard Biener <rguenther@suse.de>
4992 Kelvin Nilsen <kelvin@gcc.gnu.org>
4994 PR bootstrap/80867
4995 * tree-vect-stmts.c (vectorizable_call): Don't call
4996 targetm.vectorize_builtin_md_vectorized_function if callee is
4997 NULL.
4999 2018-01-22 Carl Love <cel@us.ibm.com>
5001 * doc/extend.tex: Fix typo in second arg in
5002 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
5004 2018-01-29 Richard Biener <rguenther@suse.de>
5006 PR tree-optimization/84086
5007 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
5008 (flush_ssaname_freelist): When SSA names were released reset
5009 the SCEV hash table.
5011 2018-01-29 Richard Biener <rguenther@suse.de>
5013 PR tree-optimization/84057
5014 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
5015 removed paths when removing edges.
5017 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
5019 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
5020 -mfunction-return=@var{choice}.
5022 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
5024 PR diagnostic/84034
5025 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
5026 Handle CR like TAB.
5027 (layout::print_source_line): Likewise.
5028 (test_get_line_width_without_trailing_whitespace): Add test cases.
5030 2018-01-27 Jakub Jelinek <jakub@redhat.com>
5032 PR middle-end/84040
5033 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
5034 debug insns.
5036 2018-01-26 Jim Wilson <jimw@sifive.com>
5038 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
5040 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
5041 specified.
5043 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5045 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
5046 and CMP + SUB-immediate -> SUBS.
5048 2018-01-26 Martin Sebor <msebor@redhat.com>
5050 PR tree-optimization/83896
5051 * tree-ssa-strlen.c (get_string_len): Rename...
5052 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
5053 Avoid assuming length is constant.
5054 (handle_char_store): Use HOST_WIDE_INT for string length.
5056 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
5058 PR target/81763
5059 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
5060 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
5062 2018-01-26 Richard Biener <rguenther@suse.de>
5064 PR rtl-optimization/84003
5065 * dse.c (record_store): Only record redundant stores when
5066 the earlier store aliases at least all accesses the later one does.
5068 2018-01-26 Jakub Jelinek <jakub@redhat.com>
5070 PR rtl-optimization/83985
5071 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
5072 REG_CFA_RESTORE insns.
5073 (delete_unmarked_insns): Don't ignore separate shrink wrapping
5074 REG_CFA_RESTORE insns here.
5076 PR c/83989
5077 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
5078 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
5080 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5082 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
5083 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
5084 (arc_init): Likewise.
5085 (arc_override_options): Likewise.
5086 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
5087 value.
5088 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
5089 support.
5090 * config/arc/arc.h (TARGET_DBNZ): Define.
5091 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
5092 properly set the tune attribute.
5093 (dbnz): Use TARGET_DBNZ guard.
5094 * config/arc/arc.opt (mtune): Add core3 option.
5096 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5098 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
5099 recognize new pic like addresses.
5100 (arc_delegitimize_address): Clean up.
5102 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5104 * config/arc/arc-arches.def: Option mrf16 valid for all
5105 architectures.
5106 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
5107 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
5108 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
5109 * config/arc/arc-tables.opt: Regenerate.
5110 * config/arc/arc.c (arc_conditional_register_usage): Handle
5111 reduced register file case.
5112 (arc_file_start): Set must have build attributes.
5113 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
5114 mrf16 option value.
5115 * config/arc/arc.opt (mrf16): Add new option.
5116 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
5117 * config/arc/genmultilib.awk: Handle new mrf16 option.
5118 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
5119 * config/arc/t-multilib: Regenerate.
5120 * doc/invoke.texi (ARC Options): Document mrf16 option.
5122 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5124 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
5125 * config/arc/arc.c (arc_handle_secure_attribute): New function.
5126 (arc_attribute_table): Add 'secure_call' attribute.
5127 (arc_print_operand): Print secure call operand.
5128 (arc_function_ok_for_sibcall): Don't optimize tail calls when
5129 secure.
5130 (arc_is_secure_call_p): New function. * config/arc/arc.md
5131 (call_i): Add support for sjli instruction.
5132 (call_value_i): Likewise.
5133 * config/arc/constraints.md (Csc): New constraint.
5135 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
5136 John Eric Martin <John.Martin@emmicro-us.com>
5138 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
5139 * config/arc/arc.c (_arc_jli_section): New struct.
5140 (arc_jli_section): New type.
5141 (rc_jli_sections): New static variable.
5142 (arc_handle_jli_attribute): New function.
5143 (arc_attribute_table): Add jli_always and jli_fixed attribute.
5144 (arc_file_end): New function.
5145 (TARGET_ASM_FILE_END): Define.
5146 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
5147 (arc_add_jli_section): New function.
5148 (jli_call_scan): Likewise.
5149 (arc_reorg): Call jli_call_scan.
5150 (arc_output_addsi): Remove 'S' from printing asm operand.
5151 (arc_is_jli_call_p): New function.
5152 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
5153 operand.
5154 (movhi_insn): Likewise.
5155 (movsi_insn): Likewise.
5156 (movsi_set_cc_insn): Likewise.
5157 (loadqi_update): Likewise.
5158 (load_zeroextendqisi_update): Likewise.
5159 (load_signextendqisi_update): Likewise.
5160 (loadhi_update): Likewise.
5161 (load_zeroextendhisi_update): Likewise.
5162 (load_signextendhisi_update): Likewise.
5163 (loadsi_update): Likewise.
5164 (loadsf_update): Likewise.
5165 (movsicc_insn): Likewise.
5166 (bset_insn): Likewise.
5167 (bxor_insn): Likewise.
5168 (bclr_insn): Likewise.
5169 (bmsk_insn): Likewise.
5170 (bicsi3_insn): Likewise.
5171 (cmpsi_cc_c_insn): Likewise.
5172 (movsi_ne): Likewise.
5173 (movsi_cond_exec): Likewise.
5174 (clrsbsi2): Likewise.
5175 (norm_f): Likewise.
5176 (normw): Likewise.
5177 (swap): Likewise.
5178 (divaw): Likewise.
5179 (flag): Likewise.
5180 (sr): Likewise.
5181 (kflag): Likewise.
5182 (ffs): Likewise.
5183 (ffs_f): Likewise.
5184 (fls): Likewise.
5185 (call_i): Remove 'S' asm letter, add jli instruction.
5186 (call_value_i): Likewise.
5187 * config/arc/arc.op (mjli-always): New option.
5188 * config/arc/constraints.md (Cji): New constraint.
5189 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
5190 operand.
5191 (subsf3_fpx): Likewise.
5192 (mulsf3_fpx): Likewise.
5193 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
5194 asm operand.
5195 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
5196 function attrbutes.
5197 * doc/invoke.texi (ARC): Document mjli-always option.
5199 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
5201 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
5202 avoid addition with 0 and use incw and decw where possible.
5204 2018-01-26 Richard Biener <rguenther@suse.de>
5206 PR tree-optimization/81082
5207 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
5208 association if it requires casting to unsigned.
5209 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
5210 from fold_plusminus_mult_expr to catch important cases late when
5211 range info is available.
5213 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5215 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
5216 * configure.ac (hidden_linkonce): New test.
5217 * configure: Regenerate.
5218 * config.in: Regenerate.
5220 2018-01-26 Julia Koval <julia.koval@intel.com>
5222 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
5223 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
5224 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
5225 _mm_mask_bitshuffle_epi64_mask): Fix type.
5226 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
5227 USI_FTYPE_V4DI_V4DI_USI): Remove.
5228 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
5229 __builtin_ia32_vpshufbitqmb256_mask,
5230 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
5231 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
5232 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
5234 2018-01-26 Alan Modra <amodra@gmail.com>
5236 PR target/84033
5237 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
5238 UNSPEC_VBPERMQ. Sort other unspecs.
5240 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
5242 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
5244 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
5246 PR middle-end/83055
5247 * predict.c (drop_profile): Do not push/pop cfun; update also
5248 node->count.
5249 (handle_missing_profiles): Fix logic looking for zero profiles.
5251 2018-01-25 Jakub Jelinek <jakub@redhat.com>
5253 PR middle-end/83977
5254 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
5255 on functions with #pragma omp declare simd or functions with simd
5256 attribute.
5257 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
5258 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5259 Remove trailing \n from warning_at calls.
5261 2018-01-25 Tom de Vries <tom@codesourcery.com>
5263 PR target/84028
5264 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5265 for neutered workers.
5267 2018-01-24 Joseph Myers <joseph@codesourcery.com>
5269 PR target/68467
5270 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
5271 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
5273 2018-01-24 Jeff Law <law@redhat.com>
5275 PR target/83994
5276 * i386.c (get_probe_interval): Move to earlier point.
5277 (ix86_compute_frame_layout): If -fstack-clash-protection and
5278 the frame is larger than the probe interval, then use pushes
5279 to save registers rather than reg->mem moves.
5280 (ix86_expand_prologue): Remove conditional for int_registers_saved
5281 assertion.
5283 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
5285 PR target/84014
5286 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
5287 min/max for never referenced object.
5289 2018-01-24 Jakub Jelinek <jakub@redhat.com>
5291 PR middle-end/83977
5292 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
5293 here.
5294 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
5295 attributes from DECL_ATTRIBUTES (decl) without affecting
5296 DECL_ATTRIBUTES (current_function_decl).
5297 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
5298 functions with non-NULL DECL_ABSTRACT_ORIGIN.
5300 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
5302 PR tree-optimization/83979
5303 * fold-const.c (fold_comparison): Use constant_boolean_node
5304 instead of boolean_{true,false}_node.
5306 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
5308 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
5309 with zero counts.
5311 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5313 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5314 Simplify the clause that sets the length attribute.
5315 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5316 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
5317 clause that sets the length attribute.
5318 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5320 2018-01-24 Tom de Vries <tom@codesourcery.com>
5322 PR target/83589
5323 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
5324 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
5325 Add strict parameter.
5326 (prevent_branch_around_nothing): Insert dummy insn between branch to
5327 label and label with no ptx insn inbetween.
5328 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
5330 2018-01-24 Tom de Vries <tom@codesourcery.com>
5332 PR target/81352
5333 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
5334 for neutered threads in warp.
5335 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
5337 2018-01-24 Richard Biener <rguenther@suse.de>
5339 PR tree-optimization/83176
5340 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
5341 operands.
5343 2018-01-24 Richard Biener <rguenther@suse.de>
5345 PR tree-optimization/82819
5346 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
5347 code generating pluses that are no-ops in the target precision.
5349 2018-01-24 Richard Biener <rguenther@suse.de>
5351 PR middle-end/84000
5352 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
5354 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5356 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
5357 to merge probabilities.
5358 * predict.c (probably_never_executed): Also mark as cold functions
5359 with global 0 profile and guessed local profile.
5360 * profile-count.c (profile_probability::combine_with_count): New
5361 member function.
5362 * profile-count.h (profile_probability::operator*,
5363 profile_probability::operator*=, profile_probability::operator/,
5364 profile_probability::operator/=): Reduce precision to adjusted
5365 and set value to guessed on contradictory divisions.
5366 (profile_probability::combine_with_freq): Remove.
5367 (profile_probability::combine_wiht_count): Declare.
5368 (profile_count::force_nonzero):: Set to adjusted.
5369 (profile_count::probability_in):: Set quality to adjusted.
5370 * tree-ssa-tail-merge.c (replace_block_by): Use
5371 combine_with_count.
5373 2018-01-23 Andrew Waterman <andrew@sifive.com>
5374 Jim Wilson <jimw@sifive.com>
5376 * config/riscv/riscv.c (riscv_stack_boundary): New.
5377 (riscv_option_override): Set riscv_stack_boundary. Handle
5378 riscv_preferred_stack_boundary_arg.
5379 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
5380 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
5381 (STACK_BOUNDARY): Set to riscv_stack_boundary.
5382 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
5383 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
5384 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
5386 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
5388 PR target/83905
5389 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
5390 of struct ix86_frame.
5391 (ix86_expand_epilogue): Likewise. Add a local variable for
5392 the reg_save_offset field in struct ix86_frame.
5394 2018-01-23 Bin Cheng <bin.cheng@arm.com>
5396 PR tree-optimization/82604
5397 * tree-loop-distribution.c (enum partition_kind): New enum item
5398 PKIND_PARTIAL_MEMSET.
5399 (partition_builtin_p): Support above new enum item.
5400 (generate_code_for_partition): Ditto.
5401 (compute_access_range): Differentiate cases that equality can be
5402 proven at all loops, the innermost loops or no loops.
5403 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
5404 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
5405 (finalize_partitions, distribute_loop): Don't fuse partition of
5406 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
5407 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
5408 parloop is enabled.
5410 2018-01-23 Martin Liska <mliska@suse.cz>
5412 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
5413 order to ignore the predictor.
5414 (PRED_POLYMORPHIC_CALL): Likewise.
5415 (PRED_RECURSIVE_CALL): Likewise.
5417 2018-01-23 Martin Liska <mliska@suse.cz>
5419 * tree-profile.c (tree_profiling): Print function header to
5420 aware reader which function we are working on.
5421 * value-prof.c (gimple_find_values_to_profile): Do not print
5422 not interesting value histograms.
5424 2018-01-23 Martin Liska <mliska@suse.cz>
5426 * profile-count.h (enum profile_quality): Add
5427 profile_uninitialized as the first value. Do not number values
5428 as they are zero based.
5429 (profile_count::verify): Update sanity check.
5430 (profile_probability::verify): Likewise.
5432 2018-01-23 Nathan Sidwell <nathan@acm.org>
5434 * doc/invoke.texi (ffor-scope): Deprecate.
5436 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5438 PR tree-optimization/83510
5439 * domwalk.c (set_all_edges_as_executable): New function.
5440 (dom_walker::dom_walker): Convert bool param
5441 "skip_unreachable_blocks" to enum reachability. Move setup of
5442 edge flags to set_all_edges_as_executable and only do it when
5443 reachability is REACHABLE_BLOCKS.
5444 * domwalk.h (enum dom_walker::reachability): New enum.
5445 (dom_walker::dom_walker): Convert bool param
5446 "skip_unreachable_blocks" to enum reachability.
5447 (set_all_edges_as_executable): New decl.
5448 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
5449 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
5450 "reachability".
5451 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
5452 but converting true to REACHABLE_BLOCKS.
5453 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
5454 * tree-vrp.c
5455 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
5456 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
5457 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
5458 REACHABLE_BLOCKS.
5459 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
5460 if check_all_array_refs will be called.
5462 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5464 * tree.c (selftest::test_location_wrappers): Add more test
5465 coverage.
5467 2018-01-23 David Malcolm <dmalcolm@redhat.com>
5469 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
5470 (selftest::test_bit_in_range): Likewise.
5472 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5474 PR testsuite/83888
5475 * doc/sourcebuild.texi (vect_float): Say that the selector
5476 only describes the situation when -funsafe-math-optimizations is on.
5477 (vect_float_strict): Document.
5479 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
5481 PR tree-optimization/83965
5482 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
5483 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
5484 instead of checking only for a reduction.
5485 (vect_recog_widen_sum_pattern): Likewise.
5487 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5489 * predict.c (probably_never_executed): Only use precise profile info.
5490 (compute_function_frequency): Skip after inlining hack since we now
5491 have quality checking.
5493 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
5495 * profile-count.h (profile_probability::very_unlikely,
5496 profile_probability::unlikely, profile_probability::even): Set
5497 precision to guessed.
5499 2018-01-23 Richard Biener <rguenther@suse.de>
5501 PR tree-optimization/83963
5502 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
5503 Properly terminate dominator walk when crossing the exit edge not
5504 when visiting its source block.
5506 2018-01-23 Jakub Jelinek <jakub@redhat.com>
5508 PR c++/83918
5509 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
5510 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
5512 2018-01-22 Jakub Jelinek <jakub@redhat.com>
5514 PR tree-optimization/83957
5515 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
5516 semicolon after for body surrounded by braces.
5518 PR tree-optimization/83081
5519 * profile-count.h (profile_probability::split): New method.
5520 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
5521 Use profile_probability::split.
5522 (do_compare_rtx_and_jump): Fix adjustment of probabilities
5523 when splitting a single conditional jump into 2.
5525 2018-01-22 David Malcolm <dmalcolm@redhat.com>
5527 PR tree-optimization/69452
5528 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
5529 decl.
5531 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5533 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
5534 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
5535 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
5537 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5539 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
5540 * config/rl78/rl78.md (movdi): New define_expand.
5541 * config/rl78/rl78.c (rl78_split_movdi): New function.
5543 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
5545 PR target/83862
5546 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
5547 no longer used.
5548 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
5549 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
5550 128-bit to produce an UNSPEC move to get the double word with the
5551 signbit and then a shift directly to do signbit.
5552 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
5553 implementation with a new version that just does either a direct
5554 move or a regular move. Move memory interface to separate insns.
5555 Move insns so they are next to the expander.
5556 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
5557 with signbit move. Split big and little endian case.
5558 (signbit<mode>2_dm_mem_le): Likewise.
5559 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
5560 (signbit<mode>2_dm2): Likewise.
5562 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5564 * config/rl78/rl78.md (anddi3): New define_expand.
5566 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5568 * config/rl78/rl78.md (umindi3): New define_expand.
5570 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5572 * config/rl78/rl78.md (smindi3): New define_expand.
5574 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5576 * config/rl78/rl78.md (smaxdi3): New define_expand.
5578 2018-01-22 Carl Love <cel@us.ibm.com>
5580 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
5581 LVX_V1TI): Add macro expansion.
5582 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
5583 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
5584 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
5585 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
5586 Change check to determine if the instruction is a byte reversing
5587 entry. Fix typo in comment.
5588 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
5589 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
5590 Add def_builtin calls for new builtins.
5591 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
5592 Add define_insn expansion.
5594 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5596 * config/rl78/rl78.md (umaxdi3): New define_expand.
5598 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
5600 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
5601 for non-QImode registers.
5603 2018-01-22 Richard Biener <rguenther@suse.de>
5605 PR tree-optimization/83963
5606 * graphite-scop-detection.c (scop_detection::get_sese): Delay
5607 including the loop exit block.
5608 (scop_detection::merge_sese): Likewise.
5609 (scop_detection::add_scop): Do it here instead.
5611 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5613 * doc/sourcebuild.texi (arm_softfloat): Document.
5615 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
5617 PR gcc/77734
5618 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
5619 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
5620 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
5622 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5623 David Edelsohn <dje.gcc@gmail.com>
5625 PR target/83946
5626 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5627 Change "crset eq" to "crset 2".
5628 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5629 (*call_indirect_aix<mode>_nospec): Likewise.
5630 (*call_value_indirect_aix<mode>_nospec): Likewise.
5631 (*call_indirect_elfv2<mode>_nospec): Likewise.
5632 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
5633 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
5634 change assembly output from . to $.
5635 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5636 (indirect_jump<mode>_nospec): Change assembly output from . to $.
5637 (*tablejump<mode>_internal1_nospec): Likewise.
5639 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
5641 PR target/80870
5642 * config/sh/sh_optimize_sett_clrt.cc:
5643 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
5645 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
5647 PR tree-optimization/83940
5648 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
5649 offset_dt to vect_constant_def rather than vect_unknown_def_type.
5650 (vect_check_load_store_mask): Add a mask_dt_out parameter and
5651 use it to pass back the definition type.
5652 (vect_check_store_rhs): Likewise rhs_dt_out.
5653 (vect_build_gather_load_calls): Add a mask_dt argument and use
5654 it instead of a call to vect_is_simple_use.
5655 (vectorizable_store): Update calls to vect_check_load_store_mask
5656 and vect_check_store_rhs. Use the dt returned by the latter instead
5657 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
5658 instead of calls to vect_is_simple_use. Pass the scalar rather
5659 than the vector operand to vect_is_simple_use when handling
5660 second and subsequent copies of an rhs value.
5661 (vectorizable_load): Update calls to vect_check_load_store_mask
5662 and vect_build_gather_load_calls. Use the cached mask_dt and
5663 gs_info.offset_dt instead of calls to vect_is_simple_use.
5665 2018-01-20 Jakub Jelinek <jakub@redhat.com>
5667 PR middle-end/83945
5668 * tree-emutls.c: Include gimplify.h.
5669 (lower_emutls_2): New function.
5670 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
5671 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
5672 it before further processing.
5674 PR target/83930
5675 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
5676 UINTVAL (trueop1) instead of INTVAL (op1).
5678 2018-01-19 Jakub Jelinek <jakub@redhat.com>
5680 PR debug/81570
5681 PR debug/83728
5682 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
5683 INCOMING_FRAME_SP_OFFSET if not defined.
5684 (scan_trace): Add ENTRY argument. If true and
5685 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
5686 emit a note to adjust the CFA offset.
5687 (create_cfi_notes): Adjust scan_trace callers.
5688 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
5689 INCOMING_FRAME_SP_OFFSET in the CIE.
5690 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
5691 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
5692 Likewise.
5693 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
5694 * doc/tm.texi: Regenerated.
5696 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
5698 PR rtl-optimization/83147
5699 * lra-constraints.c (remove_inheritance_pseudos): Use
5700 lra_substitute_pseudo_within_insn.
5702 2018-01-19 Tom de Vries <tom@codesourcery.com>
5703 Cesar Philippidis <cesar@codesourcery.com>
5705 PR target/83920
5706 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
5708 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
5710 PR target/83790
5711 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
5712 spaces for function labels.
5714 2018-01-19 Martin Liska <mliska@suse.cz>
5716 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
5717 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
5718 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
5719 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
5720 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
5721 (PRED_CONST_RETURN): Change from 69 to 65.
5722 (PRED_NULL_RETURN): Change from 91 to 71.
5723 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
5724 (PRED_LOOP_GUARD): Change from 66 to 73.
5726 2018-01-19 Martin Liska <mliska@suse.cz>
5728 * predict.c (predict_insn_def): Add new assert.
5729 (struct branch_predictor): Change type to signed integer.
5730 (test_prediction_value_range): Amend test to cover
5731 PROB_UNINITIALIZED.
5732 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
5733 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
5734 (PRED_LOOP_ITERATIONS_MAX): Likewise.
5735 (PRED_LOOP_IV_COMPARE): Likewise.
5736 * predict.h (PROB_UNINITIALIZED): Define new constant.
5738 2018-01-19 Martin Liska <mliska@suse.cz>
5740 * predict.c (dump_prediction): Add new format for
5741 analyze_brprob.py script which is enabled with -details
5742 suboption.
5743 * profile-count.h (precise_p): New function.
5745 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5747 PR tree-optimization/83922
5748 * tree-vect-loop.c (vect_verify_full_masking): Return false if
5749 there are no statements that need masking.
5750 (vect_active_double_reduction_p): New function.
5751 (vect_analyze_loop_operations): Use it when handling phis that
5752 are not in the loop header.
5754 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
5756 PR tree-optimization/83914
5757 * tree-vect-loop.c (vectorizable_induction): Don't convert
5758 init_expr or apply the peeling adjustment for inductions
5759 that are nested within the vectorized loop.
5761 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5763 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
5764 instead of NEG.
5766 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5768 PR sanitizer/81715
5769 PR testsuite/83882
5770 * function.h (gimplify_parameters): Add gimple_seq * argument.
5771 * function.c: Include gimple.h and options.h.
5772 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
5773 for the added local temporaries if needed.
5774 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
5775 if there are any parameter cleanups, wrap whole body into a
5776 try/finally with the cleanups.
5778 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
5780 PR target/82964
5781 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
5782 Use GET_MODE_CLASS for scalar floating point.
5784 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5786 PR ipa/82256
5787 patch by PaX Team
5788 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
5789 Fix call of call_cgraph_insertion_hooks.
5791 2018-01-18 Martin Sebor <msebor@redhat.com>
5793 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
5795 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
5797 PR ipa/83619
5798 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
5799 frequencies.
5801 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
5803 PR other/70268
5804 * common.opt: (-ffile-prefix-map): New option.
5805 * opts.c (common_handle_option): Defer it.
5806 * opts-global.c (handle_common_deferred_options): Handle it.
5807 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
5808 * file-prefix-map.h: New file.
5809 (remap_debug_filename, add_debug_prefix_map): ...here.
5810 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
5811 * final.c (debug_prefix_map, add_debug_prefix_map
5812 remap_debug_filename): Move to...
5813 * file-prefix-map.c: New file.
5814 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
5815 generalize, get rid of alloca(), use strrchr() instead of strchr().
5816 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
5817 Implement in terms of add_prefix_map().
5818 (remap_macro_filename, remap_debug_filename): Implement in term of
5819 remap_filename().
5820 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
5821 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
5822 * dbxout.c: Include file-prefix-map.h.
5823 * varasm.c: Likewise.
5824 * vmsdbgout.c: Likewise.
5825 * xcoffout.c: Likewise.
5826 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
5827 * doc/cppopts.texi (-fmacro-prefix-map): Document.
5828 * doc/invoke.texi (-ffile-prefix-map): Document.
5829 (-fdebug-prefix-map): Update description.
5831 2018-01-18 Martin Liska <mliska@suse.cz>
5833 * config/i386/i386.c (indirect_thunk_name): Document that also
5834 lfence is emitted.
5835 (output_indirect_thunk): Document why both instructions
5836 (pause and lfence) are generated.
5838 2018-01-18 Richard Biener <rguenther@suse.de>
5840 PR tree-optimization/83887
5841 * graphite-scop-detection.c
5842 (scop_detection::get_nearest_dom_with_single_entry): Remove.
5843 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
5844 (scop_detection::merge_sese): Re-implement with a flood-fill
5845 algorithm that properly finds a SESE region if it exists.
5847 2018-01-18 Jakub Jelinek <jakub@redhat.com>
5849 PR c/61240
5850 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
5851 pointer_diff optimizations use view_convert instead of convert.
5853 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5855 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
5856 Generate different code for -mno-speculate-indirect-jumps.
5857 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
5858 (*call_indirect_aix<mode>): Disable for
5859 -mno-speculate-indirect-jumps.
5860 (*call_indirect_aix<mode>_nospec): New define_insn.
5861 (*call_value_indirect_aix<mode>): Disable for
5862 -mno-speculate-indirect-jumps.
5863 (*call_value_indirect_aix<mode>_nospec): New define_insn.
5864 (*sibcall_nonlocal_sysv<mode>): Generate different code for
5865 -mno-speculate-indirect-jumps.
5866 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
5868 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5870 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
5871 long double type, set the flags for noting the default long double
5872 type, even if we don't pass or return a long double type.
5874 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
5876 PR ipa/83051
5877 * ipa-inline.c (flatten_function): Do not overwrite final inlining
5878 failure.
5880 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
5882 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
5883 support for merge[hl].
5884 (fold_mergehl_helper): New helper function.
5885 (tree-vector-builder.h): New #include for tree_vector_builder usage.
5886 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
5887 (altivec_vmrglw_direct): Add xxmrglw insn.
5889 2018-01-17 Andrew Waterman <andrew@sifive.com>
5891 * config/riscv/riscv.c (riscv_conditional_register_usage): If
5892 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
5894 2018-01-17 David Malcolm <dmalcolm@redhat.com>
5896 PR lto/83121
5897 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
5898 call the lto_location_cache before reading the
5899 DECL_SOURCE_LOCATION of the types.
5901 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
5902 Richard Sandiford <richard.sandiford@linaro.org>
5904 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
5905 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
5906 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
5907 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
5908 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
5909 Add declaration.
5910 * config/aarch64/constraints.md (aarch64_movti_operand):
5911 Limit immediates.
5912 * config/aarch64/predicates.md (Uti): Add new constraint.
5914 2018-01-17 Carl Love <cel@us.ibm.com>
5916 * config/rs6000/vsx.md (define_expand xl_len_r,
5917 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
5918 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
5919 lxvll.
5920 (define_expand, define_insn): Move the shift left from the
5921 define_insn to the define_expand for lxvl and stxvl instructions.
5922 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
5923 and XL_LEN_R definitions to PURE.
5925 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
5927 * config/i386/i386.c (indirect_thunk_name): Declare regno
5928 as unsigned int. Compare regno with INVALID_REGNUM.
5929 (output_indirect_thunk): Ditto.
5930 (output_indirect_thunk_function): Ditto.
5931 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
5932 in the call to output_indirect_thunk_function.
5934 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
5936 PR middle-end/83884
5937 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
5938 rather than the size of inner_type to determine the stack slot size
5939 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
5941 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
5943 PR target/83546
5944 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
5945 to PTA_SILVERMONT.
5947 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
5949 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
5950 endian Linux systems to optionally enable multilibs for selecting
5951 the long double type if the user configured an explicit type.
5952 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
5953 have no long double multilibs if not defined.
5954 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
5955 warn if the user used -mabi={ieee,ibm}longdouble and we built
5956 multilibs for long double.
5957 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
5958 appropriate multilib option.
5959 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
5960 multilib options.
5961 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
5962 for building long double multilibs.
5963 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
5965 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
5967 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
5968 copies.
5970 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
5971 64 bits.
5972 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
5973 128 bits.
5975 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
5976 variables.
5978 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
5979 return value.
5981 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5983 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
5984 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
5986 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
5988 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
5989 different rtl trees depending on TARGET_64BIT.
5990 (rs6000_gen_lvx): Likewise.
5992 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
5994 * config/visium/visium.md (nop): Tweak comment.
5995 (hazard_nop): Likewise.
5997 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5999 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
6000 -mspeculate-indirect-jumps.
6001 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
6002 for -mno-speculate-indirect-jumps.
6003 (*call_indirect_elfv2<mode>_nospec): New define_insn.
6004 (*call_value_indirect_elfv2<mode>): Disable for
6005 -mno-speculate-indirect-jumps.
6006 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
6007 (indirect_jump): Emit different RTL for
6008 -mno-speculate-indirect-jumps.
6009 (*indirect_jump<mode>): Disable for
6010 -mno-speculate-indirect-jumps.
6011 (*indirect_jump<mode>_nospec): New define_insn.
6012 (tablejump): Emit different RTL for
6013 -mno-speculate-indirect-jumps.
6014 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
6015 (tablejumpsi_nospec): New define_expand.
6016 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
6017 (tablejumpdi_nospec): New define_expand.
6018 (*tablejump<mode>_internal1): Disable for
6019 -mno-speculate-indirect-jumps.
6020 (*tablejump<mode>_internal1_nospec): New define_insn.
6021 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
6022 option.
6024 2018-01-16 Artyom Skrobov tyomitch@gmail.com
6026 * caller-save.c (insert_save): Drop unnecessary parameter. All
6027 callers updated.
6029 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6030 Richard Biener <rguenth@suse.de>
6032 PR libgomp/83590
6033 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
6034 return early, inline manually is_gimple_sizepos. Make sure if we
6035 call gimplify_expr we don't end up with a gimple constant.
6036 * tree.c (variably_modified_type_p): Don't return true for
6037 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
6038 * gimplify.h (is_gimple_sizepos): Remove.
6040 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6042 PR tree-optimization/83857
6043 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
6044 vectorizable_live_operation for pure SLP statements.
6045 (vectorizable_live_operation): Handle PHIs.
6047 2018-01-16 Richard Biener <rguenther@suse.de>
6049 PR tree-optimization/83867
6050 * tree-vect-stmts.c (vect_transform_stmt): Precompute
6051 nested_in_vect_loop_p since the scalar stmt may get invalidated.
6053 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6055 PR c/83844
6056 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
6057 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
6058 If off is not INTEGER_CST, issue a may not be aligned warning
6059 rather than isn't aligned. Use isn%'t rather than isn't.
6060 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
6061 into MULT_EXPR.
6062 <case MULT_EXPR>: Improve the case when bottom and one of the
6063 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
6064 operand, in that case check if the other operand is multiple of
6065 bottom divided by the INTEGER_CST operand.
6067 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6069 PR target/83858
6070 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
6071 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
6072 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
6073 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
6074 * config/pa/pa.c (pa_function_arg_advance): Likewise.
6075 (pa_function_arg, pa_arg_partial_bytes): Likewise.
6076 (pa_function_arg_size): New function.
6078 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6080 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
6081 in a separate statement.
6083 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
6085 PR tree-optimization/83847
6086 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
6087 group gathers and scatters.
6089 2018-01-16 Jakub Jelinek <jakub@redhat.com>
6091 PR rtl-optimization/86620
6092 * params.def (max-sched-ready-insns): Bump minimum value to 1.
6094 PR rtl-optimization/83213
6095 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
6096 to last if both are JUMP_INSNs.
6098 PR tree-optimization/83843
6099 * gimple-ssa-store-merging.c
6100 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
6101 store_immediate_info for bswap/nop orig_stores.
6103 2018-01-15 Andrew Waterman <andrew@sifive.com>
6105 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
6106 !TARGET_MUL.
6107 <UDIV>: Increase cost if !TARGET_DIV.
6109 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
6111 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
6112 (define_attr "cr_logical_3op"): New.
6113 (cceq_ior_compare): Adjust.
6114 (cceq_ior_compare_complement): Adjust.
6115 (*cceq_rev_compare): Adjust.
6116 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
6117 (is_cracked_insn): Adjust.
6118 (insn_must_be_first_in_group): Adjust.
6119 * config/rs6000/40x.md: Adjust.
6120 * config/rs6000/440.md: Adjust.
6121 * config/rs6000/476.md: Adjust.
6122 * config/rs6000/601.md: Adjust.
6123 * config/rs6000/603.md: Adjust.
6124 * config/rs6000/6xx.md: Adjust.
6125 * config/rs6000/7450.md: Adjust.
6126 * config/rs6000/7xx.md: Adjust.
6127 * config/rs6000/8540.md: Adjust.
6128 * config/rs6000/cell.md: Adjust.
6129 * config/rs6000/e300c2c3.md: Adjust.
6130 * config/rs6000/e500mc.md: Adjust.
6131 * config/rs6000/e500mc64.md: Adjust.
6132 * config/rs6000/e5500.md: Adjust.
6133 * config/rs6000/e6500.md: Adjust.
6134 * config/rs6000/mpc.md: Adjust.
6135 * config/rs6000/power4.md: Adjust.
6136 * config/rs6000/power5.md: Adjust.
6137 * config/rs6000/power6.md: Adjust.
6138 * config/rs6000/power7.md: Adjust.
6139 * config/rs6000/power8.md: Adjust.
6140 * config/rs6000/power9.md: Adjust.
6141 * config/rs6000/rs64.md: Adjust.
6142 * config/rs6000/titan.md: Adjust.
6144 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6146 * config/i386/predicates.md (indirect_branch_operand): Rewrite
6147 ix86_indirect_branch_register logic.
6149 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6151 * config/i386/constraints.md (Bs): Update
6152 ix86_indirect_branch_register check. Don't check
6153 ix86_indirect_branch_register with GOT_memory_operand.
6154 (Bw): Likewise.
6155 * config/i386/predicates.md (GOT_memory_operand): Don't check
6156 ix86_indirect_branch_register here.
6157 (GOT32_symbol_operand): Likewise.
6159 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6161 * config/i386/predicates.md (constant_call_address_operand):
6162 Rewrite ix86_indirect_branch_register logic.
6163 (sibcall_insn_operand): Likewise.
6165 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6167 * config/i386/constraints.md (Bs): Replace
6168 ix86_indirect_branch_thunk_register with
6169 ix86_indirect_branch_register.
6170 (Bw): Likewise.
6171 * config/i386/i386.md (indirect_jump): Likewise.
6172 (tablejump): Likewise.
6173 (*sibcall_memory): Likewise.
6174 (*sibcall_value_memory): Likewise.
6175 Peepholes of indirect call and jump via memory: Likewise.
6176 * config/i386/i386.opt: Likewise.
6177 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6178 (GOT_memory_operand): Likewise.
6179 (call_insn_operand): Likewise.
6180 (sibcall_insn_operand): Likewise.
6181 (GOT32_symbol_operand): Likewise.
6183 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6185 PR middle-end/83837
6186 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
6187 type rather than type addr's type points to.
6188 (expand_omp_atomic_mutex): Likewise.
6189 (expand_omp_atomic): Likewise.
6191 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
6193 PR target/83839
6194 * config/i386/i386.c (output_indirect_thunk_function): Use
6195 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
6196 for __x86_return_thunk.
6198 2018-01-15 Richard Biener <rguenther@suse.de>
6200 PR middle-end/83850
6201 * expmed.c (extract_bit_field_1): Fix typo.
6203 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6205 PR target/83687
6206 * config/arm/iterators.md (VF): New mode iterator.
6207 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
6208 Remove integer-related logic from pattern.
6209 (neon_vabd<mode>_3): Likewise.
6211 2018-01-15 Jakub Jelinek <jakub@redhat.com>
6213 PR middle-end/82694
6214 * common.opt (fstrict-overflow): No longer an alias.
6215 (fwrapv-pointer): New option.
6216 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
6217 also for pointer types based on flag_wrapv_pointer.
6218 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
6219 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
6220 opts->x_flag_wrapv got set.
6221 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
6222 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
6223 POINTER_TYPE_OVERFLOW_UNDEFINED.
6224 * match.pd: Likewise in address comparison pattern.
6225 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
6227 2018-01-15 Richard Biener <rguenther@suse.de>
6229 PR lto/83804
6230 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
6231 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
6232 Reset type names to their identifier if their TYPE_DECL doesn't
6233 have linkage (and thus is used for ODR and devirt).
6234 (save_debug_info_for_decl): Remove.
6235 (save_debug_info_for_type): Likewise.
6236 (add_tree_to_fld_list): Adjust.
6237 * tree-pretty-print.c (dump_generic_node): Make dumping of
6238 type names more robust.
6240 2018-01-15 Richard Biener <rguenther@suse.de>
6242 * BASE-VER: Bump to 8.0.1.
6244 2018-01-14 Martin Sebor <msebor@redhat.com>
6246 PR other/83508
6247 * builtins.c (check_access): Avoid warning when the no-warning bit
6248 is set.
6250 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
6252 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
6253 * ira-color (allocno_hard_regs_compare): Likewise.
6255 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
6257 PR target/83013
6258 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
6259 Use .pushsection/.popsection.
6261 2018-01-14 Martin Sebor <msebor@redhat.com>
6263 PR c++/81327
6264 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
6266 2018-01-14 Jakub Jelinek <jakub@redhat.com>
6268 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
6269 entry from extra_headers.
6270 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
6271 extra_headers, make the list bitwise identical to the i?86-*-* one.
6273 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6275 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
6276 -mcmodel=large with -mindirect-branch=thunk,
6277 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
6278 -mfunction-return=thunk-extern.
6279 * doc/invoke.texi: Document -mcmodel=large is incompatible with
6280 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
6281 -mfunction-return=thunk and -mfunction-return=thunk-extern.
6283 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6285 * config/i386/i386.c (print_reg): Print the name of the full
6286 integer register without '%'.
6287 (ix86_print_operand): Handle 'V'.
6288 * doc/extend.texi: Document 'V' modifier.
6290 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6292 * config/i386/constraints.md (Bs): Disallow memory operand for
6293 -mindirect-branch-register.
6294 (Bw): Likewise.
6295 * config/i386/predicates.md (indirect_branch_operand): Likewise.
6296 (GOT_memory_operand): Likewise.
6297 (call_insn_operand): Likewise.
6298 (sibcall_insn_operand): Likewise.
6299 (GOT32_symbol_operand): Likewise.
6300 * config/i386/i386.md (indirect_jump): Call convert_memory_address
6301 for -mindirect-branch-register.
6302 (tablejump): Likewise.
6303 (*sibcall_memory): Likewise.
6304 (*sibcall_value_memory): Likewise.
6305 Disallow peepholes of indirect call and jump via memory for
6306 -mindirect-branch-register.
6307 (*call_pop): Replace m with Bw.
6308 (*call_value_pop): Likewise.
6309 (*sibcall_pop_memory): Replace m with Bs.
6310 * config/i386/i386.opt (mindirect-branch-register): New option.
6311 * doc/invoke.texi: Document -mindirect-branch-register option.
6313 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6315 * config/i386/i386-protos.h (ix86_output_function_return): New.
6316 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
6317 set function_return_type.
6318 (indirect_thunk_name): Add ret_p to indicate thunk for function
6319 return.
6320 (output_indirect_thunk_function): Pass false to
6321 indirect_thunk_name.
6322 (ix86_output_indirect_branch_via_reg): Likewise.
6323 (ix86_output_indirect_branch_via_push): Likewise.
6324 (output_indirect_thunk_function): Create alias for function
6325 return thunk if regno < 0.
6326 (ix86_output_function_return): New function.
6327 (ix86_handle_fndecl_attribute): Handle function_return.
6328 (ix86_attribute_table): Add function_return.
6329 * config/i386/i386.h (machine_function): Add
6330 function_return_type.
6331 * config/i386/i386.md (simple_return_internal): Use
6332 ix86_output_function_return.
6333 (simple_return_internal_long): Likewise.
6334 * config/i386/i386.opt (mfunction-return=): New option.
6335 (indirect_branch): Mention -mfunction-return=.
6336 * doc/extend.texi: Document function_return function attribute.
6337 * doc/invoke.texi: Document -mfunction-return= option.
6339 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
6341 * config/i386/i386-opts.h (indirect_branch): New.
6342 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
6343 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
6344 with local indirect jump when converting indirect call and jump.
6345 (ix86_set_indirect_branch_type): New.
6346 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
6347 (indirectlabelno): New.
6348 (indirect_thunk_needed): Likewise.
6349 (indirect_thunk_bnd_needed): Likewise.
6350 (indirect_thunks_used): Likewise.
6351 (indirect_thunks_bnd_used): Likewise.
6352 (INDIRECT_LABEL): Likewise.
6353 (indirect_thunk_name): Likewise.
6354 (output_indirect_thunk): Likewise.
6355 (output_indirect_thunk_function): Likewise.
6356 (ix86_output_indirect_branch_via_reg): Likewise.
6357 (ix86_output_indirect_branch_via_push): Likewise.
6358 (ix86_output_indirect_branch): Likewise.
6359 (ix86_output_indirect_jmp): Likewise.
6360 (ix86_code_end): Call output_indirect_thunk_function if needed.
6361 (ix86_output_call_insn): Call ix86_output_indirect_branch if
6362 needed.
6363 (ix86_handle_fndecl_attribute): Handle indirect_branch.
6364 (ix86_attribute_table): Add indirect_branch.
6365 * config/i386/i386.h (machine_function): Add indirect_branch_type
6366 and has_local_indirect_jump.
6367 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
6368 to true.
6369 (tablejump): Likewise.
6370 (*indirect_jump): Use ix86_output_indirect_jmp.
6371 (*tablejump_1): Likewise.
6372 (simple_return_indirect_internal): Likewise.
6373 * config/i386/i386.opt (mindirect-branch=): New option.
6374 (indirect_branch): New.
6375 (keep): Likewise.
6376 (thunk): Likewise.
6377 (thunk-inline): Likewise.
6378 (thunk-extern): Likewise.
6379 * doc/extend.texi: Document indirect_branch function attribute.
6380 * doc/invoke.texi: Document -mindirect-branch= option.
6382 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
6384 PR ipa/83051
6385 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
6387 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
6389 * ipa-inline.c (want_inline_small_function_p): Return false if
6390 inlining has already failed with CIF_FINAL_ERROR.
6391 (update_caller_keys): Call want_inline_small_function_p before
6392 can_inline_edge_p.
6393 (update_callee_keys): Likewise.
6395 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6397 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
6398 New function.
6399 (rs6000_quadword_masked_address_p): Likewise.
6400 (quad_aligned_load_p): Likewise.
6401 (quad_aligned_store_p): Likewise.
6402 (const_load_sequence_p): Add comment to describe the outer-most loop.
6403 (mimic_memory_attributes_and_flags): New function.
6404 (rs6000_gen_stvx): Likewise.
6405 (replace_swapped_aligned_store): Likewise.
6406 (rs6000_gen_lvx): Likewise.
6407 (replace_swapped_aligned_load): Likewise.
6408 (replace_swapped_load_constant): Capitalize argument name in
6409 comment describing this function.
6410 (rs6000_analyze_swaps): Add a third pass to search for vector loads
6411 and stores that access quad-word aligned addresses and replace
6412 with stvx or lvx instructions when appropriate.
6413 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
6414 New function prototype.
6415 (rs6000_quadword_masked_address_p): Likewise.
6416 (rs6000_gen_lvx): Likewise.
6417 (rs6000_gen_stvx): Likewise.
6418 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
6419 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
6420 when memory address is aligned.
6421 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
6422 this split to select lvx instruction when memory address is aligned.
6423 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
6424 instruction when memory address is aligned.
6425 (*vsx_le_perm_load_v16qi): Likewise.
6426 (four unnamed splitters): Modify to select the stvx instruction
6427 when memory is aligned.
6429 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
6431 * predict.c (determine_unlikely_bbs): Handle correctly BBs
6432 which appears in the queue multiple times.
6434 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6435 Alan Hayward <alan.hayward@arm.com>
6436 David Sherwood <david.sherwood@arm.com>
6438 * tree-vectorizer.h (vec_lower_bound): New structure.
6439 (_loop_vec_info): Add check_nonzero and lower_bounds.
6440 (LOOP_VINFO_CHECK_NONZERO): New macro.
6441 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
6442 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
6443 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
6444 fields. Make seg_len the distance travelled, not including the
6445 access size.
6446 (dr_direction_indicator): Declare.
6447 (dr_zero_step_indicator): Likewise.
6448 (dr_known_forward_stride_p): Likewise.
6449 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
6450 tree-ssanames.h.
6451 (runtime_alias_check_p): Allow runtime alias checks with
6452 variable strides.
6453 (operator ==): Compare access_size and align.
6454 (prune_runtime_alias_test_list): Rework for new distinction between
6455 the access_size and seg_len.
6456 (create_intersect_range_checks_index): Likewise. Cope with polynomial
6457 segment lengths.
6458 (get_segment_min_max): New function.
6459 (create_intersect_range_checks): Use it.
6460 (dr_step_indicator): New function.
6461 (dr_direction_indicator): Likewise.
6462 (dr_zero_step_indicator): Likewise.
6463 (dr_known_forward_stride_p): Likewise.
6464 * tree-loop-distribution.c (data_ref_segment_size): Return
6465 DR_STEP * (niters - 1).
6466 (compute_alias_check_pairs): Update call to the dr_with_seg_len
6467 constructor.
6468 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
6469 (vect_preserves_scalar_order_p): New function, split out from...
6470 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
6471 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
6472 (vect_vfa_access_size): New function.
6473 (vect_vfa_align): Likewise.
6474 (vect_compile_time_alias): Take access_size_a and access_b arguments.
6475 (dump_lower_bound): New function.
6476 (vect_check_lower_bound): Likewise.
6477 (vect_small_gap_p): Likewise.
6478 (vectorizable_with_step_bound_p): Likewise.
6479 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
6480 depencies if the vectorization factor is 1. Convert the checks
6481 for nonzero steps into checks on the bounds of DR_STEP. Try using
6482 a bunds check for variable steps if the minimum required step is
6483 relatively small. Update calls to the dr_with_seg_len
6484 constructor and to vect_compile_time_alias.
6485 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
6486 function.
6487 (vect_loop_versioning): Call it.
6488 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
6489 when retrying.
6490 (vect_estimate_min_profitable_iters): Account for any bounds checks.
6492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6493 Alan Hayward <alan.hayward@arm.com>
6494 David Sherwood <david.sherwood@arm.com>
6496 * doc/sourcebuild.texi (vect_scatter_store): Document.
6497 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
6498 optabs.
6499 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
6500 Document.
6501 * genopinit.c (main): Add supports_vec_scatter_store and
6502 supports_vec_scatter_store_cached to target_optabs.
6503 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
6504 IFN_MASK_SCATTER_STORE.
6505 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
6506 functions.
6507 * internal-fn.h (internal_store_fn_p): Declare.
6508 (internal_fn_stored_value_index): Likewise.
6509 * internal-fn.c (scatter_store_direct): New macro.
6510 (expand_scatter_store_optab_fn): New function.
6511 (direct_scatter_store_optab_supported_p): New macro.
6512 (internal_store_fn_p): New function.
6513 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
6514 IFN_MASK_SCATTER_STORE.
6515 (internal_fn_mask_index): Likewise.
6516 (internal_fn_stored_value_index): New function.
6517 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
6518 for scatter stores.
6519 * optabs-query.h (supports_vec_scatter_store_p): Declare.
6520 * optabs-query.c (supports_vec_scatter_store_p): New function.
6521 * tree-vectorizer.h (vect_get_store_rhs): Declare.
6522 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
6523 true for scatter stores.
6524 (vect_gather_scatter_fn_p): Handle scatter stores too.
6525 (vect_check_gather_scatter): Consider using scatter stores if
6526 supports_vec_scatter_store_p.
6527 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
6528 scatter stores too.
6529 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6530 internal_fn_stored_value_index.
6531 (check_load_store_masking): Handle scatter stores too.
6532 (vect_get_store_rhs): Make public.
6533 (vectorizable_call): Use internal_store_fn_p.
6534 (vectorizable_store): Handle scatter store internal functions.
6535 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
6536 when deciding whether the end of the group has been reached.
6537 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
6538 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
6539 (mask_scatter_store<mode>): New insns.
6541 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6542 Alan Hayward <alan.hayward@arm.com>
6543 David Sherwood <david.sherwood@arm.com>
6545 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
6546 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
6547 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
6548 function.
6549 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
6550 Use vect_truncate_gather_scatter_offset if we can't treat the
6551 operation as a normal gather load or scatter store.
6552 (get_group_load_store_type): Take the gather_scatter_info
6553 as argument. Try using a gather load or scatter store for
6554 single-element groups.
6555 (get_load_store_type): Update calls to get_group_load_store_type
6556 and vect_use_strided_gather_scatters_p.
6558 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6559 Alan Hayward <alan.hayward@arm.com>
6560 David Sherwood <david.sherwood@arm.com>
6562 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
6563 optional tree argument.
6564 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
6565 null target hooks.
6566 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
6567 but continue to use the current value as a fallback.
6568 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
6569 to compare the updates.
6570 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
6571 (get_load_store_type): Use it when handling a strided access.
6572 (vect_get_strided_load_store_ops): New function.
6573 (vect_get_data_ptr_increment): Likewise.
6574 (vectorizable_load): Handle strided gather loads. Always pass
6575 a step to vect_create_data_ref_ptr and bump_vector_ptr.
6577 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6578 Alan Hayward <alan.hayward@arm.com>
6579 David Sherwood <david.sherwood@arm.com>
6581 * doc/md.texi (gather_load@var{m}): Document.
6582 (mask_gather_load@var{m}): Likewise.
6583 * genopinit.c (main): Add supports_vec_gather_load and
6584 supports_vec_gather_load_cached to target_optabs.
6585 * optabs-tree.c (init_tree_optimization_optabs): Use
6586 ggc_cleared_alloc to allocate target_optabs.
6587 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
6588 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
6589 functions.
6590 * internal-fn.h (internal_load_fn_p): Declare.
6591 (internal_gather_scatter_fn_p): Likewise.
6592 (internal_fn_mask_index): Likewise.
6593 (internal_gather_scatter_fn_supported_p): Likewise.
6594 * internal-fn.c (gather_load_direct): New macro.
6595 (expand_gather_load_optab_fn): New function.
6596 (direct_gather_load_optab_supported_p): New macro.
6597 (direct_internal_fn_optab): New function.
6598 (internal_load_fn_p): Likewise.
6599 (internal_gather_scatter_fn_p): Likewise.
6600 (internal_fn_mask_index): Likewise.
6601 (internal_gather_scatter_fn_supported_p): Likewise.
6602 * optabs-query.c (supports_at_least_one_mode_p): New function.
6603 (supports_vec_gather_load_p): Likewise.
6604 * optabs-query.h (supports_vec_gather_load_p): Declare.
6605 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
6606 and memory_type field.
6607 (NUM_PATTERNS): Bump to 15.
6608 * tree-vect-data-refs.c: Include internal-fn.h.
6609 (vect_gather_scatter_fn_p): New function.
6610 (vect_describe_gather_scatter_call): Likewise.
6611 (vect_check_gather_scatter): Try using internal functions for
6612 gather loads. Recognize existing calls to a gather load function.
6613 (vect_analyze_data_refs): Consider using gather loads if
6614 supports_vec_gather_load_p.
6615 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
6616 (vect_get_gather_scatter_offset_type): Likewise.
6617 (vect_convert_mask_for_vectype): Likewise.
6618 (vect_add_conversion_to_patterm): Likewise.
6619 (vect_try_gather_scatter_pattern): Likewise.
6620 (vect_recog_gather_scatter_pattern): New pattern recognizer.
6621 (vect_vect_recog_func_ptrs): Add it.
6622 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
6623 internal_fn_mask_index and internal_gather_scatter_fn_p.
6624 (check_load_store_masking): Take the gather_scatter_info as an
6625 argument and handle gather loads.
6626 (vect_get_gather_scatter_ops): New function.
6627 (vectorizable_call): Check internal_load_fn_p.
6628 (vectorizable_load): Likewise. Handle gather load internal
6629 functions.
6630 (vectorizable_store): Update call to check_load_store_masking.
6631 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
6632 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
6633 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
6634 (aarch64_gather_scale_operand_d): New predicates.
6635 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
6636 (mask_gather_load<mode>): New insns.
6638 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6639 Alan Hayward <alan.hayward@arm.com>
6640 David Sherwood <david.sherwood@arm.com>
6642 * optabs.def (fold_left_plus_optab): New optab.
6643 * doc/md.texi (fold_left_plus_@var{m}): Document.
6644 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
6645 * internal-fn.c (fold_left_direct): Define.
6646 (expand_fold_left_optab_fn): Likewise.
6647 (direct_fold_left_optab_supported_p): Likewise.
6648 * fold-const-call.c (fold_const_fold_left): New function.
6649 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
6650 * tree-parloops.c (valid_reduction_p): New function.
6651 (gather_scalar_reductions): Use it.
6652 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
6653 (vect_finish_replace_stmt): Declare.
6654 * tree-vect-loop.c (fold_left_reduction_fn): New function.
6655 (needs_fold_left_reduction_p): New function, split out from...
6656 (vect_is_simple_reduction): ...here. Accept reductions that
6657 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
6658 (vect_force_simple_reduction): Also store the reduction type in
6659 the assignment's STMT_VINFO_REDUC_TYPE.
6660 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
6661 (merge_with_identity): New function.
6662 (vect_expand_fold_left): Likewise.
6663 (vectorize_fold_left_reduction): Likewise.
6664 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
6665 scalar phi in place for it. Check for target support and reject
6666 cases that would reassociate the operation. Defer the transform
6667 phase to vectorize_fold_left_reduction.
6668 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
6669 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
6670 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
6672 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6674 * tree-if-conv.c (predicate_mem_writes): Remove redundant
6675 call to ifc_temp_var.
6677 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6678 Alan Hayward <alan.hayward@arm.com>
6679 David Sherwood <david.sherwood@arm.com>
6681 * target.def (legitimize_address_displacement): Take the original
6682 offset as a poly_int.
6683 * targhooks.h (default_legitimize_address_displacement): Update
6684 accordingly.
6685 * targhooks.c (default_legitimize_address_displacement): Likewise.
6686 * doc/tm.texi: Regenerate.
6687 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
6688 as an argument, moving assert of ad->disp == ad->disp_term to...
6689 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
6690 Try calling targetm.legitimize_address_displacement before expanding
6691 the address rather than afterwards, and adjust for the new interface.
6692 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
6693 Match the new hook interface. Handle SVE addresses.
6694 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
6695 new hook interface.
6697 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6699 * Makefile.in (OBJS): Add early-remat.o.
6700 * target.def (select_early_remat_modes): New hook.
6701 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
6702 * doc/tm.texi: Regenerate.
6703 * targhooks.h (default_select_early_remat_modes): Declare.
6704 * targhooks.c (default_select_early_remat_modes): New function.
6705 * timevar.def (TV_EARLY_REMAT): New timevar.
6706 * passes.def (pass_early_remat): New pass.
6707 * tree-pass.h (make_pass_early_remat): Declare.
6708 * early-remat.c: New file.
6709 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
6710 function.
6711 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
6713 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6714 Alan Hayward <alan.hayward@arm.com>
6715 David Sherwood <david.sherwood@arm.com>
6717 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
6718 vfm1 with a bound_epilog parameter.
6719 (vect_do_peeling): Update calls accordingly, and move the prologue
6720 call earlier in the function. Treat the base bound_epilog as 0 for
6721 fully-masked loops and retain vf - 1 for other loops. Add 1 to
6722 this base when peeling for gaps.
6723 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
6724 with fully-masked loops.
6725 (vect_estimate_min_profitable_iters): Handle the single peeled
6726 iteration in that case.
6728 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6729 Alan Hayward <alan.hayward@arm.com>
6730 David Sherwood <david.sherwood@arm.com>
6732 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
6733 single-element interleaving even if the size is not a power of 2.
6734 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
6735 accesses for single-element interleaving if the group size is
6736 not a power of 2.
6738 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6739 Alan Hayward <alan.hayward@arm.com>
6740 David Sherwood <david.sherwood@arm.com>
6742 * doc/md.texi (fold_extract_last_@var{m}): Document.
6743 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
6744 * optabs.def (fold_extract_last_optab): New optab.
6745 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
6746 * internal-fn.c (fold_extract_direct): New macro.
6747 (expand_fold_extract_optab_fn): Likewise.
6748 (direct_fold_extract_optab_supported_p): Likewise.
6749 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
6750 * tree-vect-loop.c (vect_model_reduction_cost): Handle
6751 EXTRACT_LAST_REDUCTION.
6752 (get_initial_def_for_reduction): Do not create an initial vector
6753 for EXTRACT_LAST_REDUCTION reductions.
6754 (vectorizable_reduction): Leave the scalar phi in place for
6755 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
6756 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
6757 epilogue code for EXTRACT_LAST_REDUCTION and defer the
6758 transform phase to vectorizable_condition.
6759 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
6760 split out from...
6761 (vect_finish_stmt_generation): ...here.
6762 (vect_finish_replace_stmt): New function.
6763 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
6764 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
6765 pattern.
6766 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
6768 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6769 Alan Hayward <alan.hayward@arm.com>
6770 David Sherwood <david.sherwood@arm.com>
6772 * doc/md.texi (extract_last_@var{m}): Document.
6773 * optabs.def (extract_last_optab): New optab.
6774 * internal-fn.def (EXTRACT_LAST): New internal function.
6775 * internal-fn.c (cond_unary_direct): New macro.
6776 (expand_cond_unary_optab_fn): Likewise.
6777 (direct_cond_unary_optab_supported_p): Likewise.
6778 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
6779 loops using EXTRACT_LAST.
6780 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
6781 (extract_last_<mode>): ...this optab.
6782 (vec_extract<mode><Vel>): Update accordingly.
6784 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6785 Alan Hayward <alan.hayward@arm.com>
6786 David Sherwood <david.sherwood@arm.com>
6788 * target.def (empty_mask_is_expensive): New hook.
6789 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
6790 * doc/tm.texi: Regenerate.
6791 * targhooks.h (default_empty_mask_is_expensive): Declare.
6792 * targhooks.c (default_empty_mask_is_expensive): New function.
6793 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
6794 if the target says that empty masks are expensive.
6795 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
6796 New function.
6797 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
6799 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6800 Alan Hayward <alan.hayward@arm.com>
6801 David Sherwood <david.sherwood@arm.com>
6803 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
6804 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
6805 (vect_use_loop_mask_for_alignment_p): New function.
6806 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
6807 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
6808 niters_skip argument. Make sure that the first niters_skip elements
6809 of the first iteration are inactive.
6810 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
6811 Update call to vect_set_loop_masks_directly.
6812 (get_misalign_in_elems): New function, split out from...
6813 (vect_gen_prolog_loop_niters): ...here.
6814 (vect_update_init_of_dr): Take a code argument that specifies whether
6815 the adjustment should be added or subtracted.
6816 (vect_update_init_of_drs): Likewise.
6817 (vect_prepare_for_masked_peels): New function.
6818 (vect_do_peeling): Skip prologue peeling if we're using a mask
6819 instead. Update call to vect_update_inits_of_drs.
6820 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6821 mask_skip_niters.
6822 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
6823 alignment. Do not include the number of peeled iterations in
6824 the minimum threshold in that case.
6825 (vectorizable_induction): Adjust the start value down by
6826 LOOP_VINFO_MASK_SKIP_NITERS iterations.
6827 (vect_transform_loop): Call vect_prepare_for_masked_peels.
6828 Take the number of skipped iterations into account when calculating
6829 the loop bounds.
6830 * tree-vect-stmts.c (vect_gen_while_not): New function.
6832 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6833 Alan Hayward <alan.hayward@arm.com>
6834 David Sherwood <david.sherwood@arm.com>
6836 * doc/sourcebuild.texi (vect_fully_masked): Document.
6837 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
6838 default value to 0.
6839 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
6840 split out from...
6841 (vect_analyze_loop_2): ...here. Don't check the vectorization
6842 factor against the number of loop iterations if the loop is
6843 fully-masked.
6845 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6846 Alan Hayward <alan.hayward@arm.com>
6847 David Sherwood <david.sherwood@arm.com>
6849 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
6850 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
6851 (dump_groups): Update accordingly.
6852 (iv_use::mem_type): New member variable.
6853 (address_p): New function.
6854 (record_use): Add a mem_type argument and initialize the new
6855 mem_type field.
6856 (record_group_use): Add a mem_type argument. Use address_p.
6857 Remove obsolete null checks of base_object. Update call to record_use.
6858 (find_interesting_uses_op): Update call to record_group_use.
6859 (find_interesting_uses_cond): Likewise.
6860 (find_interesting_uses_address): Likewise.
6861 (get_mem_type_for_internal_fn): New function.
6862 (find_address_like_use): Likewise.
6863 (find_interesting_uses_stmt): Try find_address_like_use before
6864 calling find_interesting_uses_op.
6865 (addr_offset_valid_p): Use the iv mem_type field as the type
6866 of the addressed memory.
6867 (add_autoinc_candidates): Likewise.
6868 (get_address_cost): Likewise.
6869 (split_small_address_groups_p): Use address_p.
6870 (split_address_groups): Likewise.
6871 (add_iv_candidate_for_use): Likewise.
6872 (autoinc_possible_for_pair): Likewise.
6873 (rewrite_groups): Likewise.
6874 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
6875 (determine_group_iv_cost): Update after split of USE_ADDRESS.
6876 (get_alias_ptr_type_for_ptr_address): New function.
6877 (rewrite_use_address): Rewrite address uses in calls that were
6878 identified by find_address_like_use.
6880 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6881 Alan Hayward <alan.hayward@arm.com>
6882 David Sherwood <david.sherwood@arm.com>
6884 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
6885 TARGET_MEM_REFs.
6886 * gimple-expr.h (is_gimple_addressable: Likewise.
6887 * gimple-expr.c (is_gimple_address): Likewise.
6888 * internal-fn.c (expand_call_mem_ref): New function.
6889 (expand_mask_load_optab_fn): Use it.
6890 (expand_mask_store_optab_fn): Likewise.
6892 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6893 Alan Hayward <alan.hayward@arm.com>
6894 David Sherwood <david.sherwood@arm.com>
6896 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
6897 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
6898 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
6899 (cond_umax@var{mode}): Document.
6900 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
6901 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
6902 (cond_umin_optab, cond_umax_optab): New optabs.
6903 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
6904 (COND_IOR, COND_XOR): New internal functions.
6905 * internal-fn.h (get_conditional_internal_fn): Declare.
6906 * internal-fn.c (cond_binary_direct): New macro.
6907 (expand_cond_binary_optab_fn): Likewise.
6908 (direct_cond_binary_optab_supported_p): Likewise.
6909 (get_conditional_internal_fn): New function.
6910 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
6911 Cope with reduction statements that are vectorized as calls rather
6912 than assignments.
6913 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
6914 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
6915 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
6916 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
6917 (UNSPEC_COND_EOR): New unspecs.
6918 (optab): Add mappings for them.
6919 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
6920 (sve_int_op, sve_fp_op): New int attributes.
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 * optabs.def (while_ult_optab): New optab.
6927 * doc/md.texi (while_ult@var{m}@var{n}): Document.
6928 * internal-fn.def (WHILE_ULT): New internal function.
6929 * internal-fn.h (direct_internal_fn_supported_p): New override
6930 that takes two types as argument.
6931 * internal-fn.c (while_direct): New macro.
6932 (expand_while_optab_fn): New function.
6933 (convert_optab_supported_p): Likewise.
6934 (direct_while_optab_supported_p): New macro.
6935 * wide-int.h (wi::udiv_ceil): New function.
6936 * tree-vectorizer.h (rgroup_masks): New structure.
6937 (vec_loop_masks): New typedef.
6938 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
6939 and fully_masked_p.
6940 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
6941 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
6942 (vect_max_vf): New function.
6943 (slpeel_make_loop_iterate_ntimes): Delete.
6944 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
6945 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
6946 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
6947 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
6948 internal-fn.h, stor-layout.h and optabs-query.h.
6949 (vect_set_loop_mask): New function.
6950 (add_preheader_seq): Likewise.
6951 (add_header_seq): Likewise.
6952 (interleave_supported_p): Likewise.
6953 (vect_maybe_permute_loop_masks): Likewise.
6954 (vect_set_loop_masks_directly): Likewise.
6955 (vect_set_loop_condition_masked): Likewise.
6956 (vect_set_loop_condition_unmasked): New function, split out from
6957 slpeel_make_loop_iterate_ntimes.
6958 (slpeel_make_loop_iterate_ntimes): Rename to..
6959 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
6960 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
6961 (vect_do_peeling): Update call accordingly.
6962 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
6963 loops.
6964 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
6965 mask_compare_type, can_fully_mask_p and fully_masked_p.
6966 (release_vec_loop_masks): New function.
6967 (_loop_vec_info): Use it to free the loop masks.
6968 (can_produce_all_loop_masks_p): New function.
6969 (vect_get_max_nscalars_per_iter): Likewise.
6970 (vect_verify_full_masking): Likewise.
6971 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
6972 retries, and free the mask rgroups before retrying. Check loop-wide
6973 reasons for disallowing fully-masked loops. Make the final decision
6974 about whether use a fully-masked loop or not.
6975 (vect_estimate_min_profitable_iters): Do not assume that peeling
6976 for the number of iterations will be needed for fully-masked loops.
6977 (vectorizable_reduction): Disable fully-masked loops.
6978 (vectorizable_live_operation): Likewise.
6979 (vect_halve_mask_nunits): New function.
6980 (vect_double_mask_nunits): Likewise.
6981 (vect_record_loop_mask): Likewise.
6982 (vect_get_loop_mask): Likewise.
6983 (vect_transform_loop): Handle the case in which the final loop
6984 iteration might handle a partial vector. Call vect_set_loop_condition
6985 instead of slpeel_make_loop_iterate_ntimes.
6986 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
6987 (check_load_store_masking): New function.
6988 (prepare_load_store_mask): Likewise.
6989 (vectorizable_store): Handle fully-masked loops.
6990 (vectorizable_load): Likewise.
6991 (supportable_widening_operation): Use vect_halve_mask_nunits for
6992 booleans.
6993 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
6994 (vect_gen_while): New function.
6995 * config/aarch64/aarch64.md (umax<mode>3): New expander.
6996 (aarch64_uqdec<mode>): New insn.
6998 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
6999 Alan Hayward <alan.hayward@arm.com>
7000 David Sherwood <david.sherwood@arm.com>
7002 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
7003 (reduc_xor_scal_optab): New optabs.
7004 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
7005 (reduc_xor_scal_@var{m}): Document.
7006 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
7007 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
7008 internal functions.
7009 * fold-const-call.c (fold_const_call): Handle them.
7010 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
7011 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
7012 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
7013 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
7014 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
7015 (UNSPEC_XORV): New unspecs.
7016 (optab): Add entries for them.
7017 (BITWISEV): New int iterator.
7018 (bit_reduc_op): New int attributes.
7020 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7021 Alan Hayward <alan.hayward@arm.com>
7022 David Sherwood <david.sherwood@arm.com>
7024 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
7025 * internal-fn.def (VEC_SHL_INSERT): New internal function.
7026 * optabs.def (vec_shl_insert_optab): New optab.
7027 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
7028 (duplicate_and_interleave): Likewise.
7029 * tree-vect-loop.c: Include internal-fn.h.
7030 (neutral_op_for_slp_reduction): New function, split out from
7031 get_initial_defs_for_reduction.
7032 (get_initial_def_for_reduction): Handle option 2 for variable-length
7033 vectors by loading the neutral value into a vector and then shifting
7034 the initial value into element 0.
7035 (get_initial_defs_for_reduction): Replace the code argument with
7036 the neutral value calculated by neutral_op_for_slp_reduction.
7037 Use gimple_build_vector for constant-length vectors.
7038 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
7039 but the first group_size elements have a neutral value.
7040 Use duplicate_and_interleave otherwise.
7041 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
7042 Update call to get_initial_defs_for_reduction. Handle SLP
7043 reductions for variable-length vectors by creating one vector
7044 result for each scalar result, with the elements associated
7045 with other scalar results stubbed out with the neutral value.
7046 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
7047 Require IFN_VEC_SHL_INSERT for double reductions on
7048 variable-length vectors, or SLP reductions that have
7049 a neutral value. Require can_duplicate_and_interleave_p
7050 support for variable-length unchained SLP reductions if there
7051 is no neutral value, such as for MIN/MAX reductions. Also require
7052 the number of vector elements to be a multiple of the number of
7053 SLP statements when doing variable-length unchained SLP reductions.
7054 Update call to vect_create_epilog_for_reduction.
7055 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
7056 and remove initial values.
7057 (duplicate_and_interleave): Make public.
7058 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
7059 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
7061 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7062 Alan Hayward <alan.hayward@arm.com>
7063 David Sherwood <david.sherwood@arm.com>
7065 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
7066 (can_duplicate_and_interleave_p): New function.
7067 (vect_get_and_check_slp_defs): Take the vector of statements
7068 rather than just the current one. Remove excess parentheses.
7069 Restriction rejectinon of vect_constant_def and vect_external_def
7070 for variable-length vectors to boolean types, or types for which
7071 can_duplicate_and_interleave_p is false.
7072 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
7073 (duplicate_and_interleave): New function.
7074 (vect_get_constant_vectors): Use gimple_build_vector for
7075 constant-length vectors and suitable variable-length constant
7076 vectors. Use duplicate_and_interleave for other variable-length
7077 vectors. Don't defer the update when inserting new statements.
7079 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7080 Alan Hayward <alan.hayward@arm.com>
7081 David Sherwood <david.sherwood@arm.com>
7083 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
7084 min_profitable_iters doesn't go negative.
7086 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7087 Alan Hayward <alan.hayward@arm.com>
7088 David Sherwood <david.sherwood@arm.com>
7090 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
7091 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
7092 * optabs.def (vec_mask_load_lanes_optab): New optab.
7093 (vec_mask_store_lanes_optab): Likewise.
7094 * internal-fn.def (MASK_LOAD_LANES): New internal function.
7095 (MASK_STORE_LANES): Likewise.
7096 * internal-fn.c (mask_load_lanes_direct): New macro.
7097 (mask_store_lanes_direct): Likewise.
7098 (expand_mask_load_optab_fn): Handle masked operations.
7099 (expand_mask_load_lanes_optab_fn): New macro.
7100 (expand_mask_store_optab_fn): Handle masked operations.
7101 (expand_mask_store_lanes_optab_fn): New macro.
7102 (direct_mask_load_lanes_optab_supported_p): Likewise.
7103 (direct_mask_store_lanes_optab_supported_p): Likewise.
7104 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
7105 parameter.
7106 (vect_load_lanes_supported): Likewise.
7107 * tree-vect-data-refs.c (strip_conversion): New function.
7108 (can_group_stmts_p): Likewise.
7109 (vect_analyze_data_ref_accesses): Use it instead of checking
7110 for a pair of assignments.
7111 (vect_store_lanes_supported): Take a masked_p parameter.
7112 (vect_load_lanes_supported): Likewise.
7113 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
7114 vect_store_lanes_supported and vect_load_lanes_supported.
7115 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
7116 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
7117 parameter. Don't allow gaps for masked accesses.
7118 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
7119 and vect_load_lanes_supported.
7120 (get_load_store_type): Take a masked_p parameter and update
7121 call to get_group_load_store_type.
7122 (vectorizable_store): Update call to get_load_store_type.
7123 Handle IFN_MASK_STORE_LANES.
7124 (vectorizable_load): Update call to get_load_store_type.
7125 Handle IFN_MASK_LOAD_LANES.
7127 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7128 Alan Hayward <alan.hayward@arm.com>
7129 David Sherwood <david.sherwood@arm.com>
7131 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
7132 modes for SVE.
7133 * config/aarch64/aarch64-protos.h
7134 (aarch64_sve_struct_memory_operand_p): Declare.
7135 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
7136 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
7137 (VPRED, vpred): Handle SVE structure modes.
7138 * config/aarch64/constraints.md (Utx): New constraint.
7139 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
7140 (aarch64_sve_struct_nonimmediate_operand): New predicates.
7141 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
7142 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
7143 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
7144 structure modes. Split into pieces after RA.
7145 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
7146 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
7147 New patterns.
7148 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
7149 SVE structure modes.
7150 (aarch64_classify_address): Likewise.
7151 (sizetochar): Move earlier in file.
7152 (aarch64_print_operand): Handle SVE register lists.
7153 (aarch64_array_mode): New function.
7154 (aarch64_sve_struct_memory_operand_p): Likewise.
7155 (TARGET_ARRAY_MODE): Redefine.
7157 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7158 Alan Hayward <alan.hayward@arm.com>
7159 David Sherwood <david.sherwood@arm.com>
7161 * target.def (array_mode): New target hook.
7162 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
7163 * doc/tm.texi: Regenerate.
7164 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
7165 * hooks.c (hook_optmode_mode_uhwi_none): New function.
7166 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
7167 targetm.array_mode.
7168 * stor-layout.c (mode_for_array): Likewise. Support polynomial
7169 type sizes.
7171 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7172 Alan Hayward <alan.hayward@arm.com>
7173 David Sherwood <david.sherwood@arm.com>
7175 * fold-const.c (fold_binary_loc): Check the argument types
7176 rather than the result type when testing for a vector operation.
7178 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7180 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
7181 * doc/tm.texi: Regenerate.
7183 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7184 Alan Hayward <alan.hayward@arm.com>
7185 David Sherwood <david.sherwood@arm.com>
7187 * doc/invoke.texi (-msve-vector-bits=): Document new option.
7188 (sve): Document new AArch64 extension.
7189 * doc/md.texi (w): Extend the description of the AArch64
7190 constraint to include SVE vectors.
7191 (Upl, Upa): Document new AArch64 predicate constraints.
7192 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
7193 enum.
7194 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
7195 (msve-vector-bits=): New option.
7196 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
7197 SVE when these are disabled.
7198 (sve): New extension.
7199 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
7200 modes. Adjust their number of units based on aarch64_sve_vg.
7201 (MAX_BITSIZE_MODE_ANY_MODE): Define.
7202 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
7203 aarch64_addr_query_type.
7204 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
7205 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
7206 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
7207 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
7208 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
7209 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
7210 (aarch64_simd_imm_zero_p): Delete.
7211 (aarch64_check_zero_based_sve_index_immediate): Declare.
7212 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7213 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7214 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7215 (aarch64_sve_float_mul_immediate_p): Likewise.
7216 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7217 rather than an rtx.
7218 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
7219 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
7220 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
7221 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
7222 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
7223 (aarch64_regmode_natural_size): Likewise.
7224 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
7225 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
7226 left one place.
7227 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
7228 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
7229 for VG and the SVE predicate registers.
7230 (V_ALIASES): Add a "z"-prefixed alias.
7231 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
7232 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
7233 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
7234 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
7235 (REG_CLASS_NAMES): Add entries for them.
7236 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
7237 and the predicate registers.
7238 (aarch64_sve_vg): Declare.
7239 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
7240 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
7241 (REGMODE_NATURAL_SIZE): Define.
7242 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
7243 SVE macros.
7244 * config/aarch64/aarch64.c: Include cfgrtl.h.
7245 (simd_immediate_info): Add a constructor for series vectors,
7246 and an associated step field.
7247 (aarch64_sve_vg): New variable.
7248 (aarch64_dbx_register_number): Handle VG and the predicate registers.
7249 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
7250 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
7251 (VEC_ANY_DATA, VEC_STRUCT): New constants.
7252 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
7253 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
7254 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
7255 (aarch64_get_mask_mode): New functions.
7256 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
7257 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
7258 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
7259 predicate modes and predicate registers. Explicitly restrict
7260 GPRs to modes of 16 bytes or smaller. Only allow FP registers
7261 to store a vector mode if it is recognized by
7262 aarch64_classify_vector_mode.
7263 (aarch64_regmode_natural_size): New function.
7264 (aarch64_hard_regno_caller_save_mode): Return the original mode
7265 for predicates.
7266 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
7267 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
7268 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
7269 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
7270 functions.
7271 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
7272 does not overlap dest if the function is frame-related. Handle
7273 SVE constants.
7274 (aarch64_split_add_offset): New function.
7275 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
7276 them aarch64_add_offset.
7277 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
7278 and update call to aarch64_sub_sp.
7279 (aarch64_add_cfa_expression): New function.
7280 (aarch64_expand_prologue): Pass extra temporary registers to the
7281 functions above. Handle the case in which we need to emit new
7282 DW_CFA_expressions for registers that were originally saved
7283 relative to the stack pointer, but now have to be expressed
7284 relative to the frame pointer.
7285 (aarch64_output_mi_thunk): Pass extra temporary registers to the
7286 functions above.
7287 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
7288 IP0 and IP1 values for SVE frames.
7289 (aarch64_expand_vec_series): New function.
7290 (aarch64_expand_sve_widened_duplicate): Likewise.
7291 (aarch64_expand_sve_const_vector): Likewise.
7292 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
7293 Handle SVE constants. Use emit_move_insn to move a force_const_mem
7294 into the register, rather than emitting a SET directly.
7295 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
7296 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
7297 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
7298 (offset_9bit_signed_scaled_p): New functions.
7299 (aarch64_replicate_bitmask_imm): New function.
7300 (aarch64_bitmask_imm): Use it.
7301 (aarch64_cannot_force_const_mem): Reject expressions involving
7302 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
7303 (aarch64_classify_index): Handle SVE indices, by requiring
7304 a plain register index with a scale that matches the element size.
7305 (aarch64_classify_address): Handle SVE addresses. Assert that
7306 the mode of the address is VOIDmode or an integer mode.
7307 Update call to aarch64_classify_symbol.
7308 (aarch64_classify_symbolic_expression): Update call to
7309 aarch64_classify_symbol.
7310 (aarch64_const_vec_all_in_range_p): New function.
7311 (aarch64_print_vector_float_operand): Likewise.
7312 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
7313 "vN" for FP registers with SVE modes. Handle (const ...) vectors
7314 and the FP immediates 1.0 and 0.5.
7315 (aarch64_print_address_internal): Handle SVE addresses.
7316 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
7317 (aarch64_regno_regclass): Handle predicate registers.
7318 (aarch64_secondary_reload): Handle big-endian reloads of SVE
7319 data modes.
7320 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
7321 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
7322 (aarch64_convert_sve_vector_bits): New function.
7323 (aarch64_override_options): Use it to handle -msve-vector-bits=.
7324 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
7325 rather than an rtx.
7326 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
7327 Handle SVE vector and predicate modes. Accept VL-based constants
7328 that need only one temporary register, and VL offsets that require
7329 no temporary registers.
7330 (aarch64_conditional_register_usage): Mark the predicate registers
7331 as fixed if SVE isn't available.
7332 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
7333 Return true for SVE vector and predicate modes.
7334 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
7335 rather than an unsigned int. Handle SVE modes.
7336 (aarch64_preferred_simd_mode): Update call accordingly. Handle
7337 SVE modes.
7338 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
7339 if SVE is enabled.
7340 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
7341 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
7342 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
7343 (aarch64_sve_float_mul_immediate_p): New functions.
7344 (aarch64_sve_valid_immediate): New function.
7345 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
7346 Explicitly reject structure modes. Check for INDEX constants.
7347 Handle PTRUE and PFALSE constants.
7348 (aarch64_check_zero_based_sve_index_immediate): New function.
7349 (aarch64_simd_imm_zero_p): Delete.
7350 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
7351 vector modes. Accept constants in the range of CNT[BHWD].
7352 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
7353 ask for an Advanced SIMD mode.
7354 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
7355 (aarch64_simd_vector_alignment): Handle SVE predicates.
7356 (aarch64_vectorize_preferred_vector_alignment): New function.
7357 (aarch64_simd_vector_alignment_reachable): Use it instead of
7358 the vector size.
7359 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
7360 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
7361 functions.
7362 (MAX_VECT_LEN): Delete.
7363 (expand_vec_perm_d): Add a vec_flags field.
7364 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
7365 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
7366 (aarch64_evpc_ext): Don't apply a big-endian lane correction
7367 for SVE modes.
7368 (aarch64_evpc_rev): Rename to...
7369 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
7370 (aarch64_evpc_rev_global): New function.
7371 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
7372 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
7373 MAX_VECT_LEN.
7374 (aarch64_evpc_sve_tbl): New function.
7375 (aarch64_expand_vec_perm_const_1): Update after rename of
7376 aarch64_evpc_rev. Handle SVE permutes too, trying
7377 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
7378 than aarch64_evpc_tbl.
7379 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
7380 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
7381 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
7382 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
7383 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
7384 (aarch64_expand_sve_vcond): New functions.
7385 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
7386 of aarch64_vector_mode_p.
7387 (aarch64_dwarf_poly_indeterminate_value): New function.
7388 (aarch64_compute_pressure_classes): Likewise.
7389 (aarch64_can_change_mode_class): Likewise.
7390 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
7391 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
7392 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
7393 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
7394 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
7395 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
7396 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
7397 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
7398 constraints.
7399 (Dn, Dl, Dr): Accept const as well as const_vector.
7400 (Dz): Likewise. Compare against CONST0_RTX.
7401 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
7402 of "vector" where appropriate.
7403 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
7404 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
7405 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
7406 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
7407 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
7408 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
7409 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
7410 (v_int_equiv): Extend to SVE modes.
7411 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
7412 mode attributes.
7413 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
7414 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
7415 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
7416 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
7417 (SVE_COND_FP_CMP): New int iterators.
7418 (perm_hilo): Handle the new unpack unspecs.
7419 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
7420 attributes.
7421 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
7422 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
7423 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
7424 (aarch64_equality_operator, aarch64_constant_vector_operand)
7425 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
7426 (aarch64_sve_nonimmediate_operand): Likewise.
7427 (aarch64_sve_general_operand): Likewise.
7428 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
7429 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
7430 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
7431 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
7432 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
7433 (aarch64_sve_float_arith_immediate): Likewise.
7434 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
7435 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
7436 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
7437 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
7438 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
7439 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
7440 (aarch64_sve_float_arith_operand): Likewise.
7441 (aarch64_sve_float_arith_with_sub_operand): Likewise.
7442 (aarch64_sve_float_mul_operand): Likewise.
7443 (aarch64_sve_vec_perm_operand): Likewise.
7444 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
7445 (aarch64_mov_operand): Accept const_poly_int and const_vector.
7446 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
7447 as well as const_vector.
7448 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
7449 in file. Use CONST0_RTX and CONSTM1_RTX.
7450 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
7451 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
7452 Use aarch64_simd_imm_zero.
7453 * config/aarch64/aarch64-sve.md: New file.
7454 * config/aarch64/aarch64.md: Include it.
7455 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
7456 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
7457 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
7458 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
7459 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
7460 (sve): New attribute.
7461 (enabled): Disable instructions with the sve attribute unless
7462 TARGET_SVE.
7463 (movqi, movhi): Pass CONST_POLY_INT operaneds through
7464 aarch64_expand_mov_immediate.
7465 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
7466 CNT[BHSD] immediates.
7467 (movti): Split CONST_POLY_INT moves into two halves.
7468 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
7469 Split additions that need a temporary here if the destination
7470 is the stack pointer.
7471 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
7472 (*add<mode>3_poly_1): New instruction.
7473 (set_clobber_cc): New expander.
7475 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7477 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
7478 parameter and use it instead of GET_MODE_SIZE (innermode). Use
7479 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
7480 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
7481 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
7482 Change innermode from fixed_mode_size to machine_mode.
7483 (simplify_subreg): Update call accordingly. Handle a constant-sized
7484 subreg of a variable-length CONST_VECTOR.
7486 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
7487 Alan Hayward <alan.hayward@arm.com>
7488 David Sherwood <david.sherwood@arm.com>
7490 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
7491 (add_offset_to_base): New function, split out from...
7492 (create_mem_ref): ...here. When handling a scale other than 1,
7493 check first whether the address is valid without the offset.
7494 Add it into the base if so, leaving the index and scale as-is.
7496 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7498 PR c++/83778
7499 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
7500 fold_for_warn before checking if arg2 is INTEGER_CST.
7502 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
7504 * config/rs6000/predicates.md (load_multiple_operation): Delete.
7505 (store_multiple_operation): Delete.
7506 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
7507 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
7508 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
7509 guarded by TARGET_STRING.
7510 (rs6000_output_load_multiple): Delete.
7511 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
7512 OPTION_MASK_STRING / TARGET_STRING handling.
7513 (print_operand) <'N', 'O'>: Add comment that these are unused now.
7514 (const rs6000_opt_masks) <"string">: Change mask to 0.
7515 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
7516 (MASK_STRING): Delete.
7517 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
7518 parts. Simplify.
7519 (load_multiple): Delete.
7520 (*ldmsi8): Delete.
7521 (*ldmsi7): Delete.
7522 (*ldmsi6): Delete.
7523 (*ldmsi5): Delete.
7524 (*ldmsi4): Delete.
7525 (*ldmsi3): Delete.
7526 (store_multiple): Delete.
7527 (*stmsi8): Delete.
7528 (*stmsi7): Delete.
7529 (*stmsi6): Delete.
7530 (*stmsi5): Delete.
7531 (*stmsi4): Delete.
7532 (*stmsi3): Delete.
7533 (movmemsi_8reg): Delete.
7534 (corresponding unnamed define_insn): Delete.
7535 (movmemsi_6reg): Delete.
7536 (corresponding unnamed define_insn): Delete.
7537 (movmemsi_4reg): Delete.
7538 (corresponding unnamed define_insn): Delete.
7539 (movmemsi_2reg): Delete.
7540 (corresponding unnamed define_insn): Delete.
7541 (movmemsi_1reg): Delete.
7542 (corresponding unnamed define_insn): Delete.
7543 * config/rs6000/rs6000.opt (mno-string): New.
7544 (mstring): Replace by deprecation warning stub.
7545 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
7547 2018-01-12 Jakub Jelinek <jakub@redhat.com>
7549 * regrename.c (regrename_do_replace): If replacing the same
7550 reg multiple times, try to reuse last created gen_raw_REG.
7552 PR debug/81155
7553 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
7554 main to workaround a bug in GDB.
7556 2018-01-12 Tom de Vries <tom@codesourcery.com>
7558 PR target/83737
7559 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
7561 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
7563 PR rtl-optimization/80481
7564 * ira-color.c (get_cap_member): New function.
7565 (allocnos_conflict_by_live_ranges_p): Use it.
7566 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
7567 (setup_slot_coalesced_allocno_live_ranges): Ditto.
7569 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
7571 PR target/83628
7572 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
7573 (*saddl_se_1): Ditto.
7574 (*ssubsi_1): Ditto.
7575 (*ssubl_se_1): Ditto.
7577 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7579 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
7580 rather than wi::to_widest for DR_INITs.
7581 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
7582 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
7583 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
7584 INTEGER_CSTs.
7585 (vect_analyze_group_access_1): Note that here.
7587 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7589 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
7590 polynomial type sizes.
7592 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
7594 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
7595 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
7596 (gimple_add_tmp_var): Likewise.
7598 2018-01-12 Martin Liska <mliska@suse.cz>
7600 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
7601 (gimple_alloc_sizes): Likewise.
7602 (dump_gimple_statistics): Use PRIu64 in printf format.
7603 * gimple.h: Change uint64_t to int.
7605 2018-01-12 Martin Liska <mliska@suse.cz>
7607 * tree-core.h: Use uint64_t instead of int.
7608 * tree.c (tree_node_counts): Likewise.
7609 (tree_node_sizes): Likewise.
7610 (dump_tree_statistics): Use PRIu64 in printf format.
7612 2018-01-12 Martin Liska <mliska@suse.cz>
7614 * Makefile.in: As qsort_chk is implemented in vec.c, add
7615 vec.o to linkage of gencfn-macros.
7616 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
7617 passing the info to record_node_allocation_statistics.
7618 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
7619 and pass the info.
7620 * ggc-common.c (struct ggc_usage): Add operator== and use
7621 it in operator< and compare function.
7622 * mem-stats.h (struct mem_usage): Likewise.
7623 * vec.c (struct vec_usage): Remove operator< and compare
7624 function. Can be simply inherited.
7626 2018-01-12 Martin Jambor <mjambor@suse.cz>
7628 PR target/81616
7629 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
7630 * tree-ssa-math-opts.c: Include domwalk.h.
7631 (convert_mult_to_fma_1): New function.
7632 (fma_transformation_info): New type.
7633 (fma_deferring_state): Likewise.
7634 (cancel_fma_deferring): New function.
7635 (result_of_phi): Likewise.
7636 (last_fma_candidate_feeds_initial_phi): Likewise.
7637 (convert_mult_to_fma): Added deferring logic, split actual
7638 transformation to convert_mult_to_fma_1.
7639 (math_opts_dom_walker): New type.
7640 (math_opts_dom_walker::after_dom_children): New method, body moved
7641 here from pass_optimize_widening_mul::execute, added deferring logic
7642 bits.
7643 (pass_optimize_widening_mul::execute): Moved most of code to
7644 math_opts_dom_walker::after_dom_children.
7645 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
7646 * config/i386/i386.c (ix86_option_override_internal): Added
7647 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
7649 2018-01-12 Richard Biener <rguenther@suse.de>
7651 PR debug/83157
7652 * dwarf2out.c (gen_variable_die): Do not reset old_die for
7653 inline instance vars.
7655 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
7657 PR target/81819
7658 * config/rx/rx.c (rx_is_restricted_memory_address):
7659 Handle SUBREG case.
7661 2018-01-12 Richard Biener <rguenther@suse.de>
7663 PR tree-optimization/80846
7664 * target.def (split_reduction): New target hook.
7665 * targhooks.c (default_split_reduction): New function.
7666 * targhooks.h (default_split_reduction): Declare.
7667 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
7668 target requests first reduce vectors by combining low and high
7669 parts.
7670 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
7671 (get_vectype_for_scalar_type_and_size): Export.
7672 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
7673 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
7674 * doc/tm.texi: Regenerate.
7675 * config/i386/i386.c (ix86_split_reduction): Implement
7676 TARGET_VECTORIZE_SPLIT_REDUCTION.
7678 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7680 PR target/83368
7681 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
7682 in PIC mode except for TARGET_VXWORKS_RTP.
7683 * config/sparc/sparc.c: Include cfgrtl.h.
7684 (TARGET_INIT_PIC_REG): Define.
7685 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
7686 (sparc_pic_register_p): New predicate.
7687 (sparc_legitimate_address_p): Use it.
7688 (sparc_legitimize_pic_address): Likewise.
7689 (sparc_delegitimize_address): Likewise.
7690 (sparc_mode_dependent_address_p): Likewise.
7691 (gen_load_pcrel_sym): Remove 4th parameter.
7692 (load_got_register): Adjust call to above. Remove obsolete stuff.
7693 (sparc_expand_prologue): Do not call load_got_register here.
7694 (sparc_flat_expand_prologue): Likewise.
7695 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
7696 (sparc_use_pseudo_pic_reg): New function.
7697 (sparc_init_pic_reg): Likewise.
7698 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
7699 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
7701 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
7703 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
7704 Add item for branch_cost.
7706 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
7708 PR rtl-optimization/83565
7709 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
7710 not extend the result to a larger mode for rotate operations.
7711 (num_sign_bit_copies1): Likewise.
7713 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7715 PR target/40411
7716 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
7717 -symbolic.
7718 Use values-Xc.o for -pedantic.
7719 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
7721 2018-01-12 Martin Liska <mliska@suse.cz>
7723 PR ipa/83054
7724 * ipa-devirt.c (final_warning_record::grow_type_warnings):
7725 New function.
7726 (possible_polymorphic_call_targets): Use it.
7727 (ipa_devirt): Likewise.
7729 2018-01-12 Martin Liska <mliska@suse.cz>
7731 * profile-count.h (enum profile_quality): Use 0 as invalid
7732 enum value of profile_quality.
7734 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
7736 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
7737 -mext-string options.
7739 2018-01-12 Richard Biener <rguenther@suse.de>
7741 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
7742 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
7743 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
7744 Likewise.
7745 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
7747 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
7749 * configure.ac (--with-long-double-format): Add support for the
7750 configuration option to change the default long double format on
7751 PowerPC systems.
7752 * config.gcc (powerpc*-linux*-*): Likewise.
7753 * configure: Regenerate.
7754 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
7755 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
7756 used without modification.
7758 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7760 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
7761 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
7762 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
7763 MISC_BUILTIN_SPEC_BARRIER.
7764 (rs6000_init_builtins): Likewise.
7765 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
7766 enum value.
7767 (speculation_barrier): New define_insn.
7768 * doc/extend.texi: Document __builtin_speculation_barrier.
7770 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7772 PR target/83203
7773 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
7774 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
7775 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
7776 iterators.
7777 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
7778 integral modes instead of "ss" and "sd".
7779 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
7780 vectors with 32-bit and 64-bit elements.
7781 (vecdupssescalarmodesuffix): New mode attribute.
7782 (vec_dup<mode>): Use it.
7784 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
7786 PR target/83330
7787 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
7788 frame if argument is passed on stack.
7790 2018-01-11 Jakub Jelinek <jakub@redhat.com>
7792 PR target/82682
7793 * ree.c (combine_reaching_defs): Optimize also
7794 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
7795 reg2=any_extend(exp); reg1=reg2;, formatting fix.
7797 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7799 PR middle-end/83189
7800 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
7802 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
7804 PR middle-end/83718
7805 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
7806 after they are computed.
7808 2018-01-11 Bin Cheng <bin.cheng@arm.com>
7810 PR tree-optimization/83695
7811 * gimple-loop-linterchange.cc
7812 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
7813 reset cached scev information after interchange.
7814 (pass_linterchange::execute): Remove call to scev_reset_htab.
7816 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7818 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
7819 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
7820 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
7821 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
7822 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
7823 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
7824 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
7825 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
7826 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
7827 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
7828 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
7829 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
7830 (V_lane_reg): Likewise.
7831 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
7832 New define_expand.
7833 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
7834 (vfmal_lane_low<mode>_intrinsic,
7835 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
7836 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
7837 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
7838 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
7839 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
7840 vfmsl_lane_high<mode>_intrinsic): New define_insns.
7842 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7844 * config/arm/arm-cpus.in (fp16fml): New feature.
7845 (ALL_SIMD): Add fp16fml.
7846 (armv8.2-a): Add fp16fml as an option.
7847 (armv8.3-a): Likewise.
7848 (armv8.4-a): Add fp16fml as part of fp16.
7849 * config/arm/arm.h (TARGET_FP16FML): Define.
7850 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
7851 when appropriate.
7852 * config/arm/arm-modes.def (V2HF): Define.
7853 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
7854 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
7855 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
7856 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
7857 vfmsl_low, vfmsl_high): New set of builtins.
7858 * config/arm/iterators.md (PLUSMINUS): New code iterator.
7859 (vfml_op): New code attribute.
7860 (VFMLHALVES): New int iterator.
7861 (VFML, VFMLSEL): New mode attributes.
7862 (V_reg): Define mapping for V2HF.
7863 (V_hi, V_lo): New mode attributes.
7864 (VF_constraint): Likewise.
7865 (vfml_half, vfml_half_selector): New int attributes.
7866 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
7867 define_expand.
7868 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
7869 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
7870 New define_insn.
7871 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
7872 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
7873 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
7874 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
7875 documentation.
7876 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
7877 Document new effective target and option set.
7879 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7881 * config/arm/arm-cpus.in (armv8_4): New feature.
7882 (ARMv8_4a): New fgroup.
7883 (armv8.4-a): New arch.
7884 * config/arm/arm-tables.opt: Regenerate.
7885 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
7886 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
7887 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
7888 Add matching rules for -march=armv8.4-a and extensions.
7889 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
7891 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
7893 PR target/81821
7894 * config/rx/rx.md (BW): New mode attribute.
7895 (sync_lock_test_and_setsi): Add mode suffix to insn output.
7897 2018-01-11 Richard Biener <rguenther@suse.de>
7899 PR tree-optimization/83435
7900 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
7901 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
7902 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
7904 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7905 Alan Hayward <alan.hayward@arm.com>
7906 David Sherwood <david.sherwood@arm.com>
7908 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
7909 field.
7910 (aarch64_classify_address): Initialize it. Track polynomial offsets.
7911 (aarch64_print_address_internal): Use it to check for a zero offset.
7913 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7914 Alan Hayward <alan.hayward@arm.com>
7915 David Sherwood <david.sherwood@arm.com>
7917 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
7918 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
7919 Return a poly_int64 rather than a HOST_WIDE_INT.
7920 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
7921 rather than a HOST_WIDE_INT.
7922 * config/aarch64/aarch64.h (aarch64_frame): Protect with
7923 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
7924 hard_fp_offset, frame_size, initial_adjust, callee_offset and
7925 final_offset from HOST_WIDE_INT to poly_int64.
7926 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
7927 to_constant when getting the number of units in an Advanced SIMD
7928 mode.
7929 (aarch64_builtin_vectorized_function): Check for a constant number
7930 of units.
7931 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
7932 GET_MODE_SIZE.
7933 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
7934 attribute instead of GET_MODE_NUNITS.
7935 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
7936 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
7937 GET_MODE_SIZE for fixed-size registers.
7938 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
7939 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
7940 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
7941 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
7942 (aarch64_print_operand, aarch64_print_address_internal)
7943 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
7944 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
7945 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
7946 Handle polynomial GET_MODE_SIZE.
7947 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
7948 wider than SImode without modification.
7949 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
7950 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
7951 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
7952 passing and returning SVE modes.
7953 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
7954 rather than GEN_INT.
7955 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
7956 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
7957 (aarch64_allocate_and_probe_stack_space): Likewise.
7958 (aarch64_layout_frame): Cope with polynomial offsets.
7959 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
7960 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
7961 polynomial offsets.
7962 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
7963 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
7964 poly_int64 rather than a HOST_WIDE_INT.
7965 (aarch64_get_separate_components, aarch64_process_components)
7966 (aarch64_expand_prologue, aarch64_expand_epilogue)
7967 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
7968 (aarch64_anchor_offset): New function, split out from...
7969 (aarch64_legitimize_address): ...here.
7970 (aarch64_builtin_vectorization_cost): Handle polynomial
7971 TYPE_VECTOR_SUBPARTS.
7972 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
7973 GET_MODE_NUNITS.
7974 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
7975 number of elements from the PARALLEL rather than the mode.
7976 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
7977 rather than GET_MODE_BITSIZE.
7978 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
7979 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
7980 (aarch64_expand_vec_perm_const_1): Handle polynomial
7981 d->perm.length () and d->perm elements.
7982 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
7983 Apply to_constant to d->perm elements.
7984 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
7985 polynomial CONST_VECTOR_NUNITS.
7986 (aarch64_move_pointer): Take amount as a poly_int64 rather
7987 than an int.
7988 (aarch64_progress_pointer): Avoid temporary variable.
7989 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
7990 the mode attribute instead of GET_MODE.
7992 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
7993 Alan Hayward <alan.hayward@arm.com>
7994 David Sherwood <david.sherwood@arm.com>
7996 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
7997 x exists before using it.
7998 (aarch64_add_constant_internal): Rename to...
7999 (aarch64_add_offset_1): ...this. Replace regnum with separate
8000 src and dest rtxes. Handle the case in which they're different,
8001 including when the offset is zero. Replace scratchreg with an rtx.
8002 Use 2 additions if there is no spare register into which we can
8003 move a 16-bit constant.
8004 (aarch64_add_constant): Delete.
8005 (aarch64_add_offset): Replace reg with separate src and dest
8006 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
8007 Use aarch64_add_offset_1.
8008 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
8009 an rtx rather than an int. Take the delta as a poly_int64
8010 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
8011 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
8012 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
8013 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
8014 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
8015 and aarch64_add_sp.
8016 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
8017 aarch64_add_constant.
8019 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8021 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
8022 Use scalar_float_mode.
8024 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8026 * config/aarch64/aarch64-simd.md
8027 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
8028 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
8029 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
8030 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
8031 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
8032 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
8033 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
8034 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
8035 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
8036 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
8038 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8040 PR target/83514
8041 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
8042 targ_options->x_arm_arch_string is non NULL.
8044 2018-01-11 Tamar Christina <tamar.christina@arm.com>
8046 * config/aarch64/aarch64.h
8047 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
8049 2018-01-11 Sudakshina Das <sudi.das@arm.com>
8051 PR target/82096
8052 * expmed.c (emit_store_flag_force): Swap if const op0
8053 and change VOIDmode to mode of op0.
8055 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
8057 PR rtl-optimization/83761
8058 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
8059 than bytes to mode_for_size.
8061 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8063 PR middle-end/83189
8064 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
8065 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
8066 profile.
8068 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8070 PR middle-end/83575
8071 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
8072 when in layout mode.
8073 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
8074 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
8075 partition fixup.
8077 2018-01-10 Michael Collison <michael.collison@arm.com>
8079 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
8080 * config/aarch64/aarch64-option-extension.def: Add
8081 AARCH64_OPT_EXTENSION of 'fp16fml'.
8082 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8083 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
8084 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
8085 * config/aarch64/constraints.md (Ui7): New constraint.
8086 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
8087 (VFMLA_SEL_W): Ditto.
8088 (f16quad): Ditto.
8089 (f16mac1): Ditto.
8090 (VFMLA16_LOW): New int iterator.
8091 (VFMLA16_HIGH): Ditto.
8092 (UNSPEC_FMLAL): New unspec.
8093 (UNSPEC_FMLSL): Ditto.
8094 (UNSPEC_FMLAL2): Ditto.
8095 (UNSPEC_FMLSL2): Ditto.
8096 (f16mac): New code attribute.
8097 * config/aarch64/aarch64-simd-builtins.def
8098 (aarch64_fmlal_lowv2sf): Ditto.
8099 (aarch64_fmlsl_lowv2sf): Ditto.
8100 (aarch64_fmlalq_lowv4sf): Ditto.
8101 (aarch64_fmlslq_lowv4sf): Ditto.
8102 (aarch64_fmlal_highv2sf): Ditto.
8103 (aarch64_fmlsl_highv2sf): Ditto.
8104 (aarch64_fmlalq_highv4sf): Ditto.
8105 (aarch64_fmlslq_highv4sf): Ditto.
8106 (aarch64_fmlal_lane_lowv2sf): Ditto.
8107 (aarch64_fmlsl_lane_lowv2sf): Ditto.
8108 (aarch64_fmlal_laneq_lowv2sf): Ditto.
8109 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
8110 (aarch64_fmlalq_lane_lowv4sf): Ditto.
8111 (aarch64_fmlsl_lane_lowv4sf): Ditto.
8112 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
8113 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
8114 (aarch64_fmlal_lane_highv2sf): Ditto.
8115 (aarch64_fmlsl_lane_highv2sf): Ditto.
8116 (aarch64_fmlal_laneq_highv2sf): Ditto.
8117 (aarch64_fmlsl_laneq_highv2sf): Ditto.
8118 (aarch64_fmlalq_lane_highv4sf): Ditto.
8119 (aarch64_fmlsl_lane_highv4sf): Ditto.
8120 (aarch64_fmlalq_laneq_highv4sf): Ditto.
8121 (aarch64_fmlsl_laneq_highv4sf): Ditto.
8122 * config/aarch64/aarch64-simd.md:
8123 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
8124 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8125 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
8126 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
8127 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
8128 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
8129 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
8130 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
8131 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
8132 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
8133 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
8134 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
8135 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
8136 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
8137 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
8138 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
8139 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
8140 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
8141 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
8142 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
8143 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
8144 (vfmlsl_low_u32): Ditto.
8145 (vfmlalq_low_u32): Ditto.
8146 (vfmlslq_low_u32): Ditto.
8147 (vfmlal_high_u32): Ditto.
8148 (vfmlsl_high_u32): Ditto.
8149 (vfmlalq_high_u32): Ditto.
8150 (vfmlslq_high_u32): Ditto.
8151 (vfmlal_lane_low_u32): Ditto.
8152 (vfmlsl_lane_low_u32): Ditto.
8153 (vfmlal_laneq_low_u32): Ditto.
8154 (vfmlsl_laneq_low_u32): Ditto.
8155 (vfmlalq_lane_low_u32): Ditto.
8156 (vfmlslq_lane_low_u32): Ditto.
8157 (vfmlalq_laneq_low_u32): Ditto.
8158 (vfmlslq_laneq_low_u32): Ditto.
8159 (vfmlal_lane_high_u32): Ditto.
8160 (vfmlsl_lane_high_u32): Ditto.
8161 (vfmlal_laneq_high_u32): Ditto.
8162 (vfmlsl_laneq_high_u32): Ditto.
8163 (vfmlalq_lane_high_u32): Ditto.
8164 (vfmlslq_lane_high_u32): Ditto.
8165 (vfmlalq_laneq_high_u32): Ditto.
8166 (vfmlslq_laneq_high_u32): Ditto.
8167 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
8168 (AARCH64_FL_FOR_ARCH8_4): New.
8169 (AARCH64_ISA_F16FML): New ISA flag.
8170 (TARGET_F16FML): New feature flag for fp16fml.
8171 (doc/invoke.texi): Document new fp16fml option.
8173 2018-01-10 Michael Collison <michael.collison@arm.com>
8175 * config/aarch64/aarch64-builtins.c:
8176 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
8177 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8178 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
8179 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
8180 (AARCH64_ISA_SHA3): New ISA flag.
8181 (TARGET_SHA3): New feature flag for sha3.
8182 * config/aarch64/iterators.md (sha512_op): New int attribute.
8183 (CRYPTO_SHA512): New int iterator.
8184 (UNSPEC_SHA512H): New unspec.
8185 (UNSPEC_SHA512H2): Ditto.
8186 (UNSPEC_SHA512SU0): Ditto.
8187 (UNSPEC_SHA512SU1): Ditto.
8188 * config/aarch64/aarch64-simd-builtins.def
8189 (aarch64_crypto_sha512hqv2di): New builtin.
8190 (aarch64_crypto_sha512h2qv2di): Ditto.
8191 (aarch64_crypto_sha512su0qv2di): Ditto.
8192 (aarch64_crypto_sha512su1qv2di): Ditto.
8193 (aarch64_eor3qv8hi): Ditto.
8194 (aarch64_rax1qv2di): Ditto.
8195 (aarch64_xarqv2di): Ditto.
8196 (aarch64_bcaxqv8hi): Ditto.
8197 * config/aarch64/aarch64-simd.md:
8198 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
8199 (aarch64_crypto_sha512su0qv2di): Ditto.
8200 (aarch64_crypto_sha512su1qv2di): Ditto.
8201 (aarch64_eor3qv8hi): Ditto.
8202 (aarch64_rax1qv2di): Ditto.
8203 (aarch64_xarqv2di): Ditto.
8204 (aarch64_bcaxqv8hi): Ditto.
8205 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
8206 (vsha512h2q_u64): Ditto.
8207 (vsha512su0q_u64): Ditto.
8208 (vsha512su1q_u64): Ditto.
8209 (veor3q_u16): Ditto.
8210 (vrax1q_u64): Ditto.
8211 (vxarq_u64): Ditto.
8212 (vbcaxq_u16): Ditto.
8213 * config/arm/types.md (crypto_sha512): New type attribute.
8214 (crypto_sha3): Ditto.
8215 (doc/invoke.texi): Document new sha3 option.
8217 2018-01-10 Michael Collison <michael.collison@arm.com>
8219 * config/aarch64/aarch64-builtins.c:
8220 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
8221 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8222 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
8223 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
8224 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
8225 (AARCH64_ISA_SM4): New ISA flag.
8226 (TARGET_SM4): New feature flag for sm4.
8227 * config/aarch64/aarch64-simd-builtins.def
8228 (aarch64_sm3ss1qv4si): Ditto.
8229 (aarch64_sm3tt1aq4si): Ditto.
8230 (aarch64_sm3tt1bq4si): Ditto.
8231 (aarch64_sm3tt2aq4si): Ditto.
8232 (aarch64_sm3tt2bq4si): Ditto.
8233 (aarch64_sm3partw1qv4si): Ditto.
8234 (aarch64_sm3partw2qv4si): Ditto.
8235 (aarch64_sm4eqv4si): Ditto.
8236 (aarch64_sm4ekeyqv4si): Ditto.
8237 * config/aarch64/aarch64-simd.md:
8238 (aarch64_sm3ss1qv4si): Ditto.
8239 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
8240 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
8241 (aarch64_sm4eqv4si): Ditto.
8242 (aarch64_sm4ekeyqv4si): Ditto.
8243 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
8244 (sm3part_op): Ditto.
8245 (CRYPTO_SM3TT): Ditto.
8246 (CRYPTO_SM3PART): Ditto.
8247 (UNSPEC_SM3SS1): New unspec.
8248 (UNSPEC_SM3TT1A): Ditto.
8249 (UNSPEC_SM3TT1B): Ditto.
8250 (UNSPEC_SM3TT2A): Ditto.
8251 (UNSPEC_SM3TT2B): Ditto.
8252 (UNSPEC_SM3PARTW1): Ditto.
8253 (UNSPEC_SM3PARTW2): Ditto.
8254 (UNSPEC_SM4E): Ditto.
8255 (UNSPEC_SM4EKEY): Ditto.
8256 * config/aarch64/constraints.md (Ui2): New constraint.
8257 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
8258 * config/arm/types.md (crypto_sm3): New type attribute.
8259 (crypto_sm4): Ditto.
8260 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
8261 (vsm3tt1aq_u32): Ditto.
8262 (vsm3tt1bq_u32): Ditto.
8263 (vsm3tt2aq_u32): Ditto.
8264 (vsm3tt2bq_u32): Ditto.
8265 (vsm3partw1q_u32): Ditto.
8266 (vsm3partw2q_u32): Ditto.
8267 (vsm4eq_u32): Ditto.
8268 (vsm4ekeyq_u32): Ditto.
8269 (doc/invoke.texi): Document new sm4 option.
8271 2018-01-10 Michael Collison <michael.collison@arm.com>
8273 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
8274 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
8275 (AARCH64_FL_FOR_ARCH8_4): New.
8276 (AARCH64_FL_V8_4): New flag.
8277 (doc/invoke.texi): Document new armv8.4-a option.
8279 2018-01-10 Michael Collison <michael.collison@arm.com>
8281 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
8282 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
8283 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
8284 * config/aarch64/aarch64-option-extension.def: Add
8285 AARCH64_OPT_EXTENSION of 'sha2'.
8286 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
8287 (crypto): Disable sha2 and aes if crypto disabled.
8288 (crypto): Enable aes and sha2 if enabled.
8289 (simd): Disable sha2 and aes if simd disabled.
8290 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
8291 New flags.
8292 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
8293 (TARGET_SHA2): New feature flag for sha2.
8294 (TARGET_AES): New feature flag for aes.
8295 * config/aarch64/aarch64-simd.md:
8296 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
8297 conditional on TARGET_AES.
8298 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
8299 (aarch64_crypto_sha1hsi): Make pattern conditional
8300 on TARGET_SHA2.
8301 (aarch64_crypto_sha1hv4si): Ditto.
8302 (aarch64_be_crypto_sha1hv4si): Ditto.
8303 (aarch64_crypto_sha1su1v4si): Ditto.
8304 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
8305 (aarch64_crypto_sha1su0v4si): Ditto.
8306 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
8307 (aarch64_crypto_sha256su0v4si): Ditto.
8308 (aarch64_crypto_sha256su1v4si): Ditto.
8309 (doc/invoke.texi): Document new aes and sha2 options.
8311 2018-01-10 Martin Sebor <msebor@redhat.com>
8313 PR tree-optimization/83781
8314 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
8315 as string arrays.
8317 2018-01-11 Martin Sebor <msebor@gmail.com>
8318 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8320 PR tree-optimization/83501
8321 PR tree-optimization/81703
8323 * tree-ssa-strlen.c (get_string_cst): Rename...
8324 (get_string_len): ...to this. Handle global constants.
8325 (handle_char_store): Adjust.
8327 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
8328 Jim Wilson <jimw@sifive.com>
8330 * config/riscv/riscv-protos.h (riscv_output_return): New.
8331 * config/riscv/riscv.c (struct machine_function): New naked_p field.
8332 (riscv_attribute_table, riscv_output_return),
8333 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
8334 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
8335 (riscv_compute_frame_info): Only compute frame->mask if not a naked
8336 function.
8337 (riscv_expand_prologue): Add early return for naked function.
8338 (riscv_expand_epilogue): Likewise.
8339 (riscv_function_ok_for_sibcall): Return false for naked function.
8340 (riscv_set_current_function): New.
8341 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
8342 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
8343 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
8344 * doc/extend.texi (RISC-V Function Attributes): New.
8346 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
8348 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
8349 check for 128-bit long double before checking TCmode.
8350 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
8351 128-bit long doubles before checking TFmode or TCmode.
8352 (FLOAT128_IBM_P): Likewise.
8354 2018-01-10 Martin Sebor <msebor@redhat.com>
8356 PR tree-optimization/83671
8357 * builtins.c (c_strlen): Unconditionally return zero for the empty
8358 string.
8359 Use -Warray-bounds for warnings.
8360 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
8361 for non-constant array indices with COMPONENT_REF, arrays of
8362 arrays, and pointers to arrays.
8363 (gimple_fold_builtin_strlen): Determine and set length range for
8364 non-constant character arrays.
8366 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
8368 PR middle-end/81897
8369 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
8370 empty blocks.
8372 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
8374 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
8376 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8378 PR target/83399
8379 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
8380 VECTOR_MEM_ALTIVEC_OR_VSX_P.
8381 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
8382 indexed_or_indirect_operand predicate.
8383 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
8384 (*vsx_le_perm_load_v8hi): Likewise.
8385 (*vsx_le_perm_load_v16qi): Likewise.
8386 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
8387 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
8388 (*vsx_le_perm_store_v8hi): Likewise.
8389 (*vsx_le_perm_store_v16qi): Likewise.
8390 (eight unnamed splitters): Likewise.
8392 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
8394 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
8395 * config/rs6000/emmintrin.h: Likewise.
8396 * config/rs6000/mmintrin.h: Likewise.
8397 * config/rs6000/xmmintrin.h: Likewise.
8399 2018-01-10 David Malcolm <dmalcolm@redhat.com>
8401 PR c++/43486
8402 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
8403 "public_flag".
8404 * tree.c (tree_nop_conversion): Return true for location wrapper
8405 nodes.
8406 (maybe_wrap_with_location): New function.
8407 (selftest::check_strip_nops): New function.
8408 (selftest::test_location_wrappers): New function.
8409 (selftest::tree_c_tests): Call it.
8410 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
8411 (maybe_wrap_with_location): New decl.
8412 (EXPR_LOCATION_WRAPPER_P): New macro.
8413 (location_wrapper_p): New inline function.
8414 (tree_strip_any_location_wrapper): New inline function.
8416 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
8418 PR target/83735
8419 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
8420 stack_realign_offset for the largest alignment of stack slot
8421 actually used.
8422 (ix86_find_max_used_stack_alignment): New function.
8423 (ix86_finalize_stack_frame_flags): Use it. Set
8424 max_used_stack_alignment if we don't realign stack.
8425 * config/i386/i386.h (machine_function): Add
8426 max_used_stack_alignment.
8428 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
8430 * config/arm/arm.opt (-mbranch-cost): New option.
8431 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
8432 account.
8434 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
8436 PR target/83629
8437 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
8438 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
8440 2018-01-10 Richard Biener <rguenther@suse.de>
8442 PR debug/83765
8443 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
8444 early out so it also covers the case where we have a non-NULL
8445 origin.
8447 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8449 PR tree-optimization/83753
8450 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
8451 for non-strided grouped accesses if the number of elements is 1.
8453 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
8455 PR target/81616
8456 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
8457 * i386.h (TARGET_USE_GATHER): Define.
8458 * x86-tune.def (X86_TUNE_USE_GATHER): New.
8460 2018-01-10 Martin Liska <mliska@suse.cz>
8462 PR bootstrap/82831
8463 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
8464 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
8465 partitioning.
8466 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
8467 CLEANUP_NO_PARTITIONING is not set.
8469 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
8471 * doc/rtl.texi: Remove documentation of (const ...) wrappers
8472 for vectors, as a partial revert of r254296.
8473 * rtl.h (const_vec_p): Delete.
8474 (const_vec_duplicate_p): Don't test for vector CONSTs.
8475 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
8476 * expmed.c (make_tree): Likewise.
8478 Revert:
8479 * common.md (E, F): Use CONSTANT_P instead of checking for
8480 CONST_VECTOR.
8481 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
8482 checking for CONST_VECTOR.
8484 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8486 PR middle-end/83575
8487 * predict.c (force_edge_cold): Handle in more sane way edges
8488 with no prediction.
8490 2018-01-09 Carl Love <cel@us.ibm.com>
8492 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
8493 V4SI, V4SF types.
8494 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
8495 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
8496 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
8497 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
8498 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
8499 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
8500 * config/rs6000/rs6000-protos.h: Add extern defition for
8501 rs6000_generate_float2_double_code.
8502 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
8503 function.
8504 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
8505 (float2_v2df): Add define_expand.
8507 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
8509 PR target/83628
8510 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
8511 op_mode in the force_to_mode call.
8513 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8515 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
8516 instead of checking each element individually.
8517 (aarch64_evpc_uzp): Likewise.
8518 (aarch64_evpc_zip): Likewise.
8519 (aarch64_evpc_ext): Likewise.
8520 (aarch64_evpc_rev): Likewise.
8521 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
8522 instead of checking each element individually. Return true without
8523 generating rtl if
8524 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
8525 whether all selected elements come from the same input, instead of
8526 checking each element individually. Remove calls to gen_rtx_REG,
8527 start_sequence and end_sequence and instead assert that no rtl is
8528 generated.
8530 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8532 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
8533 order of HIGH and CONST checks.
8535 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
8537 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
8538 if the destination isn't an SSA_NAME.
8540 2018-01-09 Richard Biener <rguenther@suse.de>
8542 PR tree-optimization/83668
8543 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
8544 move prologue...
8545 (canonicalize_loop_form): ... here, renamed from ...
8546 (canonicalize_loop_closed_ssa_form): ... this and amended to
8547 swap successor edges for loop exit blocks to make us use
8548 the RPO order we need for initial schedule generation.
8550 2018-01-09 Joseph Myers <joseph@codesourcery.com>
8552 PR tree-optimization/64811
8553 * match.pd: When optimizing comparisons with Inf, avoid
8554 introducing or losing exceptions from comparisons with NaN.
8556 2018-01-09 Martin Liska <mliska@suse.cz>
8558 PR sanitizer/82517
8559 * asan.c (shadow_mem_size): Add gcc_assert.
8561 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
8563 Don't save registers in main().
8565 PR target/83738
8566 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
8567 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
8568 * config/avr/avr.c (avr_set_current_function): Don't error if
8569 naked, OS_task or OS_main are specified at the same time.
8570 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
8571 OS_main.
8572 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
8573 attribute.
8574 * common/config/avr/avr-common.c (avr_option_optimization_table):
8575 Switch on -mmain-is-OS_task for optimizing compilations.
8577 2018-01-09 Richard Biener <rguenther@suse.de>
8579 PR tree-optimization/83572
8580 * graphite.c: Include cfganal.h.
8581 (graphite_transform_loops): Connect infinite loops to exit
8582 and remove fake edges at the end.
8584 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8586 * ipa-inline.c (edge_badness): Revert accidental checkin.
8588 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
8590 PR ipa/80763
8591 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
8592 symbols; not inline clones.
8594 2018-01-09 Jakub Jelinek <jakub@redhat.com>
8596 PR target/83507
8597 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
8598 hard registers. Formatting fixes.
8600 PR preprocessor/83722
8601 * gcc.c (try_generate_repro): Pass
8602 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
8603 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
8604 do_report_bug.
8606 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
8607 Kito Cheng <kito.cheng@gmail.com>
8609 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
8610 (riscv_leaf_function_p): Delete.
8611 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
8613 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8615 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
8616 function.
8617 (do_ifelse): New function.
8618 (do_isel): New function.
8619 (do_sub3): New function.
8620 (do_add3): New function.
8621 (do_load_mask_compare): New function.
8622 (do_overlap_load_compare): New function.
8623 (expand_compare_loop): New function.
8624 (expand_block_compare): Call expand_compare_loop() when appropriate.
8625 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
8626 option description.
8627 (-mblock-compare-inline-loop-limit): New option.
8629 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8631 PR target/83677
8632 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
8633 Reverse order of second and third operands in first alternative.
8634 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
8635 of first and second elements in UNSPEC_VPERMR vector.
8636 (altivec_expand_vec_perm_le): Likewise.
8638 2018-01-08 Jeff Law <law@redhat.com>
8640 PR rtl-optimizatin/81308
8641 * tree-switch-conversion.c (cfg_altered): New file scoped static.
8642 (process_switch): If group_case_labels makes a change, then set
8643 cfg_altered.
8644 (pass_convert_switch::execute): If a switch is converted, then
8645 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
8647 PR rtl-optimization/81308
8648 * recog.c (split_all_insns): Conditionally cleanup the CFG after
8649 splitting insns.
8651 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
8653 PR target/83663 - Revert r255946
8654 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
8655 generation for cases where splatting a value is not useful.
8656 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8657 across a vec_duplicate and a paradoxical subreg forming a vector
8658 mode to a vec_concat.
8660 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8662 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
8663 -march=armv8.3-a variants.
8664 * config/arm/t-multilib: Likewise.
8665 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
8667 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8669 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
8670 to generate rtl.
8671 (cceq_ior_compare_complement): Give it a name so I can use it, and
8672 change boolean_or_operator predicate to boolean_operator so it can
8673 be used to generate a crand.
8674 (eqne): New code iterator.
8675 (bd/bd_neg): New code_attrs.
8676 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
8677 a single define_insn.
8678 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
8679 decrement (bdnzt/bdnzf/bdzt/bdzf).
8680 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
8681 with the new names of the branch decrement patterns, and added the
8682 names of the branch decrement conditional patterns.
8684 2018-01-08 Richard Biener <rguenther@suse.de>
8686 PR tree-optimization/83563
8687 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
8688 cache.
8690 2018-01-08 Richard Biener <rguenther@suse.de>
8692 PR middle-end/83713
8693 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
8695 2018-01-08 Richard Biener <rguenther@suse.de>
8697 PR tree-optimization/83685
8698 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
8699 references to abnormals.
8701 2018-01-08 Richard Biener <rguenther@suse.de>
8703 PR lto/83719
8704 * dwarf2out.c (output_indirect_strings): Handle empty
8705 skeleton_debug_str_hash.
8706 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
8708 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8710 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
8711 (emit_store_direct): Likewise.
8712 (arc_trampoline_adjust_address): Likewise.
8713 (arc_asm_trampoline_template): New function.
8714 (arc_initialize_trampoline): Use asm_trampoline_template.
8715 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
8716 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
8717 * config/arc/arc.md (flush_icache): Delete pattern.
8719 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
8721 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
8722 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
8723 munaligned-access.
8725 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8727 PR target/83681
8728 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
8729 by not USED_FOR_TARGET.
8730 (make_pass_resolve_sw_modes): Likewise.
8732 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
8734 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
8735 USED_FOR_TARGET.
8737 2018-01-08 Richard Biener <rguenther@suse.de>
8739 PR middle-end/83580
8740 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
8742 2018-01-08 Richard Biener <rguenther@suse.de>
8744 PR middle-end/83517
8745 * match.pd ((t * 2) / 2) -> t): Add missing :c.
8747 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
8749 PR middle-end/81897
8750 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
8751 basic blocks with a small number of successors.
8752 (convert_control_dep_chain_into_preds): Improve handling of
8753 forwarder blocks.
8754 (dump_predicates): Split apart into...
8755 (dump_pred_chain): ...here...
8756 (dump_pred_info): ...and here.
8757 (can_one_predicate_be_invalidated_p): Add debugging printfs.
8758 (can_chain_union_be_invalidated_p): Improve check for invalidation
8759 of paths.
8760 (uninit_uses_cannot_happen): Avoid unnecessary if
8761 convert_control_dep_chain_into_preds yielded nothing.
8763 2018-01-06 Martin Sebor <msebor@redhat.com>
8765 PR tree-optimization/83640
8766 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
8767 subtracting negative offset from size.
8768 (builtin_access::overlap): Adjust offset bounds of the access to fall
8769 within the size of the object if possible.
8771 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
8773 PR rtl-optimization/83699
8774 * expmed.c (extract_bit_field_1): Restrict the vector usage of
8775 extract_bit_field_as_subreg to cases in which the extracted
8776 value is also a vector.
8778 * lra-constraints.c (process_alt_operands): Test for the equivalence
8779 substitutions when detecting a possible reload cycle.
8781 2018-01-06 Jakub Jelinek <jakub@redhat.com>
8783 PR debug/83480
8784 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
8785 by default if flag_selective_schedling{,2}. Formatting fixes.
8787 PR rtl-optimization/83682
8788 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
8789 if it has non-VECTOR_MODE element mode.
8790 (vec_duplicate_p): Likewise.
8792 PR middle-end/83694
8793 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
8794 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
8796 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8798 PR target/83604
8799 * config/i386/i386-builtin.def
8800 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8801 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
8802 Require also OPTION_MASK_ISA_AVX512F in addition to
8803 OPTION_MASK_ISA_GFNI.
8804 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
8805 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
8806 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
8807 to OPTION_MASK_ISA_GFNI.
8808 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
8809 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
8810 OPTION_MASK_ISA_AVX512BW.
8811 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
8812 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
8813 addition to OPTION_MASK_ISA_GFNI.
8814 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
8815 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
8816 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
8817 to OPTION_MASK_ISA_GFNI.
8818 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
8819 a requirement for all ISAs rather than any of them with a few
8820 exceptions.
8821 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
8822 processing.
8823 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
8824 bitmasks to be enabled with 3 exceptions, instead of requiring any
8825 enabled ISA with lots of exceptions.
8826 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
8827 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
8828 Change avx512bw in isa attribute to avx512f.
8829 * config/i386/sgxintrin.h: Add license boilerplate.
8830 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
8831 to __AVX512F__ and __AVX512VL to __AVX512VL__.
8832 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
8833 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
8834 defined.
8835 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
8836 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
8837 temporarily sse2 rather than sse if not enabled already.
8839 PR target/83604
8840 * config/i386/sse.md (VI248_VLBW): Rename to ...
8841 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
8842 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
8843 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
8844 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
8845 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
8846 mode iterator instead of VI248_VLBW.
8848 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
8850 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
8851 (record_modified): Skip clobbers; add debug output.
8852 (param_change_prob): Use sreal frequencies.
8854 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8856 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
8857 punt for user-aligned variables.
8859 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
8861 * tree-chrec.c (chrec_contains_symbols): Return true for
8862 POLY_INT_CST.
8864 2018-01-05 Sudakshina Das <sudi.das@arm.com>
8866 PR target/82439
8867 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
8868 of (x|y) == x for BICS pattern.
8870 2018-01-05 Jakub Jelinek <jakub@redhat.com>
8872 PR tree-optimization/83605
8873 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
8874 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
8875 can throw.
8877 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
8879 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
8880 * config/epiphany/rtems.h: New file.
8882 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8883 Uros Bizjak <ubizjak@gmail.com>
8885 PR target/83554
8886 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
8887 QIreg_operand instead of register_operand predicate.
8888 * config/i386/i386.c (ix86_rop_should_change_byte_p,
8889 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
8890 comments instead of -fmitigate[-_]rop.
8892 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
8894 PR bootstrap/81926
8895 * cgraphunit.c (symbol_table::compile): Switch to text_section
8896 before calling assembly_start debug hook.
8897 * run-rtl-passes.c (run_rtl_passes): Likewise.
8898 Include output.h.
8900 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8902 * tree-vrp.c (extract_range_from_binary_expr_1): Check
8903 range_int_cst_p rather than !symbolic_range_p before calling
8904 extract_range_from_multiplicative_op_1.
8906 2018-01-04 Jeff Law <law@redhat.com>
8908 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
8909 redundant test in assertion.
8911 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8913 * doc/rtl.texi: Document machine_mode wrapper classes.
8915 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8917 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
8918 using tree_to_uhwi.
8920 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8922 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
8923 the VEC_PERM_EXPR fold to fail.
8925 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8927 PR debug/83585
8928 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
8929 to switched_sections.
8931 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8933 PR target/83680
8934 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
8935 test for d.testing.
8937 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
8939 PR target/83387
8940 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
8941 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
8943 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8945 PR debug/83666
8946 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
8947 is BLKmode and bitpos not zero or mode change is needed.
8949 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
8951 PR target/83675
8952 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
8953 TARGET_VIS2.
8955 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
8957 PR target/83628
8958 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
8959 instead of MULT rtx. Update all corresponding splitters.
8960 (*saddl_se): Ditto.
8961 (*ssub<modesuffix>): Ditto.
8962 (*ssubl_se): Ditto.
8963 (*cmp_sadd_di): Update split patterns.
8964 (*cmp_sadd_si): Ditto.
8965 (*cmp_sadd_sidi): Ditto.
8966 (*cmp_ssub_di): Ditto.
8967 (*cmp_ssub_si): Ditto.
8968 (*cmp_ssub_sidi): Ditto.
8969 * config/alpha/predicates.md (const23_operand): New predicate.
8970 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
8971 Look for ASHIFT, not MULT inner operand.
8972 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
8974 2018-01-04 Martin Liska <mliska@suse.cz>
8976 PR gcov-profile/83669
8977 * gcov.c (output_intermediate_file): Add version to intermediate
8978 gcov file.
8979 * doc/gcov.texi: Document new field 'version' in intermediate
8980 file format. Fix location of '-k' option of gcov command.
8982 2018-01-04 Martin Liska <mliska@suse.cz>
8984 PR ipa/82352
8985 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
8987 2018-01-04 Jakub Jelinek <jakub@redhat.com>
8989 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
8991 2018-01-03 Martin Sebor <msebor@redhat.com>
8993 PR tree-optimization/83655
8994 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
8995 checking calls with invalid arguments.
8997 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
8999 * tree-vect-stmts.c (vect_get_store_rhs): New function.
9000 (vectorizable_mask_load_store): Delete.
9001 (vectorizable_call): Return false for masked loads and stores.
9002 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
9003 instead of gimple_assign_rhs1.
9004 (vectorizable_load): Handle IFN_MASK_LOAD.
9005 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
9007 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9009 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
9010 split out from..,
9011 (vectorizable_mask_load_store): ...here.
9012 (vectorizable_load): ...and here.
9014 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9016 * tree-vect-stmts.c (vect_build_all_ones_mask)
9017 (vect_build_zero_merge_argument): New functions, split out from...
9018 (vectorizable_load): ...here.
9020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9022 * tree-vect-stmts.c (vect_check_store_rhs): New function,
9023 split out from...
9024 (vectorizable_mask_load_store): ...here.
9025 (vectorizable_store): ...and here.
9027 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9029 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
9030 split out from...
9031 (vectorizable_mask_load_store): ...here.
9033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9035 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
9036 (vect_model_store_cost): Take a vec_load_store_type instead of a
9037 vect_def_type.
9038 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
9039 (vect_model_store_cost): Take a vec_load_store_type instead of a
9040 vect_def_type.
9041 (vectorizable_mask_load_store): Update accordingly.
9042 (vectorizable_store): Likewise.
9043 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
9045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9047 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
9048 IFN_MASK_LOAD calls here rather than...
9049 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
9051 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9052 Alan Hayward <alan.hayward@arm.com>
9053 David Sherwood <david.sherwood@arm.com>
9055 * expmed.c (extract_bit_field_1): For vector extracts,
9056 fall back to extract_bit_field_as_subreg if vec_extract
9057 isn't available.
9059 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9060 Alan Hayward <alan.hayward@arm.com>
9061 David Sherwood <david.sherwood@arm.com>
9063 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
9064 they are variable or constant sized.
9065 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
9066 slots for constant-sized data.
9068 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9069 Alan Hayward <alan.hayward@arm.com>
9070 David Sherwood <david.sherwood@arm.com>
9072 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
9073 handling COND_EXPRs with boolean comparisons, try to find a better
9074 basis for the mask type than the boolean itself.
9076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9078 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
9079 is calculated and how it can be overridden.
9080 * genmodes.c (max_bitsize_mode_any_mode): New variable.
9081 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
9082 if defined.
9083 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
9084 if nonzero.
9086 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9087 Alan Hayward <alan.hayward@arm.com>
9088 David Sherwood <david.sherwood@arm.com>
9090 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
9091 Remove the mode argument.
9092 (aarch64_simd_valid_immediate): Remove the mode and inverse
9093 arguments.
9094 * config/aarch64/iterators.md (bitsize): New iterator.
9095 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
9096 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
9097 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
9098 aarch64_simd_valid_immediate.
9099 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
9100 (aarch64_reg_or_bic_imm): Likewise.
9101 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
9102 with an insn_type enum and msl with a modifier_type enum.
9103 Replace element_width with a scalar_mode. Change the shift
9104 to unsigned int. Add constructors for scalar_float_mode and
9105 scalar_int_mode elements.
9106 (aarch64_vect_float_const_representable_p): Delete.
9107 (aarch64_can_const_movi_rtx_p)
9108 (aarch64_simd_scalar_immediate_valid_for_move)
9109 (aarch64_simd_make_constant): Update call to
9110 aarch64_simd_valid_immediate.
9111 (aarch64_advsimd_valid_immediate_hs): New function.
9112 (aarch64_advsimd_valid_immediate): Likewise.
9113 (aarch64_simd_valid_immediate): Remove mode and inverse
9114 arguments. Rewrite to use the above. Use const_vec_duplicate_p
9115 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
9116 and aarch64_float_const_representable_p on the result.
9117 (aarch64_output_simd_mov_immediate): Remove mode argument.
9118 Update call to aarch64_simd_valid_immediate and use of
9119 simd_immediate_info.
9120 (aarch64_output_scalar_simd_mov_immediate): Update call
9121 accordingly.
9123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9124 Alan Hayward <alan.hayward@arm.com>
9125 David Sherwood <david.sherwood@arm.com>
9127 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
9128 (mode_nunits): Likewise CONST_MODE_NUNITS.
9129 * machmode.def (ADJUST_NUNITS): Document.
9130 * genmodes.c (mode_data::need_nunits_adj): New field.
9131 (blank_mode): Update accordingly.
9132 (adj_nunits): New variable.
9133 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
9134 parameter.
9135 (emit_mode_size_inline): Set need_bytesize_adj for all modes
9136 listed in adj_nunits.
9137 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
9138 listed in adj_nunits. Don't emit case statements for such modes.
9139 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
9140 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
9141 nothing if adj_nunits is nonnull.
9142 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
9143 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
9144 (emit_mode_fbit): Update use of print_maybe_const_decl.
9145 (emit_move_size): Likewise. Treat the array as non-const
9146 if adj_nunits.
9147 (emit_mode_adjustments): Handle adj_nunits.
9149 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9151 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
9152 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
9153 (VECTOR_MODES): Use it.
9154 (make_vector_modes): Take the prefix as an argument.
9156 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9157 Alan Hayward <alan.hayward@arm.com>
9158 David Sherwood <david.sherwood@arm.com>
9160 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
9161 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
9162 for MODE_VECTOR_BOOL.
9163 * machmode.def (VECTOR_BOOL_MODE): Document.
9164 * genmodes.c (VECTOR_BOOL_MODE): New macro.
9165 (make_vector_bool_mode): New function.
9166 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
9167 MODE_VECTOR_BOOL.
9168 * lto-streamer-in.c (lto_input_mode_table): Likewise.
9169 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
9170 Likewise.
9171 * stor-layout.c (int_mode_for_mode): Likewise.
9172 * tree.c (build_vector_type_for_mode): Likewise.
9173 * varasm.c (output_constant_pool_2): Likewise.
9174 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
9175 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
9176 for MODE_VECTOR_BOOL.
9177 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
9178 of mode class checks.
9179 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
9180 instead of a list of mode class checks.
9181 (expand_vector_scalar_condition): Likewise.
9182 (type_for_widest_vector_mode): Handle BImode as an inner mode.
9184 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9185 Alan Hayward <alan.hayward@arm.com>
9186 David Sherwood <david.sherwood@arm.com>
9188 * machmode.h (mode_size): Change from unsigned short to
9189 poly_uint16_pod.
9190 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
9191 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9192 or if measurement_type is not polynomial.
9193 (fixed_size_mode::includes_p): Check for constant-sized modes.
9194 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
9195 return a poly_uint16 rather than an unsigned short.
9196 (emit_mode_size): Change the type of mode_size from unsigned short
9197 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
9198 (emit_mode_adjustments): Cope with polynomial vector sizes.
9199 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9200 for GET_MODE_SIZE.
9201 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9202 for GET_MODE_SIZE.
9203 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
9204 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
9205 * caller-save.c (setup_save_areas): Likewise.
9206 (replace_reg_with_saved_mem): Likewise.
9207 * calls.c (emit_library_call_value_1): Likewise.
9208 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
9209 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
9210 (gen_lowpart_for_combine): Likewise.
9211 * convert.c (convert_to_integer_1): Likewise.
9212 * cse.c (equiv_constant, cse_insn): Likewise.
9213 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
9214 (cselib_subst_to_values): Likewise.
9215 * dce.c (word_dce_process_block): Likewise.
9216 * df-problems.c (df_word_lr_mark_ref): Likewise.
9217 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
9218 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
9219 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
9220 (rtl_for_decl_location): Likewise.
9221 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
9222 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
9223 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
9224 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
9225 (expand_expr_real_1): Likewise.
9226 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
9227 (pad_below): Likewise.
9228 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9229 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
9230 * ira.c (get_subreg_tracking_sizes): Likewise.
9231 * ira-build.c (ira_create_allocno_objects): Likewise.
9232 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
9233 (ira_sort_regnos_for_alter_reg): Likewise.
9234 * ira-costs.c (record_operand_costs): Likewise.
9235 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
9236 (resolve_simple_move): Likewise.
9237 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
9238 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
9239 (lra_constraints): Likewise.
9240 (CONST_POOL_OK_P): Reject variable-sized modes.
9241 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
9242 (add_pseudo_to_slot, lra_spill): Likewise.
9243 * omp-low.c (omp_clause_aligned_alignment): Likewise.
9244 * optabs-query.c (get_best_extraction_insn): Likewise.
9245 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9246 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
9247 (expand_mult_highpart, valid_multiword_target_p): Likewise.
9248 * recog.c (offsettable_address_addr_space_p): Likewise.
9249 * regcprop.c (maybe_mode_change): Likewise.
9250 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
9251 * regrename.c (build_def_use): Likewise.
9252 * regstat.c (dump_reg_info): Likewise.
9253 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
9254 (find_reloads, find_reloads_subreg_address): Likewise.
9255 * reload1.c (eliminate_regs_1): Likewise.
9256 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
9257 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
9258 (simplify_binary_operation_1, simplify_subreg): Likewise.
9259 * targhooks.c (default_function_arg_padding): Likewise.
9260 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
9261 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
9262 (verify_gimple_assign_ternary): Likewise.
9263 * tree-inline.c (estimate_move_cost): Likewise.
9264 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9265 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
9266 (get_address_cost_ainc): Likewise.
9267 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
9268 (vect_supportable_dr_alignment): Likewise.
9269 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9270 (vectorizable_reduction): Likewise.
9271 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
9272 (vectorizable_operation, vectorizable_load): Likewise.
9273 * tree.c (build_same_sized_truth_vector_type): Likewise.
9274 * valtrack.c (cleanup_auto_inc_dec): Likewise.
9275 * var-tracking.c (emit_note_insn_var_location): Likewise.
9276 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
9277 (ADDR_VEC_ALIGN): Likewise.
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 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
9284 unsigned short.
9285 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
9286 or if measurement_type is polynomial.
9287 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
9288 * combine.c (make_extraction): Likewise.
9289 * dse.c (find_shift_sequence): Likewise.
9290 * dwarf2out.c (mem_loc_descriptor): Likewise.
9291 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
9292 (extract_bit_field, extract_low_bits): Likewise.
9293 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
9294 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
9295 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
9296 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
9297 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
9298 * reload.c (find_reloads): Likewise.
9299 * reload1.c (alter_reg): Likewise.
9300 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
9301 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
9302 * tree-if-conv.c (predicate_mem_writes): Likewise.
9303 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
9304 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
9305 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9306 * valtrack.c (dead_debug_insert_temp): Likewise.
9307 * varasm.c (mergeable_constant_section): Likewise.
9308 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
9310 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9311 Alan Hayward <alan.hayward@arm.com>
9312 David Sherwood <david.sherwood@arm.com>
9314 * expr.c (expand_assignment): Cope with polynomial mode sizes
9315 when assigning to a CONCAT.
9317 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9318 Alan Hayward <alan.hayward@arm.com>
9319 David Sherwood <david.sherwood@arm.com>
9321 * machmode.h (mode_precision): Change from unsigned short to
9322 poly_uint16_pod.
9323 (mode_to_precision): Return a poly_uint16 rather than an unsigned
9324 short.
9325 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
9326 or if measurement_type is not polynomial.
9327 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
9328 in which the mode is already known to be a scalar_int_mode.
9329 * genmodes.c (emit_mode_precision): Change the type of mode_precision
9330 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
9331 initializer.
9332 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9333 for GET_MODE_PRECISION.
9334 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9335 for GET_MODE_PRECISION.
9336 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
9337 as polynomial.
9338 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
9339 (expand_field_assignment, make_extraction): Likewise.
9340 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
9341 (get_last_value): Likewise.
9342 * convert.c (convert_to_integer_1): Likewise.
9343 * cse.c (cse_insn): Likewise.
9344 * expr.c (expand_expr_real_1): Likewise.
9345 * lra-constraints.c (simplify_operand_subreg): Likewise.
9346 * optabs-query.c (can_atomic_load_p): Likewise.
9347 * optabs.c (expand_atomic_load): Likewise.
9348 (expand_atomic_store): Likewise.
9349 * ree.c (combine_reaching_defs): Likewise.
9350 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
9351 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
9352 * tree.h (type_has_mode_precision_p): Likewise.
9353 * ubsan.c (instrument_si_overflow): Likewise.
9355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9356 Alan Hayward <alan.hayward@arm.com>
9357 David Sherwood <david.sherwood@arm.com>
9359 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
9360 polynomial numbers of units.
9361 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
9362 (valid_vector_subparts_p): New function.
9363 (build_vector_type): Remove temporary shim and take the number
9364 of units as a poly_uint64 rather than an int.
9365 (build_opaque_vector_type): Take the number of units as a
9366 poly_uint64 rather than an int.
9367 * tree.c (build_vector_from_ctor): Handle polynomial
9368 TYPE_VECTOR_SUBPARTS.
9369 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
9370 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
9371 (build_vector_from_val): If the number of units is variable,
9372 use build_vec_duplicate_cst for constant operands and
9373 VEC_DUPLICATE_EXPR otherwise.
9374 (make_vector_type): Remove temporary is_constant ().
9375 (build_vector_type, build_opaque_vector_type): Take the number of
9376 units as a poly_uint64 rather than an int.
9377 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
9378 VECTOR_CST_NELTS.
9379 * cfgexpand.c (expand_debug_expr): Likewise.
9380 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
9381 (store_constructor, expand_expr_real_1): Likewise.
9382 (const_scalar_mask_from_tree): Likewise.
9383 * fold-const-call.c (fold_const_reduction): Likewise.
9384 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
9385 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
9386 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
9387 (fold_relational_const): Likewise.
9388 (native_interpret_vector): Likewise. Change the size from an
9389 int to an unsigned int.
9390 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
9391 TYPE_VECTOR_SUBPARTS.
9392 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
9393 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
9394 duplicating a non-constant operand into a variable-length vector.
9395 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
9396 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
9397 * ipa-icf.c (sem_variable::equals): Likewise.
9398 * match.pd: Likewise.
9399 * omp-simd-clone.c (simd_clone_subparts): Likewise.
9400 * print-tree.c (print_node): Likewise.
9401 * stor-layout.c (layout_type): Likewise.
9402 * targhooks.c (default_builtin_vectorization_cost): Likewise.
9403 * tree-cfg.c (verify_gimple_comparison): Likewise.
9404 (verify_gimple_assign_binary): Likewise.
9405 (verify_gimple_assign_ternary): Likewise.
9406 (verify_gimple_assign_single): Likewise.
9407 * tree-pretty-print.c (dump_generic_node): Likewise.
9408 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
9409 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
9410 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
9411 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
9412 (vect_shift_permute_load_chain): Likewise.
9413 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
9414 (expand_vector_condition, optimize_vector_constructor): Likewise.
9415 (lower_vec_perm, get_compute_type): Likewise.
9416 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
9417 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
9418 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
9419 (vect_recog_mask_conversion_pattern): Likewise.
9420 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
9421 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
9422 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
9423 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
9424 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
9425 (vectorizable_shift, vectorizable_operation, vectorizable_store)
9426 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
9427 (supportable_widening_operation): Likewise.
9428 (supportable_narrowing_operation): Likewise.
9429 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
9430 Likewise.
9431 * varasm.c (output_constant): Likewise.
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 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
9438 so that both the length == 3 and length != 3 cases set up their
9439 own permute vectors. Add comments explaining why we know the
9440 number of elements is constant.
9441 (vect_permute_load_chain): Likewise.
9443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9444 Alan Hayward <alan.hayward@arm.com>
9445 David Sherwood <david.sherwood@arm.com>
9447 * machmode.h (mode_nunits): Change from unsigned char to
9448 poly_uint16_pod.
9449 (ONLY_FIXED_SIZE_MODES): New macro.
9450 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
9451 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
9452 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
9453 New typedefs.
9454 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
9455 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
9456 or if measurement_type is not polynomial.
9457 * genmodes.c (ZERO_COEFFS): New macro.
9458 (emit_mode_nunits_inline): Make mode_nunits_inline return a
9459 poly_uint16.
9460 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
9461 Use ZERO_COEFFS when emitting initializers.
9462 * data-streamer.h (bp_pack_poly_value): New function.
9463 (bp_unpack_poly_value): Likewise.
9464 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
9465 for GET_MODE_NUNITS.
9466 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
9467 for GET_MODE_NUNITS.
9468 * tree.c (make_vector_type): Remove temporary shim and make
9469 the real function take the number of units as a poly_uint64
9470 rather than an int.
9471 (build_vector_type_for_mode): Handle polynomial nunits.
9472 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
9473 * emit-rtl.c (const_vec_series_p_1): Likewise.
9474 (gen_rtx_CONST_VECTOR): Likewise.
9475 * fold-const.c (test_vec_duplicate_folding): Likewise.
9476 * genrecog.c (validate_pattern): Likewise.
9477 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
9478 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
9479 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
9480 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
9481 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
9482 * rtlanal.c (subreg_get_info): Likewise.
9483 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
9484 (vect_grouped_load_supported): Likewise.
9485 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
9486 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
9487 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
9488 (simplify_const_unary_operation, simplify_binary_operation_1)
9489 (simplify_const_binary_operation, simplify_ternary_operation)
9490 (test_vector_ops_duplicate, test_vector_ops): Likewise.
9491 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
9492 instead of CONST_VECTOR_NUNITS.
9493 * varasm.c (output_constant_pool_2): Likewise.
9494 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
9495 explicit-encoded elements in the XVEC for variable-length vectors.
9497 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9499 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
9501 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9502 Alan Hayward <alan.hayward@arm.com>
9503 David Sherwood <david.sherwood@arm.com>
9505 * coretypes.h (fixed_size_mode): Declare.
9506 (fixed_size_mode_pod): New typedef.
9507 * builtins.h (target_builtins::x_apply_args_mode)
9508 (target_builtins::x_apply_result_mode): Change type to
9509 fixed_size_mode_pod.
9510 * builtins.c (apply_args_size, apply_result_size, result_vector)
9511 (expand_builtin_apply_args_1, expand_builtin_apply)
9512 (expand_builtin_return): Update accordingly.
9514 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9516 * cse.c (hash_rtx_cb): Hash only the encoded elements.
9517 * cselib.c (cselib_hash_rtx): Likewise.
9518 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
9519 CONST_VECTOR encoding.
9521 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9522 Jeff Law <law@redhat.com>
9524 PR target/83641
9525 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
9526 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
9527 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
9528 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
9530 PR target/83641
9531 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
9532 explicitly probe *sp in a noreturn function if there were any callee
9533 register saves or frame pointer is needed.
9535 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9537 PR debug/83621
9538 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
9539 BLKmode for ternary, binary or unary expressions.
9541 PR debug/83645
9542 * var-tracking.c (delete_vta_debug_insn): New inline function.
9543 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
9544 insns from get_insns () to NULL instead of each bb separately.
9545 Use delete_vta_debug_insn. No longer static.
9546 (vt_debug_insns_local, variable_tracking_main_1): Adjust
9547 delete_vta_debug_insns callers.
9548 * rtl.h (delete_vta_debug_insns): Declare.
9549 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
9550 instead of variable_tracking_main.
9552 2018-01-03 Martin Sebor <msebor@redhat.com>
9554 PR tree-optimization/83603
9555 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
9556 arguments past the endof the argument list in functions declared
9557 without a prototype.
9558 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
9559 Avoid checking when arguments are null.
9561 2018-01-03 Martin Sebor <msebor@redhat.com>
9563 PR c/83559
9564 * doc/extend.texi (attribute const): Fix a typo.
9565 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
9566 issuing -Wsuggest-attribute for void functions.
9568 2018-01-03 Martin Sebor <msebor@redhat.com>
9570 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
9571 offset_int::from instead of wide_int::to_shwi.
9572 (maybe_diag_overlap): Remove assertion.
9573 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
9574 * gimple-ssa-sprintf.c (format_directive): Same.
9575 (parse_directive): Same.
9576 (sprintf_dom_walker::compute_format_length): Same.
9577 (try_substitute_return_value): Same.
9579 2018-01-03 Jeff Law <law@redhat.com>
9581 PR middle-end/83654
9582 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
9583 non-constant residual for zero at runtime and avoid probing in
9584 that case. Reorganize code for trailing problem to mirror handling
9585 of the residual.
9587 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
9589 PR tree-optimization/83501
9590 * tree-ssa-strlen.c (get_string_cst): New.
9591 (handle_char_store): Call get_string_cst.
9593 2018-01-03 Martin Liska <mliska@suse.cz>
9595 PR tree-optimization/83593
9596 * tree-ssa-strlen.c: Include tree-cfg.h.
9597 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
9598 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
9599 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
9600 to false.
9601 (strlen_dom_walker::before_dom_children): Call
9602 gimple_purge_dead_eh_edges. Dump tranformation with details
9603 dump flags.
9604 (strlen_dom_walker::before_dom_children): Update call by adding
9605 new argument cleanup_eh.
9606 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
9608 2018-01-03 Martin Liska <mliska@suse.cz>
9610 PR ipa/83549
9611 * cif-code.def (VARIADIC_THUNK): New enum value.
9612 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
9613 thunks.
9615 2018-01-03 Jan Beulich <jbeulich@suse.com>
9617 * sse.md (mov<mode>_internal): Tighten condition for when to use
9618 vmovdqu<ssescalarsize> for TI and OI modes.
9620 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9622 Update copyright years.
9624 2018-01-03 Martin Liska <mliska@suse.cz>
9626 PR ipa/83594
9627 * ipa-visibility.c (function_and_variable_visibility): Skip
9628 functions with noipa attribure.
9630 2018-01-03 Jakub Jelinek <jakub@redhat.com>
9632 * gcc.c (process_command): Update copyright notice dates.
9633 * gcov-dump.c (print_version): Ditto.
9634 * gcov.c (print_version): Ditto.
9635 * gcov-tool.c (print_version): Ditto.
9636 * gengtype.c (create_file): Ditto.
9637 * doc/cpp.texi: Bump @copying's copyright year.
9638 * doc/cppinternals.texi: Ditto.
9639 * doc/gcc.texi: Ditto.
9640 * doc/gccint.texi: Ditto.
9641 * doc/gcov.texi: Ditto.
9642 * doc/install.texi: Ditto.
9643 * doc/invoke.texi: Ditto.
9645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9647 * vector-builder.h (vector_builder::m_full_nelts): Change from
9648 unsigned int to poly_uint64.
9649 (vector_builder::full_nelts): Update prototype accordingly.
9650 (vector_builder::new_vector): Likewise.
9651 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
9652 (vector_builder::operator ==): Likewise.
9653 (vector_builder::finalize): Likewise.
9654 * int-vector-builder.h (int_vector_builder::int_vector_builder):
9655 Take the number of elements as a poly_uint64 rather than an
9656 unsigned int.
9657 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
9658 from unsigned int to poly_uint64.
9659 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
9660 (vec_perm_indices::new_vector): Likewise.
9661 (vec_perm_indices::length): Likewise.
9662 (vec_perm_indices::nelts_per_input): Likewise.
9663 (vec_perm_indices::input_nelts): Likewise.
9664 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
9665 number of elements per input as a poly_uint64 rather than an
9666 unsigned int. Use the original encoding for variable-length
9667 vectors, rather than clamping each individual element.
9668 For the second and subsequent elements in each pattern,
9669 clamp the step and base before clamping their sum.
9670 (vec_perm_indices::series_p): Handle polynomial element counts.
9671 (vec_perm_indices::all_in_range_p): Likewise.
9672 (vec_perm_indices_to_tree): Likewise.
9673 (vec_perm_indices_to_rtx): Likewise.
9674 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
9675 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
9676 (tree_vector_builder::new_binary_operation): Handle polynomial
9677 element counts. Return false if we need to know the number
9678 of elements at compile time.
9679 * fold-const.c (fold_vec_perm): Punt if the number of elements
9680 isn't known at compile time.
9682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9684 * vec-perm-indices.h (vec_perm_builder): Change element type
9685 from HOST_WIDE_INT to poly_int64.
9686 (vec_perm_indices::element_type): Update accordingly.
9687 (vec_perm_indices::clamp): Handle polynomial element_types.
9688 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
9689 (vec_perm_indices::all_in_range_p): Likewise.
9690 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
9691 than shwi trees.
9692 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
9693 polynomial vec_perm_indices element types.
9694 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
9695 * fold-const.c (fold_vec_perm): Likewise.
9696 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
9697 * tree-vect-generic.c (lower_vec_perm): Likewise.
9698 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
9699 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
9700 element type to HOST_WIDE_INT.
9702 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9703 Alan Hayward <alan.hayward@arm.com>
9704 David Sherwood <david.sherwood@arm.com>
9706 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
9707 rather than an int. Use plus_constant.
9708 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
9709 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
9711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9712 Alan Hayward <alan.hayward@arm.com>
9713 David Sherwood <david.sherwood@arm.com>
9715 * calls.c (emit_call_1, expand_call): Change struct_value_size from
9716 a HOST_WIDE_INT to a poly_int64.
9718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9719 Alan Hayward <alan.hayward@arm.com>
9720 David Sherwood <david.sherwood@arm.com>
9722 * calls.c (load_register_parameters): Cope with polynomial
9723 mode sizes. Require a constant size for BLKmode parameters
9724 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
9725 forces a parameter to be padded at the lsb end in order to
9726 fill a complete number of words, require the parameter size
9727 to be ordered wrt UNITS_PER_WORD.
9729 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9730 Alan Hayward <alan.hayward@arm.com>
9731 David Sherwood <david.sherwood@arm.com>
9733 * reload1.c (spill_stack_slot_width): Change element type
9734 from unsigned int to poly_uint64_pod.
9735 (alter_reg): Treat mode sizes as polynomial.
9737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9738 Alan Hayward <alan.hayward@arm.com>
9739 David Sherwood <david.sherwood@arm.com>
9741 * reload.c (complex_word_subreg_p): New function.
9742 (reload_inner_reg_of_subreg, push_reload): Use it.
9744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9745 Alan Hayward <alan.hayward@arm.com>
9746 David Sherwood <david.sherwood@arm.com>
9748 * lra-constraints.c (process_alt_operands): Reject matched
9749 operands whose sizes aren't ordered.
9750 (match_reload): Refer to this check here.
9752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9753 Alan Hayward <alan.hayward@arm.com>
9754 David Sherwood <david.sherwood@arm.com>
9756 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
9757 that the mode size is in the set {1, 2, 4, 8, 16}.
9759 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9760 Alan Hayward <alan.hayward@arm.com>
9761 David Sherwood <david.sherwood@arm.com>
9763 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
9764 Use plus_constant instead of gen_rtx_PLUS.
9766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9767 Alan Hayward <alan.hayward@arm.com>
9768 David Sherwood <david.sherwood@arm.com>
9770 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
9771 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
9772 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
9773 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
9774 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
9775 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
9776 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
9777 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
9778 * config/i386/i386.c (ix86_push_rounding): ...this new function.
9779 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
9780 a poly_int64.
9781 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
9782 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
9783 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
9784 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
9785 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
9786 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
9787 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
9788 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
9789 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
9790 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
9791 function.
9792 * expr.c (emit_move_resolve_push): Treat the input and result
9793 of PUSH_ROUNDING as a poly_int64.
9794 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
9795 (emit_push_insn): Likewise.
9796 * lra-eliminations.c (mark_not_eliminable): Likewise.
9797 * recog.c (push_operand): Likewise.
9798 * reload1.c (elimination_effects): Likewise.
9799 * rtlanal.c (nonzero_bits1): Likewise.
9800 * calls.c (store_one_arg): Likewise. Require the padding to be
9801 known at compile time.
9803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9804 Alan Hayward <alan.hayward@arm.com>
9805 David Sherwood <david.sherwood@arm.com>
9807 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
9808 Use plus_constant instead of gen_rtx_PLUS.
9810 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9811 Alan Hayward <alan.hayward@arm.com>
9812 David Sherwood <david.sherwood@arm.com>
9814 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
9815 rather than an int.
9817 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9818 Alan Hayward <alan.hayward@arm.com>
9819 David Sherwood <david.sherwood@arm.com>
9821 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
9822 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
9823 via stack temporaries. Treat the mode size as polynomial too.
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 * expr.c (expand_expr_real_2): When handling conversions involving
9830 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
9831 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
9832 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
9840 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9841 Alan Hayward <alan.hayward@arm.com>
9842 David Sherwood <david.sherwood@arm.com>
9844 * combine.c (can_change_dest_mode): Handle polynomial
9845 REGMODE_NATURAL_SIZE.
9846 * expmed.c (store_bit_field_1): Likewise.
9847 * expr.c (store_constructor): Likewise.
9848 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
9849 and polynomial REGMODE_NATURAL_SIZE.
9850 (gen_lowpart_common): Likewise.
9851 * reginfo.c (record_subregs_of_mode): Likewise.
9852 * rtlanal.c (read_modify_subreg_p): Likewise.
9854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9855 Alan Hayward <alan.hayward@arm.com>
9856 David Sherwood <david.sherwood@arm.com>
9858 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
9859 numbers of elements.
9861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9862 Alan Hayward <alan.hayward@arm.com>
9863 David Sherwood <david.sherwood@arm.com>
9865 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
9872 in a POINTER_PLUS_EXPR.
9874 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9875 Alan Hayward <alan.hayward@arm.com>
9876 David Sherwood <david.sherwood@arm.com>
9878 * omp-simd-clone.c (simd_clone_subparts): New function.
9879 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
9880 (ipa_simd_modify_function_body): Likewise.
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 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
9887 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
9888 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
9889 (expand_vector_condition, vector_element): Likewise.
9890 (subparts_gt): New function.
9891 (get_compute_type): Use subparts_gt.
9892 (count_type_subparts): Delete.
9893 (expand_vector_operations_1): Use subparts_gt instead of
9894 count_type_subparts.
9896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9897 Alan Hayward <alan.hayward@arm.com>
9898 David Sherwood <david.sherwood@arm.com>
9900 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
9901 (vect_compile_time_alias): ...this new function. Do the calculation
9902 on poly_ints rather than trees.
9903 (vect_prune_runtime_alias_test_list): Update call accordingly.
9905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9906 Alan Hayward <alan.hayward@arm.com>
9907 David Sherwood <david.sherwood@arm.com>
9909 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
9910 numbers of units.
9911 (vect_schedule_slp_instance): Likewise.
9913 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9914 Alan Hayward <alan.hayward@arm.com>
9915 David Sherwood <david.sherwood@arm.com>
9917 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
9918 constant and extern definitions for variable-length vectors.
9919 (vect_get_constant_vectors): Note that the number of units
9920 is known to be constant.
9922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9923 Alan Hayward <alan.hayward@arm.com>
9924 David Sherwood <david.sherwood@arm.com>
9926 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
9927 of units as polynomial. Choose between WIDE and NARROW based
9928 on multiple_p.
9930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9931 Alan Hayward <alan.hayward@arm.com>
9932 David Sherwood <david.sherwood@arm.com>
9934 * tree-vect-stmts.c (simd_clone_subparts): New function.
9935 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
9937 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9938 Alan Hayward <alan.hayward@arm.com>
9939 David Sherwood <david.sherwood@arm.com>
9941 * tree-vect-stmts.c (vectorizable_call): Treat the number of
9942 vectors as polynomial. Use build_index_vector for
9943 IFN_GOMP_SIMD_LANE.
9945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9946 Alan Hayward <alan.hayward@arm.com>
9947 David Sherwood <david.sherwood@arm.com>
9949 * tree-vect-stmts.c (get_load_store_type): Treat the number of
9950 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
9951 for variable-length vectors.
9952 (vectorizable_mask_load_store): Treat the number of units as
9953 polynomial, asserting that it is constant if the condition has
9954 already been enforced.
9955 (vectorizable_store, vectorizable_load): Likewise.
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 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
9962 of units as polynomial. Punt if we can't tell at compile time
9963 which vector contains the final result.
9965 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9966 Alan Hayward <alan.hayward@arm.com>
9967 David Sherwood <david.sherwood@arm.com>
9969 * tree-vect-loop.c (vectorizable_induction): Treat the number
9970 of units as polynomial. Punt on SLP inductions. Use an integer
9971 VEC_SERIES_EXPR for variable-length integer reductions. Use a
9972 cast of such a series for variable-length floating-point
9973 reductions.
9975 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9976 Alan Hayward <alan.hayward@arm.com>
9977 David Sherwood <david.sherwood@arm.com>
9979 * tree.h (build_index_vector): Declare.
9980 * tree.c (build_index_vector): New function.
9981 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
9982 of units as polynomial, forcibly converting it to a constant if
9983 vectorizable_reduction has already enforced the condition.
9984 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
9985 to create a {1,2,3,...} vector.
9986 (vectorizable_reduction): Treat the number of units as polynomial.
9987 Choose vectype_in based on the largest scalar element size rather
9988 than the smallest number of units. Enforce the restrictions
9989 relied on above.
9991 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9992 Alan Hayward <alan.hayward@arm.com>
9993 David Sherwood <david.sherwood@arm.com>
9995 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
9996 number of units as polynomial.
9998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
9999 Alan Hayward <alan.hayward@arm.com>
10000 David Sherwood <david.sherwood@arm.com>
10002 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
10003 * target.def (autovectorize_vector_sizes): Return the vector sizes
10004 by pointer, using vector_sizes rather than a bitmask.
10005 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
10006 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
10007 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
10008 Likewise.
10009 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
10010 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
10011 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
10012 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
10013 * omp-general.c (omp_max_vf): Likewise.
10014 * omp-low.c (omp_clause_aligned_alignment): Likewise.
10015 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
10016 * tree-vect-loop.c (vect_analyze_loop): Likewise.
10017 * tree-vect-slp.c (vect_slp_bb): Likewise.
10018 * doc/tm.texi: Regenerate.
10019 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
10020 to a poly_uint64.
10021 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
10022 the vector size as a poly_uint64 rather than an unsigned int.
10023 (current_vector_size): Change from an unsigned int to a poly_uint64.
10024 (get_vectype_for_scalar_type): Update accordingly.
10025 * tree.h (build_truth_vector_type): Take the size and number of
10026 units as a poly_uint64 rather than an unsigned int.
10027 (build_vector_type): Add a temporary overload that takes
10028 the number of units as a poly_uint64 rather than an unsigned int.
10029 * tree.c (make_vector_type): Likewise.
10030 (build_truth_vector_type): Take the number of units as a poly_uint64
10031 rather than an unsigned int.
10033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10034 Alan Hayward <alan.hayward@arm.com>
10035 David Sherwood <david.sherwood@arm.com>
10037 * target.def (get_mask_mode): Take the number of units and length
10038 as poly_uint64s rather than unsigned ints.
10039 * targhooks.h (default_get_mask_mode): Update accordingly.
10040 * targhooks.c (default_get_mask_mode): Likewise.
10041 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
10042 * doc/tm.texi: Regenerate.
10044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10045 Alan Hayward <alan.hayward@arm.com>
10046 David Sherwood <david.sherwood@arm.com>
10048 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
10049 * omp-general.c (omp_max_vf): Likewise.
10050 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
10051 (expand_omp_simd): Handle polynomial safelen.
10052 * omp-low.c (omplow_simd_context): Add a default constructor.
10053 (omplow_simd_context::max_vf): Change from int to poly_uint64.
10054 (lower_rec_simd_input_clauses): Update accordingly.
10055 (lower_rec_input_clauses): Likewise.
10057 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10058 Alan Hayward <alan.hayward@arm.com>
10059 David Sherwood <david.sherwood@arm.com>
10061 * tree-vectorizer.h (vect_nunits_for_cost): New function.
10062 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
10063 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
10064 (vect_analyze_slp_cost): Likewise.
10065 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
10066 (vect_model_load_cost): Likewise.
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-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
10073 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
10074 from an unsigned int * to a poly_uint64_pod *.
10075 (calculate_unrolling_factor): New function.
10076 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
10078 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10079 Alan Hayward <alan.hayward@arm.com>
10080 David Sherwood <david.sherwood@arm.com>
10082 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
10083 from an unsigned int to a poly_uint64.
10084 (_loop_vec_info::slp_unrolling_factor): Likewise.
10085 (_loop_vec_info::vectorization_factor): Change from an int
10086 to a poly_uint64.
10087 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
10088 (vect_get_num_vectors): New function.
10089 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
10090 (vect_get_num_copies): Use vect_get_num_vectors.
10091 (vect_analyze_data_ref_dependences): Change max_vf from an int *
10092 to an unsigned int *.
10093 (vect_analyze_data_refs): Change min_vf from an int * to a
10094 poly_uint64 *.
10095 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10096 than an unsigned HOST_WIDE_INT.
10097 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10098 (vect_analyze_data_ref_dependence): Change max_vf from an int *
10099 to an unsigned int *.
10100 (vect_analyze_data_ref_dependences): Likewise.
10101 (vect_compute_data_ref_alignment): Handle polynomial vf.
10102 (vect_enhance_data_refs_alignment): Likewise.
10103 (vect_prune_runtime_alias_test_list): Likewise.
10104 (vect_shift_permute_load_chain): Likewise.
10105 (vect_supportable_dr_alignment): Likewise.
10106 (dependence_distance_ge_vf): Take the vectorization factor as a
10107 poly_uint64 rather than an unsigned HOST_WIDE_INT.
10108 (vect_analyze_data_refs): Change min_vf from an int * to a
10109 poly_uint64 *.
10110 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
10111 vfm1 as a poly_uint64 rather than an int. Make the same change
10112 for the returned bound_scalar.
10113 (vect_gen_vector_loop_niters): Handle polynomial vf.
10114 (vect_do_peeling): Likewise. Update call to
10115 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
10116 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
10117 be constant.
10118 * tree-vect-loop.c (vect_determine_vectorization_factor)
10119 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
10120 (vect_get_known_peeling_cost): Likewise.
10121 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
10122 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
10123 (vect_transform_loop): Likewise. Use the lowest possible VF when
10124 updating the upper bounds of the loop.
10125 (vect_min_worthwhile_factor): Make static. Return an unsigned int
10126 rather than an int.
10127 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
10128 polynomial unroll factors.
10129 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
10130 (vect_make_slp_decision): Likewise.
10131 (vect_supported_load_permutation_p): Likewise, and polynomial
10132 vf too.
10133 (vect_analyze_slp_cost): Handle polynomial vf.
10134 (vect_slp_analyze_node_operations): Likewise.
10135 (vect_slp_analyze_bb_1): Likewise.
10136 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
10137 than an unsigned HOST_WIDE_INT.
10138 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
10139 (vectorizable_load): Handle polynomial vf.
10140 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
10141 a poly_uint64.
10142 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
10144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10145 Alan Hayward <alan.hayward@arm.com>
10146 David Sherwood <david.sherwood@arm.com>
10148 * match.pd: Handle bit operations involving three constants
10149 and try to fold one pair.
10151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
10153 * tree-vect-loop-manip.c: Include gimple-fold.h.
10154 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
10155 niters_maybe_zero parameters. Handle other cases besides a step of 1.
10156 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
10157 Add a path that uses a step of VF instead of 1, but disable it
10158 for now.
10159 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
10160 and niters_no_overflow parameters. Update calls to
10161 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
10162 Create a new SSA name if the latter choses to use a ste other
10163 than zero, and return it via niters_vector_mult_vf_var.
10164 * tree-vect-loop.c (vect_transform_loop): Update calls to
10165 vect_do_peeling, vect_gen_vector_loop_niters and
10166 slpeel_make_loop_iterate_ntimes.
10167 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
10168 (vect_gen_vector_loop_niters): Update declarations after above changes.
10170 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
10172 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
10173 128-bit round to integer instructions.
10174 (ceil<mode>2): Likewise.
10175 (btrunc<mode>2): Likewise.
10176 (round<mode>2): Likewise.
10178 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10180 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
10181 unaligned VSX load/store on P8/P9.
10182 (expand_block_clear): Allow the use of unaligned VSX
10183 load/store on P8/P9.
10185 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10187 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
10188 New function.
10189 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
10190 swap associated with both a load and a store.
10192 2018-01-02 Andrew Waterman <andrew@sifive.com>
10194 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
10195 * config/riscv/riscv.md (clear_cache): Use it.
10197 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
10199 * web.c: Remove out-of-date comment.
10201 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10203 * expr.c (fixup_args_size_notes): Check that any existing
10204 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
10205 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
10206 (emit_single_push_insn): ...here.
10208 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10210 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
10211 (const_vector_encoded_nelts): New function.
10212 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
10213 (const_vector_int_elt, const_vector_elt): Declare.
10214 * emit-rtl.c (const_vector_int_elt_1): New function.
10215 (const_vector_elt): Likewise.
10216 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
10217 of CONST_VECTOR_ELT.
10219 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10221 * expr.c: Include rtx-vector-builder.h.
10222 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
10223 directly on the tree encoding.
10224 (const_vector_from_tree): Likewise.
10225 * optabs.c: Include rtx-vector-builder.h.
10226 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
10227 sequence of "u" values.
10228 * vec-perm-indices.c: Include rtx-vector-builder.h.
10229 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
10230 directly on the vec_perm_indices encoding.
10232 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10234 * doc/rtl.texi (const_vector): Describe new encoding scheme.
10235 * Makefile.in (OBJS): Add rtx-vector-builder.o.
10236 * rtx-vector-builder.h: New file.
10237 * rtx-vector-builder.c: Likewise.
10238 * rtl.h (rtx_def::u2): Add a const_vector field.
10239 (CONST_VECTOR_NPATTERNS): New macro.
10240 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
10241 (CONST_VECTOR_DUPLICATE_P): Likewise.
10242 (CONST_VECTOR_STEPPED_P): Likewise.
10243 (CONST_VECTOR_ENCODED_ELT): Likewise.
10244 (const_vec_duplicate_p): Check for a duplicated vector encoding.
10245 (unwrap_const_vec_duplicate): Likewise.
10246 (const_vec_series_p): Check for a non-duplicated vector encoding.
10247 Say that the function only returns true for integer vectors.
10248 * emit-rtl.c: Include rtx-vector-builder.h.
10249 (gen_const_vec_duplicate_1): Delete.
10250 (gen_const_vector): Call gen_const_vec_duplicate instead of
10251 gen_const_vec_duplicate_1.
10252 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
10253 (gen_const_vec_duplicate): Use rtx_vector_builder.
10254 (gen_const_vec_series): Likewise.
10255 (gen_rtx_CONST_VECTOR): Likewise.
10256 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
10257 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10258 Build a new vector rather than modifying a CONST_VECTOR in-place.
10259 (handle_special_swappables): Update call accordingly.
10260 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
10261 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
10262 Build a new vector rather than modifying a CONST_VECTOR in-place.
10263 (handle_special_swappables): Update call accordingly.
10265 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10267 * simplify-rtx.c (simplify_const_binary_operation): Use
10268 CONST_VECTOR_ELT instead of XVECEXP.
10270 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10272 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
10273 the selector elements to be different from the data elements
10274 if the selector is a VECTOR_CST.
10275 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
10276 ssizetype for the selector.
10278 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10280 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
10281 before testing each element individually.
10282 * tree-vect-generic.c (lower_vec_perm): Likewise.
10284 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10286 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
10287 * selftest-run-tests.c (selftest::run_tests): Call it.
10288 * vector-builder.h (vector_builder::operator ==): New function.
10289 (vector_builder::operator !=): Likewise.
10290 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
10291 (vec_perm_indices::all_from_input_p): New function.
10292 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
10293 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
10294 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
10295 instead of reading the VECTOR_CST directly. Detect whether both
10296 vector inputs are the same before constructing the vec_perm_indices,
10297 and update the number of inputs argument accordingly. Use the
10298 utility functions added above. Only construct sel2 if we need to.
10300 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10302 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
10303 the broadcast of the low byte.
10304 (expand_mult_highpart): Use an explicit encoding for the permutes.
10305 * optabs-query.c (can_mult_highpart_p): Likewise.
10306 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
10307 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10308 (vectorizable_bswap): Likewise.
10309 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
10310 explicit encoding for the power-of-2 permutes.
10311 (vect_permute_store_chain): Likewise.
10312 (vect_grouped_load_supported): Likewise.
10313 (vect_permute_load_chain): Likewise.
10315 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10317 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
10318 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
10319 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
10320 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
10321 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
10322 (vect_gen_perm_mask_any): Likewise.
10324 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10326 * int-vector-builder.h: New file.
10327 * vec-perm-indices.h: Include int-vector-builder.h.
10328 (vec_perm_indices): Redefine as an int_vector_builder.
10329 (auto_vec_perm_indices): Delete.
10330 (vec_perm_builder): Redefine as a stand-alone class.
10331 (vec_perm_indices::vec_perm_indices): New function.
10332 (vec_perm_indices::clamp): Likewise.
10333 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
10334 (vec_perm_indices::new_vector): New function.
10335 (vec_perm_indices::new_expanded_vector): Update for new
10336 vec_perm_indices class.
10337 (vec_perm_indices::rotate_inputs): New function.
10338 (vec_perm_indices::all_in_range_p): Operate directly on the
10339 encoded form, without computing elided elements.
10340 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
10341 encoding. Update for new vec_perm_indices class.
10342 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
10343 the given vec_perm_builder.
10344 (expand_vec_perm_var): Update vec_perm_builder constructor.
10345 (expand_mult_highpart): Use vec_perm_builder instead of
10346 auto_vec_perm_indices.
10347 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
10348 vec_perm_indices instead of auto_vec_perm_indices. Use a single
10349 or double series encoding as appropriate.
10350 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
10351 vec_perm_indices instead of auto_vec_perm_indices.
10352 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10353 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10354 (vect_permute_store_chain): Likewise.
10355 (vect_grouped_load_supported): Likewise.
10356 (vect_permute_load_chain): Likewise.
10357 (vect_shift_permute_load_chain): Likewise.
10358 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10359 (vect_transform_slp_perm_load): Likewise.
10360 (vect_schedule_slp_instance): Likewise.
10361 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10362 (vectorizable_mask_load_store): Likewise.
10363 (vectorizable_bswap): Likewise.
10364 (vectorizable_store): Likewise.
10365 (vectorizable_load): Likewise.
10366 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
10367 vec_perm_indices instead of auto_vec_perm_indices. Use
10368 tree_to_vec_perm_builder to read the vector from a tree.
10369 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
10370 vec_perm_builder instead of a vec_perm_indices.
10371 (have_whole_vector_shift): Use vec_perm_builder and
10372 vec_perm_indices instead of auto_vec_perm_indices. Leave the
10373 truncation to calc_vec_perm_mask_for_shift.
10374 (vect_create_epilog_for_reduction): Likewise.
10375 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
10376 from auto_vec_perm_indices to vec_perm_indices.
10377 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10378 instead of changing individual elements.
10379 (aarch64_vectorize_vec_perm_const): Use new_vector to install
10380 the vector in d.perm.
10381 * config/arm/arm.c (expand_vec_perm_d::perm): Change
10382 from auto_vec_perm_indices to vec_perm_indices.
10383 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
10384 instead of changing individual elements.
10385 (arm_vectorize_vec_perm_const): Use new_vector to install
10386 the vector in d.perm.
10387 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
10388 Update vec_perm_builder constructor.
10389 (rs6000_expand_interleave): Likewise.
10390 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
10391 (rs6000_expand_interleave): Likewise.
10393 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10395 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
10396 to qimode could truncate the indices.
10397 * optabs.c (expand_vec_perm_var): Likewise.
10399 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10401 * Makefile.in (OBJS): Add vec-perm-indices.o.
10402 * vec-perm-indices.h: New file.
10403 * vec-perm-indices.c: Likewise.
10404 * target.h (vec_perm_indices): Replace with a forward class
10405 declaration.
10406 (auto_vec_perm_indices): Move to vec-perm-indices.h.
10407 * optabs.h: Include vec-perm-indices.h.
10408 (expand_vec_perm): Delete.
10409 (selector_fits_mode_p, expand_vec_perm_var): Declare.
10410 (expand_vec_perm_const): Declare.
10411 * target.def (vec_perm_const_ok): Replace with...
10412 (vec_perm_const): ...this new hook.
10413 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
10414 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
10415 * doc/tm.texi: Regenerate.
10416 * optabs.def (vec_perm_const): Delete.
10417 * doc/md.texi (vec_perm_const): Likewise.
10418 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
10419 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
10420 expand_vec_perm for constant permutation vectors. Assert that
10421 the mode of variable permutation vectors is the integer equivalent
10422 of the mode that is being permuted.
10423 * optabs-query.h (selector_fits_mode_p): Declare.
10424 * optabs-query.c: Include vec-perm-indices.h.
10425 (selector_fits_mode_p): New function.
10426 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
10427 is defined, instead of checking whether the vec_perm_const_optab
10428 exists. Use targetm.vectorize.vec_perm_const instead of
10429 targetm.vectorize.vec_perm_const_ok. Check whether the indices
10430 fit in the vector mode before using a variable permute.
10431 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
10432 vec_perm_indices instead of an rtx.
10433 (expand_vec_perm): Replace with...
10434 (expand_vec_perm_const): ...this new function. Take the selector
10435 as a vec_perm_indices rather than an rtx. Also take the mode of
10436 the selector. Update call to shift_amt_for_vec_perm_mask.
10437 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
10438 Use vec_perm_indices::new_expanded_vector to expand the original
10439 selector into bytes. Check whether the indices fit in the vector
10440 mode before using a variable permute.
10441 (expand_vec_perm_var): Make global.
10442 (expand_mult_highpart): Use expand_vec_perm_const.
10443 * fold-const.c: Includes vec-perm-indices.h.
10444 * tree-ssa-forwprop.c: Likewise.
10445 * tree-vect-data-refs.c: Likewise.
10446 * tree-vect-generic.c: Likewise.
10447 * tree-vect-loop.c: Likewise.
10448 * tree-vect-slp.c: Likewise.
10449 * tree-vect-stmts.c: Likewise.
10450 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
10451 Delete.
10452 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
10453 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
10454 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
10455 (aarch64_vectorize_vec_perm_const): ...this new function.
10456 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10457 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10458 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
10459 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
10460 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10461 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10462 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
10463 into...
10464 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
10465 check for NEON modes.
10466 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
10467 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
10468 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
10469 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
10470 into...
10471 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
10472 the old VEC_PERM_CONST conditions.
10473 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
10474 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
10475 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
10476 (ia64_vectorize_vec_perm_const_ok): Merge into...
10477 (ia64_vectorize_vec_perm_const): ...this new function.
10478 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
10479 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
10480 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
10481 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
10482 * config/mips/mips.c (mips_expand_vec_perm_const)
10483 (mips_vectorize_vec_perm_const_ok): Merge into...
10484 (mips_vectorize_vec_perm_const): ...this new function.
10485 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
10486 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
10487 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
10488 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
10489 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
10490 (rs6000_expand_vec_perm_const): Delete.
10491 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
10492 Delete.
10493 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10494 (altivec_expand_vec_perm_const_le): Take each operand individually.
10495 Operate on constant selectors rather than rtxes.
10496 (altivec_expand_vec_perm_const): Likewise. Update call to
10497 altivec_expand_vec_perm_const_le.
10498 (rs6000_expand_vec_perm_const): Delete.
10499 (rs6000_vectorize_vec_perm_const_ok): Delete.
10500 (rs6000_vectorize_vec_perm_const): New function.
10501 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10502 an element count and rtx array.
10503 (rs6000_expand_extract_even): Update call accordingly.
10504 (rs6000_expand_interleave): Likewise.
10505 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
10506 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
10507 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
10508 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
10509 (rs6000_expand_vec_perm_const): Delete.
10510 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
10511 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10512 (altivec_expand_vec_perm_const_le): Take each operand individually.
10513 Operate on constant selectors rather than rtxes.
10514 (altivec_expand_vec_perm_const): Likewise. Update call to
10515 altivec_expand_vec_perm_const_le.
10516 (rs6000_expand_vec_perm_const): Delete.
10517 (rs6000_vectorize_vec_perm_const_ok): Delete.
10518 (rs6000_vectorize_vec_perm_const): New function. Remove stray
10519 reference to the SPE evmerge intructions.
10520 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
10521 an element count and rtx array.
10522 (rs6000_expand_extract_even): Update call accordingly.
10523 (rs6000_expand_interleave): Likewise.
10524 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
10525 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
10526 new function.
10527 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
10529 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10531 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
10532 vector mode and that that mode matches the mode of the data
10533 being permuted.
10534 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
10535 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
10536 directly using expand_vec_perm_1 when forcing selectors into
10537 registers.
10538 (expand_vec_perm_var): New function, split out from expand_vec_perm.
10540 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10542 * optabs-query.h (can_vec_perm_p): Delete.
10543 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
10544 * optabs-query.c (can_vec_perm_p): Split into...
10545 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
10546 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
10547 particular selector is valid.
10548 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
10549 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
10550 (vect_grouped_load_supported): Likewise.
10551 (vect_shift_permute_load_chain): Likewise.
10552 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
10553 (vect_transform_slp_perm_load): Likewise.
10554 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
10555 (vectorizable_bswap): Likewise.
10556 (vect_gen_perm_mask_checked): Likewise.
10557 * fold-const.c (fold_ternary_loc): Likewise. Don't take
10558 implementations of variable permutation vectors into account
10559 when deciding which selector to use.
10560 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
10561 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
10562 with a false third argument.
10563 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
10564 to test whether the constant selector is valid and can_vec_perm_var_p
10565 to test whether a variable selector is valid.
10567 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10569 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
10570 * optabs-query.c (can_vec_perm_p): Likewise.
10571 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
10572 instead of vec_perm_indices.
10573 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
10574 (vect_gen_perm_mask_checked): Likewise,
10575 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
10576 (vect_gen_perm_mask_checked): Likewise,
10578 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
10580 * optabs-query.h (qimode_for_vec_perm): Declare.
10581 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
10582 (qimode_for_vec_perm): ...this new function.
10583 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
10585 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
10587 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
10588 does not have a conditional at the top.
10590 2018-01-02 Richard Biener <rguenther@suse.de>
10592 * ipa-inline.c (big_speedup_p): Fix expression.
10594 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10596 PR target/81616
10597 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
10598 for generic 4->6.
10600 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
10602 PR target/81616
10603 Generic tuning.
10604 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
10605 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
10606 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
10607 cond_taken_branch_cost 3->4.
10609 2018-01-01 Jakub Jelinek <jakub@redhat.com>
10611 PR tree-optimization/83581
10612 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
10613 TODO_cleanup_cfg if any changes have been made.
10615 PR middle-end/83608
10616 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
10617 convert_modes if target mode has the right side, but different mode
10618 class.
10620 PR middle-end/83609
10621 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
10622 last argument when extracting from CONCAT. If either from_real or
10623 from_imag is NULL, use expansion through memory. If result is not
10624 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
10625 the parts directly to inner mode, if even that fails, use expansion
10626 through memory.
10628 PR middle-end/83623
10629 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
10630 check for bswap in mode rather than HImode and use that in expand_unop
10631 too.
10632 \f
10633 Copyright (C) 2018 Free Software Foundation, Inc.
10635 Copying and distribution of this file, with or without modification,
10636 are permitted in any medium without royalty provided the copyright
10637 notice and this notice are preserved.