1 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
4 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
5 when the function accesses prior frames.
7 2018-03-08 Jakub Jelinek <jakub@redhat.com>
10 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
11 gen_llsym, otherwise call maybe_gen_llsym.
14 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
15 has ',' character inside of it.
17 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
21 as clobbering CC_REGNUM.
23 2018-03-08 Richard Biener <rguenther@suse.de>
26 * tree-scalar-evolution.c: Include tree-into-ssa.h.
27 (follow_copies_to_constant): Do not follow SSA names registered
30 2018-03-08 Richard Biener <rguenther@suse.de>
32 PR tree-optimization/84178
33 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
35 (version_loop_for_if_conversion): Delay update_ssa call.
36 (tree_if_conversion): Delay update_ssa until after predicate
39 2018-03-08 David Malcolm <dmalcolm@redhat.com>
41 PR tree-optimization/84178
42 * tree-if-conv.c (release_bb_predicate): Remove the
43 the assertion that the stmts have NULL use_ops.
44 Discard the statements, asserting that they haven't
45 yet been added to a BB.
47 2018-03-08 Richard Biener <rguenther@suse.de>
49 PR tree-optimization/84746
50 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
51 (phi_translate): Pass in destination ANTIC_OUT set.
52 (phi_translate_1): Likewise. For a simplified result lookup
53 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
54 (phi_translate_set): Adjust.
55 (do_pre_regular_insertion): Likewise.
56 (do_pre_partial_partial_insertion): Likewise.
58 2018-03-08 Martin Liska <mliska@suse.cz>
61 * doc/gcov.texi: Document usage of profile files.
62 * gcov-io.h: Document changes in the format.
64 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
68 * dwarf2out.c (struct dw_line_info_table): Update comments for
70 (FORCE_RESET_NEXT_VIEW): New.
71 (FORCE_RESETTING_VIEW_P): New.
72 (RESETTING_VIEW_P): Check for -1 too.
73 (ZERO_VIEW_P): Likewise.
74 (new_line_info_table): Force-reset next view.
75 (dwarf2out_begin_function): Likewise.
76 (dwarf2out_source_line): Simplify zero_view_p initialization.
77 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
78 view directly. Omit view when omitting .loc at line 0.
80 2018-03-08 Jakub Jelinek <jakub@redhat.com>
82 PR tree-optimization/84740
83 * tree-switch-conversion.c (process_switch): Call build_constructors
84 only if info.phi_count is non-zero.
86 PR tree-optimization/84739
87 * tree-tailcall.c (find_tail_calls): Check call arguments against
88 DECL_ARGUMENTS (current_function_decl) rather than
89 DECL_ARGUMENTS (func) when checking for tail recursion.
91 2018-03-07 Jakub Jelinek <jakub@redhat.com>
93 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
94 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
95 Volker Reichelt's entry and add entries for people that perform
96 GCC fuzzy testing and report numerous bugs.
98 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
101 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
102 readonly data in sdata, if that is disabled.
103 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
104 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
105 -mreadonly-in-sdata option.
107 2018-03-07 Martin Sebor <msebor@redhat.com>
109 PR tree-optimization/84468
110 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
111 basic block when looking for nul assignment.
113 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
116 * except.h (output_function_exception_table): Adjust prototype.
117 * except.c (output_function_exception_table): Remove FNNAME parameter
118 and add SECTION parameter. Ouput one part of the table at a time.
119 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
120 the first part of the exception table and emit unwind directives.
121 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
122 (i386_pe_seh_cold_init): Likewise.
123 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
124 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
125 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
126 (ix86_output_call_insn): Emit a nop in one more case for SEH.
127 * config/i386/winnt.c: Include except.h.
128 (struct seh_frame_state): Add reg_offset, after_prologue and
129 in_cold_section fields.
130 (i386_pe_seh_end_prologue): Set seh->after_prologue.
131 (i386_pe_seh_cold_init): New function.
132 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
133 to seh->in_cold_section.
134 (seh_emit_push): Record the offset of the push.
135 (seh_emit_save): Record the offet of the save.
136 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
137 Test seh->after_prologue to disregard the epilogue.
138 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
139 (i386_pe_end_cold_function): New function.
141 2018-03-07 Jakub Jelinek <jakub@redhat.com>
144 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
145 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
148 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
150 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
151 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
154 * multiple_target.c: Include tree-inline.h and intl.h.
155 (expand_target_clones): Diagnose and fail if node->definition and
156 !tree_versionable_function_p (node->decl).
158 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
160 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
162 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
163 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
164 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
166 2018-03-06 Jakub Jelinek <jakub@redhat.com>
169 * combine.c (try_combine): Use reg_or_subregno instead of handling
170 just paradoxical SUBREGs and REGs.
172 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
174 * config/arc/arc.c (arc_finalize_pic): Remove function.
175 (arc_must_save_register): We use single base PIC register, remove
176 checks to save/restore the PIC register.
177 (arc_expand_prologue): Likewise.
178 * config/arc/arc-protos.h (arc_set_default_type_attributes):
180 (arc_verify_short): Likewise.
181 (arc_attr_type): Likewise.
182 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
183 (walk_stores): Likewise.
184 (arc_address_cost): Make it static.
185 (arc_verify_short): Likewise.
186 (branch_dest): Likewise.
187 (arc_attr_type): Likewise.
188 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
189 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
190 (arc_final_prescan_insn): Remove inserting the nops due to
191 hardware hazards. It is done in reorg step.
192 (insn_length_variant_t): Remove.
193 (insn_length_parameters_t): Likewise.
194 (arc_insn_length_parameters): Likewise.
195 (arc_get_insn_variants): Likewise.
196 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
198 2018-03-06 Jakub Jelinek <jakub@redhat.com>
201 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
204 PR tree-optimization/84687
205 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
207 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
209 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
211 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
212 Rename to ppc_speculation_barrier.
213 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
214 __builtin_ppc_speculation_barrier.
216 2018-03-05 Jakub Jelinek <jakub@redhat.com>
219 * combine.c (combine_simplify_rtx): Don't try to simplify if
220 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
223 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
225 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
226 to 32 bytes when compiling for POWER9.
228 2018-03-05 Jakub Jelinek <jakub@redhat.com>
231 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
232 regparm >= 3 with no arg reg available also for calls with
233 flag_force_indirect_call. Pass decl to ix86_function_regparm.
236 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
238 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
240 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
243 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
245 2018-03-05 Richard Biener <rguenther@suse.de>
247 PR tree-optimization/84486
248 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
249 When inserting a __builtin_assume_aligned call set the LHS
250 SSA name alignment info accordingly.
252 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
254 PR tree-optimization/84114
255 * config/aarch64/aarch64.c (aarch64_reassociation_width)
256 Avoid reassociation of FLOAT_MODE addition.
258 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
260 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
261 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
262 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
263 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
264 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
265 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
266 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
268 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
269 __builtin_ia32_wbinvd): New builtins.
270 (SPECIAL_ARGS2): New.
271 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
272 (SPECIAL_ARGS2): New.
273 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
274 (ix86_valid_target_attribute_inner_p): Ditto.
275 (ix86_init_mmx_sse_builtins): Add special_args2.
276 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
277 TARGET_WBNOINVD_P): New.
278 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
279 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
280 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
281 * config/i386/immintrin.h (_wbinvd): New intrinsic.
282 * config/i386/pconfigintrin.h: New file.
283 * config/i386/wbnoinvdintrin.h: Ditto.
284 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
285 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
287 2018-03-05 Richard Biener <rguenther@suse.de>
289 PR tree-optimization/84670
290 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
292 (BB_VISITED_WITH_VISITED_SUCCS): New define.
293 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
294 (compute_antic_aux): Only assert the number of values in ANTIC_IN
295 doesn't grow if all successors (recursively) were visited at least
298 2018-03-05 Richard Biener <rguenther@suse.de>
300 PR tree-optimization/84650
301 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
302 if executed in the loop pipeline.
304 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
306 * doc/configfiles.texi (Configuration Files): Move info about
307 conditionalizing $target-protos.h to...
308 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
309 differs from $target-protos.h.
311 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
312 Chung-Ju Wu <jasonwucj@gmail.com>
314 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
315 * config/nds32/nds32-multiple.md (setmemsi): Define.
316 * config/nds32/nds32-memory-manipulation.c
317 (nds32_gen_dup_4_byte_to_word_value): New.
318 (emit_setmem_word_loop): New.
319 (emit_setmem_byte_loop): New.
320 (nds32_expand_setmem_loop): New.
321 (nds32_expand_setmem_loop_v3m): New.
322 (nds32_expand_setmem_unroll): New.
323 (nds32_expand_setmem): New.
325 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
326 Chung-Ju Wu <jasonwucj@gmail.com>
328 * config/nds32/nds32-memory-manipulation.c
329 (nds32_emit_load_store): New.
330 (nds32_emit_post_inc_load_store): New.
331 (nds32_emit_mem_move): New.
332 (nds32_emit_mem_move_block): New.
333 (nds32_expand_movmemsi_loop_unknown_size): New.
334 (nds32_expand_movmemsi_loop_known_size): New.
335 (nds32_expand_movmemsi_loop): New.
336 (nds32_expand_movmemsi_unroll): New.
337 (nds32_expand_movmemqi): Rename ...
338 (nds32_expand_movmemsi): ... to this.
339 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
340 (movmemsi): ... to this.
341 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
342 (nds32_expand_movmemsi): ... to this.
344 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
345 Monk Chiang <sh.chiang04@gmail.com>
346 Chung-Ju Wu <jasonwucj@gmail.com>
348 * config/nds32/nds32-protos.h
349 (nds32_expand_load_multiple): New arguments.
350 (nds32_expand_store_multiple): Ditto.
351 (nds32_valid_multiple_load_store): Rename ...
352 (nds32_valid_multiple_load_store_p): ... to this.
353 * config/nds32/nds32-memory-manipulation.c
354 (nds32_expand_load_multiple): Refine implementation.
355 (nds32_expand_store_multiple): Ditto.
356 * config/nds32/nds32-multiple.md
357 (load_multiple): Update nds32_expand_load_multiple interface.
358 (store_multiple): Update nds32_expand_store_multiple interface.
359 * config/nds32/nds32-predicates.c
360 (nds32_valid_multiple_load_store): Rename ...
361 (nds32_valid_multiple_load_store_p): ... to this and refine
363 * config/nds32/predicates.md
364 (nds32_load_multiple_and_update_address_operation): New predicate.
365 (nds32_store_multiple_and_update_address_operation): New predicate.
367 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
368 Chung-Ju Wu <jasonwucj@gmail.com>
370 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
371 (combo): New attribute.
372 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
374 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
376 * config/nds32/nds32.opt: Change -mcmodel= default value.
378 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
379 Monk Chiang <sh.chiang04@gmail.com>
380 Chung-Ju Wu <jasonwucj@gmail.com>
382 * config/nds32/constants.md (unspec_element): New enum.
383 * config/nds32/constraints.md (Umw): New constraint.
384 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
385 * config/nds32/nds32-intrinsic.md: Likewise.
386 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
387 (nds32_valid_smw_lwm_base_p): New.
388 (nds32_output_smw_single_word): New.
389 (nds32_output_lmw_single_word): New.
390 (nds32_expand_unaligned_load): New.
391 (nds32_expand_unaligned_store): New.
392 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
393 (nds32_output_smw_single_word): Declare.
394 (nds32_output_lmw_single_word): Declare.
395 (nds32_expand_unaligned_load): Declare.
396 (nds32_expand_unaligned_store): Declare.
397 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
398 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
399 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
400 NDS32_BUILTIN_UASTORE_DW.
401 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
404 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
405 Kito Cheng <kito.cheng@gmail.com>
406 Chung-Ju Wu <jasonwucj@gmail.com>
408 * config/nds32/nds32-intrinsic.c
409 (nds32_expand_builtin_null_ftype_reg): Delete.
410 (nds32_expand_builtin_reg_ftype_imm): Ditto.
411 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
412 (nds32_read_argument): New.
413 (nds32_legitimize_target): Ditto.
414 (nds32_legitimize_argument): Ditto.
415 (nds32_check_constant_argument): Ditto.
416 (nds32_expand_unop_builtin): Ditto.
417 (nds32_expand_unopimm_builtin): Ditto.
418 (nds32_expand_binop_builtin): Ditto.
419 (nds32_builtin_decl_impl): Ditto.
420 (builtin_description): Ditto.
421 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
422 (nds32_init_builtins_impl): Ditto.
423 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
424 (nds32_builtin_decl): New.
425 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
426 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
428 2018-03-02 Jeff Law <law@redhat.com>
430 * reorg.c (stop_search_p): Handle DEBUG_INSN.
431 (redundant_insn, fill_simple_delay_slots): Likewise.
432 (fill_slots_from_thread): Likewise.
433 * resource.c (mark_referenced_resources): Likewise.
434 (mark_set_resources, find_dead_or_set_registers): Likewise.
436 2018-03-02 Jakub Jelinek <jakub@redhat.com>
438 * substring-locations.h (format_warning_va): Formatting fix for
440 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
442 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
443 * substring-locations.c: Include intl.h.
444 (format_warning_va): Turned into small wrapper around
445 format_warning_n_va, renamed to ...
446 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
447 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
449 (format_warning_at_substring_n): New function.
450 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
451 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
452 format_warning_at_substring with just a shorter name instead of
453 const function pointer.
454 (fmtwarn_n): New function.
455 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
456 appropriate, get rid of all the fmtstr temporaries, move conditionals
457 with G_() wrapped string literals directly into fmtwarn arguments,
458 cast dir.len to (int), formatting fixes.
460 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
462 * doc/invoke.texi: Remove "Cilk Plus" references.
464 2018-03-02 Jakub Jelinek <jakub@redhat.com>
465 Richard Biener <rguenther@suse.de>
468 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
469 for error or warning attributes if CALL_FROM_THUNK_P is set.
472 2018-03-02 Jakub Jelinek <jakub@redhat.com>
475 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
476 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
479 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
480 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
482 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
483 instead of -1U in last predictors element's probability member.
485 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
488 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
489 arguments if they are comparable.
491 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
493 PR tree-optimization/84634
494 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
495 masks and masked_loop_p with a single loop_masks, making sure it's
496 null for bb vectorization.
498 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
500 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
501 (vect_analyze_data_ref_access): Use loop->safe_len rather than
502 loop->force_vectorize to check whether there is no alias.
504 2018-03-02 Jakub Jelinek <jakub@redhat.com>
507 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
509 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
511 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
512 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
513 instead of a loop around prev_real_insn.
514 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
518 * config/i386/i386.c (ix86_print_operand): Use conditional
519 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
522 2018-03-02 Richard Biener <rguenther@suse.de>
524 PR tree-optimization/84427
525 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
526 (bitmap_set_subtract_values): Rewrite to handle multiple
529 (prune_clobbered_mems): Likewise.
530 (phi_translate): Take edge instead of pred/phiblock.
531 (phi_translate_1): Likewise.
532 (phi_translate_set): Likewise. Insert all translated
533 exprs for a value into the set, keeping possibly multiple
534 expressions per value.
535 (compute_antic_aux): Adjust for phi_translate changes.
536 When intersecting union the expressions and prune those
537 not in the final value set, keeping possibly multiple
538 expressions per value. Do not use value-insertion
539 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
540 all expressions. Add verification that the value-sets
541 only shrink during iteration.
542 (compute_partial_antic_aux): Adjust for the phi_translate changes.
543 (do_pre_regular_insertion): Likewise.
544 (do_pre_partial_partial_insertion): Likewise.
546 2018-03-02 Richard Biener <rguenther@suse.de>
549 * config/darwin.c (saved_debug_info_level): New static global.
550 (darwin_asm_lto_start): Disable debug info generation for LTO out.
551 (darwin_asm_lto_end): Restore debug info generation settings.
553 2018-03-01 Martin Liska <mliska@suse.cz>
556 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
559 2018-03-01 Richard Biener <rguenther@suse.de>
562 * dwarf2out.c (gen_variable_die): Properly handle late VLA
563 type annotation with LTO when debug was disabled at compile-time.
565 2018-03-01 Matthew Fortune <mfortune@gmail.com>
567 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
569 (mips_final_postscan_insn): Likewise.
571 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
573 PR rtl-optimization/84528
574 * alias.c (init_alias_target): Add commentary.
575 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
576 a unique base value if the frame pointer is not eliminated
577 to the stack pointer.
579 2018-03-01 Tom de Vries <tom@codesourcery.com>
581 PR rtl-optimization/83327
582 * lra-int.h (hard_regs_spilled_into): Declare.
583 * lra.c (hard_regs_spilled_into): Define.
584 (init_reg_info): Init hard_regs_spilled_into.
585 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
586 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
587 (process_bb_lives): Handle hard_regs_spilled_into.
588 (lra_create_live_ranges_1): Before doing liveness propagation, clear
589 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
591 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
593 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
594 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
595 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
596 * config/rs6000/aix72.h: New file.
598 2018-02-28 Jakub Jelinek <jakub@redhat.com>
600 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
601 instead of warning_at with conditional singular and plural messages
605 * stor-layout.c (update_alignment_for_field): For
606 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
607 && !DECL_PACKED (field), do the alignment update, just use
608 only desired_align instead of MAX (type_align, desired_align)
610 (place_field): Don't do known_align < desired_align handling
611 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
612 is non-NULL, instead do it after rli->prev_field handling and
613 only if not within a bitfield word. For DECL_PACKED (field)
614 use type_align of BITS_PER_UNIT.
616 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
618 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
619 superfluous parentheses and trailing spaces.
621 2018-02-28 Richard Biener <rguenther@suse.de>
623 PR tree-optimization/84584
624 * graphite-scop-detection.c (scop_detection::add_scop): Discard
625 SCoPs with fake exit edge.
627 2018-02-28 Martin Liska <mliska@suse.cz>
630 * timevar.c (timer::print): Fix format to properly print 100%
633 2018-02-28 Richard Biener <rguenther@suse.de>
636 * genmatch.c (capture_info::walk_match): Do not mark
637 captured expressions without operands as expr_p given
638 they act more like predicates and should be subject to
639 "lost tail" side-effect preserving.
641 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
643 PR rtl-optimization/81611
644 * auto-inc-dec.c (attempt_change): Move dead note from
645 mem_insn if it's the next use of regno
646 (find_address): Take address use of reg holding
647 non-incremented value. Add parm to limit search to the named
649 (merge_in_block): Attempt to use a mem insn that is the next
650 use of the original regno.
652 2018-02-27 Martin Sebor <msebor@redhat.com>
655 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
656 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
658 2018-02-27 Martin Sebor <msebor@redhat.com>
661 * diagnostic-core.h (warning_n, error_n, inform_n): Change
662 n argument to unsigned HOST_WIDE_INT.
663 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
664 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
665 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
666 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
668 2018-02-27 Richard Biener <rguenther@suse.de>
670 PR tree-optimization/84512
671 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
672 Do not use the estimate returned from record_stmt_cost for
673 the scalar iteration cost but sum properly using add_stmt_cost.
675 2018-02-27 Richard Biener <rguenther@suse.de>
677 PR tree-optimization/84466
678 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
679 Adjust last change to less strictly validate use operands.
681 2018-02-27 Martin Liska <mliska@suse.cz>
683 PR gcov-profile/84548
684 * gcov.c (process_file): Allow partial overlap and consider it
685 also as group functions.
686 (output_lines): Properly calculate range of lines for a group.
688 2018-02-27 Martin Liska <mliska@suse.cz>
690 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
691 'ggc' suffixes. Change first column width.
692 (timer::print): Fix formatting of the column.
694 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
696 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
697 preserve inline entry blocks for the sake of debug inline
698 entry point markers alone.
699 (remove_unused_locals): Suggest in comments a better place to
700 force the preservation of inline entry blocks that are
701 otherwise unused, but do not preserve them.
703 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
705 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
707 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
710 * config/i386/constraints.md (Bs): Replace
711 ix86_indirect_branch_register with
712 TARGET_INDIRECT_BRANCH_REGISTER.
714 * config/i386/i386.md (indirect_jump): Likewise.
715 (tablejump): Likewise.
716 (*sibcall_memory): Likewise.
717 (*sibcall_value_memory): Likewise.
718 Peepholes of indirect call and jump via memory: Likewise.
719 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
720 (*sibcall_value_GOT_32): Likewise.
721 * config/i386/predicates.md (indirect_branch_operand): Likewise.
722 (GOT_memory_operand): Likewise.
723 (call_insn_operand): Likewise.
724 (sibcall_insn_operand): Likewise.
725 (GOT32_symbol_operand): Likewise.
726 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
728 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
730 PR rtl-optimization/83496
731 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
732 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
733 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
734 redundant insn, if any.
735 (relax_delay_slots): Likewise.
736 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
738 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
740 PR tree-optimization/83965
741 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
742 that grouped statements are part of a reduction chain. Return
743 true if the statement is not marked as a reduction itself but
745 (vect_recog_dot_prod_pattern): Don't check whether the statement
746 is part of a group here.
747 (vect_recog_sad_pattern): Likewise.
748 (vect_recog_widen_sum_pattern): Likewise.
750 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
753 * final.c (rest_of_clean_state): Also look for calls inside sequences.
755 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
758 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
760 (ix86_output_indirect_function_return): New prototype.
761 (ix86_split_simple_return_pop_internal): Likewise.
762 * config/i386/i386.c (indirect_return_via_cx): New.
763 (indirect_return_via_cx_bnd): Likewise.
764 (indirect_thunk_name): Handle return va CX_REG.
765 (output_indirect_thunk_function): Create alias for
766 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
767 (ix86_output_indirect_jmp): Remove the bool argument.
768 (ix86_output_indirect_function_return): New function.
769 (ix86_split_simple_return_pop_internal): Likewise.
770 * config/i386/i386.md (*indirect_jump): Don't pass false
771 to ix86_output_indirect_jmp.
772 (*tablejump_1): Likewise.
773 (simple_return_pop_internal): Change it to define_insn_and_split.
774 Call ix86_split_simple_return_pop_internal to split it for
776 (simple_return_indirect_internal): Call
777 ix86_output_indirect_function_return instead of
778 ix86_output_indirect_jmp.
780 2018-02-26 Jakub Jelinek <jakub@redhat.com>
783 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
784 memset and value initialization afterwards.
786 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
788 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
790 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
793 * common/config/aarch64/aarch64-common.c
794 (aarch_option_optimization_table[]): Switch
795 off fomit-frame-pointer
797 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
798 Chung-Ju Wu <jasonwucj@gmail.com>
800 * config/nds32/nds32-multiple.md (load_multiple): Disallow
802 (store_multiple): Ditto.
804 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
806 * config.gcc: Add --with-cpu support for nds32 target.
807 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
808 * config/nds32/nds32.opt: Add -mcpu= option.
810 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
812 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
813 isel=yes): Warn for these deprecated options.
815 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
817 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
818 ISA_2_5_MASKS_EMBEDDED.
820 2018-02-23 Jakub Jelinek <jakub@redhat.com>
822 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
823 p->max as pointers rather than using iterative_hash_expr.
825 2018-02-23 Carl Love <cel@us.ibm.com>
827 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
828 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
830 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
831 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
832 P8V_BUILTIN_VEC_VUNSIGNED2.
834 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
837 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
838 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
840 * lra-constraints.c (curr_insn_transform): Set up
841 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
843 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
844 * lra-eliminations.c (spill_pseudos): Ditto.
845 (process_insn_for_elimination): Ditto.
846 * lra-lives.c (reg_early_clobber_p): Use the new macros.
847 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
848 LRA_NON_CLOBBERED_ALT.
850 2018-02-22 Martin Sebor <msebor@redhat.com>
852 PR tree-optimization/84480
853 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
854 to maybe_diag_stxncpy_trunc. Call it.
855 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
856 from gimple_fold_builtin_strcpy. Print inlining stack.
857 (handle_builtin_stxncpy): Print inlining stack.
858 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
860 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
863 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
864 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
865 and -fcheck-pointer-bounds are used together.
866 (indirect_thunk_prefix): New enum.
867 (indirect_thunk_need_prefix): New function.
868 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
869 "_nt" instead of "_bnd" for NOTRACK prefix.
870 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
871 (output_indirect_thunk_function): Likewise.
873 (ix86_code_end): Update output_indirect_thunk_function calls.
874 (ix86_output_indirect_branch_via_reg): Replace
875 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
876 (ix86_output_indirect_branch_via_push): Likewise.
877 (ix86_output_function_return): Likewise.
878 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
879 incompatible with -fcf-protection=branch and
880 -fcheck-pointer-bounds.
882 2018-02-22 Steve Ellcey <sellcey@cavium.com>
885 * config/aarch64/aarch64.c (aarch64_print_address_internal):
886 Change gcc_assert call to output_operand_lossage.
888 2018-02-22 Steve Ellcey <sellcey@cavium.com>
890 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
892 2018-02-22 DJ Delorie <dj@redhat.com>
893 Sebastian Perta <sebastian.perta@renesas.com>
894 Oleg Endo <olegendo@gcc.gnu.org>
896 * config/rx/rx.c (rx_rtx_costs): New function.
897 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
899 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
901 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
903 2018-02-22 Martin Liska <mliska@suse.cz>
906 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
907 Add "native" as a possible value.
909 2018-02-22 Martin Liska <mliska@suse.cz>
912 * config/i386/i386.c (ix86_option_override_internal):
913 Add "native" as a possible value for -march and -mtune.
915 2018-02-22 Jakub Jelinek <jakub@redhat.com>
918 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
919 to all type variants.
921 PR tree-optimization/84503
922 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
923 width as info->bitpos + info->bitsize - start.
924 (merged_store_group::merge_overlapping): Simplify width computation.
925 (check_no_overlap): New function.
926 (imm_store_chain_info::try_coalesce_bswap): Compute expected
927 start + width and last_order of the group, fail if check_no_overlap
929 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
930 to group if check_no_overlap fails.
932 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
934 * config/rs6000/altivec.md: Delete contraint arguments to
935 define_expand, define_split, and define_peephole2, and in
936 define_insn_and_split if always unused.
937 * config/rs6000/darwin.md: Ditto.
938 * config/rs6000/dfp.md: Ditto.
939 * config/rs6000/rs6000.md: Ditto.
940 * config/rs6000/sync.md: Ditto.
941 * config/rs6000/vector.md: Ditto.
942 * config/rs6000/vsx.md: Ditto.
944 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
946 * config/rs6000/altivec.md: Write output control strings as braced
947 blocks instead of double-quoted strings.
948 * config/rs6000/darwin.md: Ditto.
949 * config/rs6000/rs6000.md: Ditto.
950 * config/rs6000/vector.md: Ditto.
951 * config/rs6000/vsx.md: Ditto.
953 2018-02-21 Jason Merrill <jason@redhat.com>
955 PR c++/84314 - ICE with templates and fastcall attribute.
956 * attribs.c (build_type_attribute_qual_variant): Remove assert.
958 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
960 * ipa-cp.c (determine_versionability): Fix comment typos.
962 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
965 * ipa-cp.c (determine_versionability): Do not version functions caling
968 2018-02-21 Martin Liska <mliska@suse.cz>
971 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
972 Add "native" as a possible value.
973 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
974 the macro when native cpu detection is available.
976 2018-02-21 Martin Liska <mliska@suse.cz>
979 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
980 Add "native" as a possible value.
981 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
982 when native cpu detection is available.
984 2018-02-21 Jakub Jelinek <jakub@redhat.com>
985 Martin Sebor <msebor@redhat.com>
987 PR tree-optimization/84478
988 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
990 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
991 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
992 support which is conservatively correct, for 2 only stay conservative
993 for maxlen. Formatting and comment capitalization fixes. Add STRICT
994 argument to the 2 argument get_range_strlen, adjust 6 arg
995 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
997 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
998 (gimple_fold_builtin_strlen): Pass true as last argument to
1001 2018-02-20 Martin Sebor <msebor@redhat.com>
1004 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
1005 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1006 (builtin_memref::builtin_memref): Factor out parts into
1007 set_base_and_offset and call it.
1009 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1012 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1013 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1014 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1015 search at the associated MODE_INT.
1017 2018-02-20 Jeff Law <law@redhat.com>
1020 PR tree-optimization/81592
1022 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1023 for range data rather than using global data.
1024 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1025 range data rather than using global data.
1026 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1027 pass it to children as needed.
1028 (struct directive::fmtresult): Similarly.
1029 (struct directive::set_width): Similarly.
1030 (struct directive::set_precision): Similarly.
1031 (format_integer, format_directive, parse_directive): Similarly.
1032 (format_none): Accept unnamed vr_values parameter.
1033 (format_percent, format_floating, format_character): Similarly.
1034 (format_string, format_plain): Similarly.
1035 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1036 the EVRP range analyzer for range data rather than using global data.
1037 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1038 gimple-ssa-evrp-analyze.h
1039 (class sprintf_dom_walker): Add after_dom_children member function.
1040 Add evrp_range_analyzer member.
1041 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1042 range analyzer as needed.
1043 (sprintf_dom_walker::after_dom_children): New member function.
1044 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1046 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1047 (evrp_range_analyzer::pop_to_marker): Likewise.
1049 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1051 PR tree-optimization/84419
1052 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1053 with the required type if its current type is compatible but
1056 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1059 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1060 after vectorization.
1062 2018-02-20 Martin Liska <mliska@suse.cz>
1065 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1066 possible values if we don't have a hint.
1068 2018-02-20 Martin Liska <mliska@suse.cz>
1072 * final.c (shorten_branches): Build align_tab array with one
1074 * opts.c (finish_options): Add alignment option limit check.
1075 (MAX_CODE_ALIGN): Likewise.
1076 (MAX_CODE_ALIGN_VALUE): Likewise.
1077 * doc/invoke.texi: Document maximum allowed option value for
1078 all -falign-* options.
1080 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1083 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1084 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1085 * var-tracking.c (emit_note_insn_var_location): Remove all references
1086 to NOTE_INSN_CALL_ARG_LOCATION.
1087 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1088 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1089 Use copy_rtx_if_shared.
1090 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1091 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1092 (dwarf2out_var_location): Remove handling of
1093 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1095 * final.c (final_scan_insn): Remove all references to
1096 NOTE_INSN_CALL_ARG_LOCATION.
1097 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1098 before dumping final insns.
1099 * except.c (emit_note_eh_region_end): Remove all references to
1100 NOTE_INSN_CALL_ARG_LOCATION.
1101 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1102 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1103 * config/arc/arc.c (hwloop_optimize): Likewise.
1104 * config/arm/arm.c (create_fix_barrier): Likewise.
1105 * config/s390/s390.c (s390_chunkify_start): Likewise.
1106 * config/sh/sh.c (find_barrier): Likewise.
1107 * config/i386/i386.c (rest_of_insert_endbranch,
1108 ix86_seh_fixup_eh_fallthru): Likewise.
1109 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1110 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1111 * config/frv/frv.c (frv_function_prologue): Likewise.
1112 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1114 (note_outside_basic_block_p): Remove all references to
1115 NOTE_INSN_CALL_ARG_LOCATION.
1116 * gengtype.c (adjust_field_rtx_def): Likewise.
1117 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1119 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1120 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1123 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1126 PR tree-optimization/84452
1127 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1128 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1131 2018-02-19 Martin Liska <mliska@suse.cz>
1134 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1136 2018-02-19 Martin Liska <mliska@suse.cz>
1137 Richard Sandiford <richard.sandiford@linaro.org>
1139 PR tree-optimization/82491
1140 * gimple-fold.c (get_base_constructor): Make earlier bail out
1143 2018-02-19 Carl Love <cel@us.ibm.com>
1145 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1146 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1148 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1149 P8V_BUILTIN_VEC_NEG.
1151 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1153 * config/rl78/rl78.md (movdf): New define expand.
1155 2018-02-19 Martin Liska <mliska@suse.cz>
1158 * doc/invoke.texi: Fix typo.
1159 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1161 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1163 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1164 handle rs6000_single_float and rs6000_double_float specially for
1167 2018-02-16 Jeff Law <law@redhat.com>
1169 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1172 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1175 * ipa-inline.c (inline_small_functions): Fix a typo.
1177 2018-02-16 Nathan Sidwell <nathan@acm.org>
1179 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1181 2018-02-16 Carl Love <cel@us.ibm.com>
1183 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1184 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1185 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1186 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1187 expansion to P8V_BUILTIN_VEC_FLOAT2.
1189 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1191 PR rtl-optimization/70023
1192 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1193 src_regno into account.
1195 2018-02-16 Carl Love <cel@us.ibm.com>
1197 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1198 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1199 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1200 * config/rs6000/rs6000.c: Remove case statements for
1201 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1202 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1203 and P9V_BUILTIN_VEC_VINSERT4B.
1204 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1205 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1206 * config/rs6000/vsx.md:
1207 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1210 2018-02-16 Carl Love <cel@us.ibm.com>
1212 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1214 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1216 * config/rs6000/rs6000-c.c: Add the definitions for
1217 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1218 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1219 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1220 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1221 definition for insert4b and define insn *insert3b_internal.
1222 * doc/extend.texi: Add documentation for vec_extract4b.
1224 2018-02-16 Nathan Sidwell <nathan@acm.org>
1226 * doc/extend.texi (Backwards Compatibility): Mention friend
1227 injection. Note for-scope is deprecated.
1228 * doc/invoke.texi (-ffriend-injection): Deprecate.
1230 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1232 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1233 that moved to I2, also allow destinations that are a paradoxical
1234 subreg (instead of a normal reg).
1236 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1239 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1242 2018-02-16 Richard Biener <rguenther@suse.de>
1244 PR tree-optimization/84037
1245 PR tree-optimization/84016
1247 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1248 Adjust vec_construct for the fact we need additional higher latency
1249 128bit inserts for AVX256 and AVX512 vector builds.
1250 (ix86_add_stmt_cost): Scale vector construction cost for
1253 2018-02-16 Richard Biener <rguenther@suse.de>
1255 PR tree-optimization/84417
1256 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1257 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1258 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1260 2018-02-16 Martin Liska <mliska@suse.cz>
1263 * internal-fn.def (ASAN_CHECK): Set proper flags.
1264 (ASAN_MARK): Likewise.
1266 2018-02-16 Julia Koval <julia.koval@intel.com>
1268 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1269 from PTA_CANNONLAKE.
1271 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1274 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1275 Use ++iter rather than iter++ for std::list iterators.
1276 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1277 defer deleting them until all nodes in the forest are processed. Do
1278 free even leaf nodes. Change to_process into auto_vec.
1281 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1282 * vec.h (vec_default_construct): Use memset instead of placement new
1283 if BROKEN_VALUE_INITIALIZATION is defined.
1284 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1285 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1288 PR rtl-optimization/83723
1289 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1290 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1291 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1293 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1295 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1297 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1299 PR rtl-optimization/81443
1300 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1301 from inner REGs to paradoxical SUBREGs.
1303 2018-02-16 Richard Biener <rguenther@suse.de>
1305 PR tree-optimization/84399
1306 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1307 For operands we can analyze at their definition make sure we can
1308 analyze them at each use as well.
1310 2018-02-16 Richard Biener <rguenther@suse.de>
1312 PR tree-optimization/84190
1313 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1314 volatile accesses if the decl isn't volatile.
1316 2018-02-15 Jason Merrill <jason@redhat.com>
1318 PR c++/84314 - ICE with templates and fastcall attribute.
1319 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1320 TYPE_CANONICAL on an existing type.
1322 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1324 PR tree-optimization/84383
1325 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1326 dstoff nor call operand_equal_p if dstbase is NULL.
1328 PR tree-optimization/84334
1329 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1330 also a CONSTANT_CLASS_P, punt.
1332 2018-02-14 Jim Wilson <jimw@sifive.com>
1334 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1335 first SMALL_OPERAND check. New local min_second_step. Move assert
1336 to where locals are set. Add TARGET_RVC support.
1337 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1339 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1341 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1343 2018-02-14 Martin Sebor <msebor@redhat.com>
1345 PR tree-optimization/83698
1346 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1347 arrays constrain the offset range to their bounds.
1348 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1349 (builtin_access::overlap): Avoid setting the size of overlap if it's
1351 (maybe_diag_overlap): Also consider arrays when deciding what values
1352 of offsets to include in diagnostics.
1354 2018-02-14 Martin Sebor <msebor@redhat.com>
1357 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1358 that correspond to the kind of a declaration.
1360 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1363 * config/pa/pa.md: Load address of PIC label using the linkage table
1364 if the label is nonlocal.
1366 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1368 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1369 warning message if user requests -maltivec=be.
1370 * doc/invoke.texi: Document deprecation of -maltivec=be.
1372 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1375 * config/rs6000/rs6000-c.c: Update definitions for
1376 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1377 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1379 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1382 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1383 add _get_ssp intrinsics. Remove argument from
1384 __builtin_ia32_rdssp[d|q].
1385 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1386 * config/i386/i386-builtin.def: Remove argument from
1387 __builtin_ia32_rdssp[d|q].
1388 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1389 ix86_expand_special_args_builtin for _rdssp[d|q].
1390 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1391 Clear register before usage.
1392 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1393 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1395 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1397 PR tree-optimization/84357
1398 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1399 operand 1 of an ARRAY_REF too.
1401 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1404 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1405 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1407 (set_of_reg): New struct.
1408 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1409 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1410 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1412 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1413 Split into bitclr, bitset, bitinvert patterns if appropriate.
1414 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1415 use rx_fuse_in_memory_bitop.
1416 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1417 to named insn, correct maximum insn length.
1419 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1422 * machmode.def: Define a complex mode for PARTIAL_INT.
1423 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1425 * doc/rtl.texi: Document CSPImode.
1426 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1428 (msp430_hard_regno_nregs_with_padding): Likewise.
1430 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1433 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1435 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1437 PR rtl-optimization/84169
1438 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1439 we generated a parallel as new i3 and we split that to new i2 and i3
1440 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1441 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1442 those to i2, not i1. Partially rewrite this scan code.
1444 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1447 * stor-layout.c (place_field): For variable length fields, adjust
1448 offset_align afterwards not just based on the field's alignment,
1449 but also on the size.
1452 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1453 of exps and logs in the use_exp2 case.
1455 2018-02-13 Jeff Law <law@redhat.com>
1457 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1460 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1463 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1467 * common.opt (gas-loc-support, gas-locview-support): New.
1468 (ginline-points, ginternal-reset-location-views): New.
1469 * doc/invoke.texi: Document them. Use @itemx where intended.
1470 (gvariable-location-views): Adjust.
1471 * target.def (reset_location_view): New.
1472 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1473 (TARGET_RESET_LOCATION_VIEW): New.
1474 * doc/tm.texi: Rebuilt.
1475 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1476 (dwarf2out_default_as_locview_support): New.
1477 (output_asm_line_debug_info): Use option variables.
1478 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1479 (output_loc_list): Likewise.
1480 (add_high_low_attributes): Check option variables.
1481 Don't output entry view attribute in strict mode.
1482 (gen_inlined_subroutine_die): Check option variables.
1483 (dwarf2out_inline_entry): Likewise.
1484 (init_sections_and_labels): Likewise.
1485 (dwarf2out_early_finish): Likewise.
1486 (maybe_reset_location_view): New, from...
1487 (dwarf2out_var_location): ... here. Call it.
1488 * debug.h (dwarf2out_default_as_loc_support): Declare.
1489 (dwarf2out_default_as_locview_support): Declare.
1490 * hooks.c (hook_int_rtx_insn_0): New.
1491 * hooks.h (hook_int_rtx_insn_0): Declare.
1492 * toplev.c (process_options): Take -gas-loc-support and
1493 -gas-locview-support from dwarf2out. Enable
1494 -gvariable-location-views by default only with locview
1495 assembler support. Enable -ginternal-reset-location-views by
1496 default only if the target defines the corresponding hook.
1497 Enable -ginline-points by default if location views are
1498 enabled; force it disabled if statement frontiers are
1500 * tree-inline.c (expand_call_inline): Check option variables.
1501 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1503 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1505 PR tree-optimization/84321
1506 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1507 handling. Also check whether the anti-range contains any values
1508 that satisfy the mask; switch to a VR_RANGE if not.
1510 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1513 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1515 2018-02-13 Martin Jambor <mjambor@suse.cz>
1518 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1519 of call statements, also set location of a load to a temporary.
1521 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1523 * config/rl78/rl78.c (add_vector_labels): New function.
1524 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1525 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1526 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1527 which checks that no arguments are passed.
1528 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1529 * doc/extend.texi: Documentation for the new attribute.
1531 2018-02-13 Andreas Schwab <schwab@suse.de>
1533 * config/riscv/linux.h (CPP_SPEC): Define.
1535 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1538 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1539 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1540 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1541 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1542 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1543 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1544 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1545 temporarily for AES and PCLMUL builtins.
1547 PR tree-optimization/84339
1548 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1549 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1553 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1554 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1555 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1557 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1559 * omp-simd-clone.h: New file.
1560 * omp-simd-clone.c: Include omp-simd-clone.h.
1561 (expand_simd_clones): No longer static.
1562 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1563 cgraph.h and omp-simd-clone.h.
1564 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1565 (vect_recog_widen_shift_pattern): Formatting fix.
1566 (vect_pattern_recog_1): Don't check optab for calls.
1569 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1570 operands[2] into a REG before using gen_lowpart on it.
1572 2018-02-12 Jeff Law <law@redhat.com>
1575 * config/sh/sh.c (find_barrier): Consider a sibling call
1578 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1579 successfully back substituting a reg.
1581 2018-02-12 Richard Biener <rguenther@suse.de>
1583 PR tree-optimization/84037
1584 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1585 parameter, move visited init to caller.
1586 (vect_slp_analyze_operations): Separate cost from validity
1587 check, initialize visited once for all instances.
1588 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1590 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1592 (vect_model_promotion_demotion_cost): Likewise.
1593 (vectorizable_bswap): Guard cost modeling with !slp_node
1594 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1596 (vectorizable_call): Likewise.
1597 (vectorizable_conversion): Likewise.
1598 (vectorizable_assignment): Likewise.
1599 (vectorizable_shift): Likewise.
1600 (vectorizable_operation): Likewise.
1601 (vectorizable_store): Likewise.
1602 (vectorizable_load): Likewise.
1603 (vectorizable_condition): Likewise.
1604 (vectorizable_comparison): Likewise.
1606 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1609 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1610 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1611 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1613 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1616 * params.def (inline-min-speedup): Increase from 8 to 15.
1617 (max-inline-insns-auto): Decrease from 40 to 30.
1618 * ipa-split.c (consider_split): Add some buffer for function to
1619 be considered inlining candidate.
1620 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1623 2018-02-12 Richard Biener <rguenther@suse.de>
1625 PR tree-optimization/84037
1626 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1627 matched stmts if we cannot swap the non-matched ones.
1629 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1631 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1632 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1633 _mm_maskz_scalef_round_ss): New intrinsics.
1634 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1635 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1636 __builtin_ia32_scalefss_round): Remove.
1637 (__builtin_ia32_scalefsd_mask_round,
1638 __builtin_ia32_scalefss_mask_round): New intrinsics.
1639 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1640 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1641 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1642 "<round_constraint>")): Changed to ...
1643 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1644 "<round_scalar_constraint>")): ... this.
1645 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1646 %0, %1, %2<round_op3>}"): Changed to ...
1647 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1648 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1649 %2<round_scalar_mask_op3>}"): ... this.
1650 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1652 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1654 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1655 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1656 (_mm_maskz_sqrt_round_ss): New intrinsics.
1657 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1658 (__builtin_ia32_sqrtsd_mask_round)
1659 (__builtin_ia32_sqrtss_mask_round): New builtins.
1660 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1661 (__builtin_ia32_sqrtss_round): Remove.
1662 (__builtin_ia32_sqrtsd_mask_round)
1663 (__builtin_ia32_sqrtss_mask_round): New builtins.
1664 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1665 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1666 ((match_operand:VF_128 1 "vector_operand"
1667 "xBm,<round_constraint>")): Changed to ...
1668 ((match_operand:VF_128 1 "vector_operand"
1669 "xBm,<round_scalar_constraint>")): ... this.
1670 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1671 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1672 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1673 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1674 %<iptr>1<round_scalar_mask_op3>}): ... this.
1675 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1676 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1678 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1681 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1682 Cast vec_cmpeq result to correct type.
1683 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1684 Cast vec_cmpgt result to correct type.
1686 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1688 * final.c (final_scan_insn_1): Renamed from...
1689 (final_scan_insn): ... this. New wrapper, to recover
1690 seen from the outermost call in recursive ones.
1691 * config/sparc/sparc.c (output_return): Drop seen from call.
1692 (output_sibcall): Likewise.
1693 * config/visium/visium.c (output_branch): Likewise.
1695 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1697 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1700 2018-02-10 Alan Modra <amodra@gmail.com>
1703 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1704 Specify LR as an input.
1706 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1709 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1710 remove_member_access_dummy_vars): New functions.
1711 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1712 lower_omp_1, execute_lower_omp): Use them.
1714 PR rtl-optimization/84308
1715 * shrink-wrap.c (spread_components): Release todo vector.
1717 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1719 PR rtl-optimization/57193
1720 * ira-color.c (struct allocno_color_data): Add member
1721 conflict_allocno_hard_prefs.
1722 (update_conflict_allocno_hard_prefs): New.
1723 (bucket_allocno_compare_func): Add a preference based on
1724 conflict_allocno_hard_prefs.
1725 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1726 (color_allocnos): Remove a dead code. Initiate
1727 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1729 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1732 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1733 constraint from =wa to wa. Avoid a subreg on the output operand,
1734 instead use a pseudo and subreg it in a move.
1735 (p9_xxbrd_<mode>): Changed to ...
1736 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1737 (p9_xxbrd_v2df): New expander.
1738 (p9_xxbrw_<mode>): Changed to ...
1739 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1740 (p9_xxbrw_v4sf): New expander.
1742 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1744 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1746 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1749 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1750 multiply in 32-bit mode.
1751 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1752 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1755 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1757 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1758 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1759 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1760 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1762 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1765 * dwarf2out.c (is_trivial_indirect_ref): New function.
1766 (dwarf2out_late_global_decl): Do not generate a location
1767 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1768 and that are not defined in the current unit.
1770 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1772 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1773 instead of a libcall for UNORDERED.
1775 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1778 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1779 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1781 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1784 * config/s390/s390.c (s390_set_current_function): Invoke
1785 s390_indirect_branch_settings also if fndecl didn't change.
1787 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1789 * config/rs6000/rs6000.md (blockage): Set length to zero.
1791 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1793 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1795 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1798 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1799 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1803 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1804 PARALLEL incoming that failed vt_get_decl_and_offset check.
1807 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1808 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1810 (get_variable_section): For decls in named .bss* sections pass true as
1811 second argument to bss_initializer_p.
1813 2018-02-09 Marek Polacek <polacek@redhat.com>
1814 Jakub Jelinek <jakub@redhat.com>
1817 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1818 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1819 Sync some changes from cxx_fold_indirect_ref.
1821 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1823 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1825 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1826 (BLOCK_INLINE_ENTRY_LABEL): New.
1827 (dwarf2out_var_location): Disregard inline entry markers.
1828 (inline_entry_data): New struct.
1829 (inline_entry_data_hasher): New hashtable type.
1830 (inline_entry_data_hasher::hash): New.
1831 (inline_entry_data_hasher::equal): New.
1832 (inline_entry_data_table): New variable.
1833 (add_high_low_attributes): Add DW_AT_entry_pc and
1834 DW_AT_GNU_entry_view attributes if a pending entry is found
1835 in inline_entry_data_table. Add old entry_pc attribute only
1836 if debug nonbinding markers are disabled.
1837 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1838 markers are enabled.
1839 (block_within_block_p, dwarf2out_inline_entry): New.
1840 (dwarf2out_finish): Check that no entries remained in
1841 inline_entry_data_table.
1842 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1843 (final_scan_insn, notice_source_line): Likewise.
1844 (rest_of_clean_state): Skip inline entry markers.
1845 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1847 * gimple.c (gimple_build_debug_inline_entry): New.
1848 * gimple.h (enum gimple_debug_subcode): Add
1849 GIMPLE_DEBUG_INLINE_ENTRY.
1850 (gimple_build_debug_inline_entry): Declare.
1851 (gimple_debug_inline_entry_p): New.
1852 (gimple_debug_nonbind_marker_p): Adjust.
1853 * insn-notes.def (INLINE_ENTRY): New.
1854 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1855 inline entry marker notes.
1856 (print_insn): Likewise.
1857 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1858 (INSN_DEBUG_MARKER_KIND): Likewise.
1859 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1860 * tree-inline.c (expand_call_inline): Build and insert
1861 debug_inline_entry stmt.
1862 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1863 inline entry blocks early, if nonbind markers are enabled.
1864 (dump_scope_block): Dump fragment info.
1865 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1866 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1867 (gimple_build_debug_inline_entry): New.
1868 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1869 Enable/disable inline entry points too.
1870 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1871 (DEBUG_INSN): Describe inline entry markers.
1873 * common.opt (gvariable-location-views): New.
1874 (gvariable-location-views=incompat5): New.
1875 * config.in: Rebuilt.
1876 * configure: Rebuilt.
1877 * configure.ac: Test assembler for view support.
1878 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1879 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1880 * dwarf2out.c (var_loc_view): New typedef.
1881 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1882 (dwarf2out_locviews_in_attribute): New.
1883 (dwarf2out_locviews_in_loclist): New.
1884 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1885 (enum dw_line_info_opcode): Add LI_adv_address.
1886 (struct dw_line_info_table): Add view.
1887 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1888 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1889 (zero_view_p): New variable.
1890 (ZERO_VIEW_P): New macro.
1891 (output_asm_line_debug_info): New.
1892 (struct var_loc_node): Add view.
1893 (add_AT_view_list, AT_loc_list): New.
1894 (add_var_loc_to_decl): Add view param. Test it against last.
1895 (new_loc_list): Add view params. Record them.
1896 (AT_loc_list_ptr): Handle loc and view lists.
1897 (view_list_to_loc_list_val_node): New.
1898 (print_dw_val): Handle dw_val_class_view_list.
1899 (size_of_die): Likewise.
1900 (value_format): Likewise.
1901 (loc_list_has_views): New.
1902 (gen_llsym): Set vl_symbol too.
1903 (maybe_gen_llsym, skip_loc_list_entry): New.
1904 (dwarf2out_maybe_output_loclist_view_pair): New.
1905 (output_loc_list): Output view list or entries too.
1906 (output_view_list_offset): New.
1907 (output_die): Handle dw_val_class_view_list.
1908 (output_dwarf_version): New.
1909 (output_compilation_unit_header): Use it.
1910 (output_skeleton_debug_sections): Likewise.
1911 (output_rnglists, output_line_info): Likewise.
1912 (output_pubnames, output_aranges): Update version comments.
1913 (output_one_line_info_table): Output view numbers in asm comments.
1914 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1915 Call maybe_gen_llsym.
1916 (loc_list_from_tree_1): Adjust.
1917 (add_AT_location_description): Create view list attribute if
1918 needed, check it's absent otherwise.
1919 (convert_cfa_to_fb_loc_list): Adjust.
1920 (maybe_emit_file): Call output_asm_line_debug_info for test.
1921 (dwarf2out_var_location): Reset views as needed. Precompute
1922 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1923 attribute. Set view.
1924 (new_line_info_table): Reset next view.
1925 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1926 (dwarf2out_source_line): Likewise. Output view resets and labels to
1927 the assembler, or select appropriate line info opcodes.
1928 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1929 (optimize_string_length): Catch it. Adjust.
1930 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1931 dw_val_class_view_list, and remove it if no longer needed.
1932 (hash_loc_list): Hash view numbers.
1933 (loc_list_hasher::equal): Compare them.
1934 (optimize_location_lists): Check whether a view list symbol is
1935 needed, and whether the locview attribute is present, and
1936 whether they match. Remove the locview attribute if no longer
1938 (index_location_lists): Call skip_loc_list_entry for test.
1939 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1940 Use output_dwarf_version.
1941 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1942 (struct dw_val_node): Add val_view_list.
1943 * final.c (SEEN_NEXT_VIEW): New.
1944 (set_next_view_needed): New.
1945 (clear_next_view_needed): New.
1946 (maybe_output_next_view): New.
1947 (final_start_function): Rename to...
1948 (final_start_function_1): ... this. Take pointer to FIRST,
1949 add SEEN parameter. Emit param bindings in the initial view.
1950 (final_start_function): Reintroduce SEEN-less interface.
1951 (final): Rename to...
1952 (final_1): ... this. Take SEEN parameter. Output final pending
1953 next view at the end.
1954 (final): Reintroduce seen-less interface.
1955 (final_scan_insn): Output pending next view before switching
1956 sections or ending a block. Mark the next view as needed when
1957 outputting variable locations. Notify debug backend of section
1958 changes, and of location view changes.
1959 (rest_of_handle_final): Adjust.
1960 * toplev.c (process_options): Autodetect value for debug variable
1961 location views option. Warn on incompat5 without -gdwarf-5.
1962 * doc/invoke.texi (gvariable-location-views): New.
1963 (gvariable-location-views=incompat5): New.
1964 (gno-variable-location-views): New.
1966 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1968 PR tree-optimization/84136
1969 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1970 that the result of find_edge is non-NULL.
1972 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1975 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1976 storing integer register in SImode. Fix cost of 256 and 512
1977 byte aligned SSE register store.
1979 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1981 * config/i386/i386.c (ix86_multiplication_cost): Fix
1982 multiplication cost for TARGET_AVX512DQ.
1984 2018-02-08 Marek Polacek <polacek@redhat.com>
1986 PR tree-optimization/84238
1987 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1990 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1992 PR tree-optimization/84265
1993 * tree-vect-stmts.c (vectorizable_store): Don't treat
1994 VMAT_CONTIGUOUS accesses as grouped.
1995 (vectorizable_load): Likewise.
1997 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1999 PR tree-optimization/81635
2000 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
2001 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
2002 (test_round_for_mask): New functions.
2003 (wide_int_cc_tests): Call test_round_for_mask.
2004 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
2005 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2006 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2007 range returned by get_range_info.
2009 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2012 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2013 * symtab.c: Include builtins.h
2014 (symtab_node::output_to_lto_symbol_table_p): Move here
2015 from lto-streamer-out.c:output_symbol_p.
2016 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2017 (output_symbol_p): Move all logic to symtab.c
2018 (produce_symtab): Update.
2020 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2022 * config/s390/s390-opts.h (enum indirect_branch): Define.
2023 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2024 (s390_indirect_branch_via_thunk)
2025 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2026 (enum s390_indirect_branch_type): Define.
2027 * config/s390/s390.c (struct s390_frame_layout, struct
2028 machine_function): Remove.
2029 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2030 (indirect_branch_table_label_no, indirect_branch_table_name):
2032 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2033 (enum s390_indirect_branch_option): Define.
2034 (s390_return_addr_from_memory): New function.
2035 (s390_handle_string_attribute): New function.
2036 (s390_attribute_table): Add new attribute handler.
2037 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2038 (s390_indirect_branch_via_thunk): New function.
2039 (s390_indirect_branch_via_inline_thunk): New function.
2040 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2041 sibling call optimization for non z10 compiles.
2042 (s390_emit_call): Force indirect branch target to be a single
2043 register. Add r1 clobber for non-z10 compiles.
2044 (s390_emit_epilogue): Emit return jump via return_use expander.
2045 (s390_reorg): Handle JUMP_INSNs as execute targets.
2046 (s390_option_override_internal): Perform validity checks for the
2047 new command line options.
2048 (s390_indirect_branch_attrvalue): New function.
2049 (s390_indirect_branch_settings): New function.
2050 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2051 (s390_output_indirect_thunk_function): New function.
2052 (s390_code_end): Implement target hook.
2053 (s390_case_values_threshold): Implement target hook.
2054 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2056 * config/s390/s390.h (struct s390_frame_layout)
2057 (struct machine_function): Move here from s390.c.
2058 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2059 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2060 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2061 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2062 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2063 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2064 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2065 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2066 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2067 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2068 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2069 (mnemonic attribute): Add values which aren't recognized
2071 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2072 pattern for branch conversion. Fix mnemonic attribute.
2073 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2074 indirect branch via thunk if requested.
2075 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2076 ("*indirect_jump"): Disable for branch conversion using out of
2078 ("indirect_jump_via_thunk<mode>_z10")
2079 ("indirect_jump_via_thunk<mode>")
2080 ("indirect_jump_via_inlinethunk<mode>_z10")
2081 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2082 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2083 ("casesi_jump_via_inlinethunk<mode>_z10")
2084 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2085 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2086 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2087 ("*indirect2_jump"): Disable for branch conversion.
2088 ("casesi_jump"): Turn into expander and expand patterns for branch
2090 ("return_use"): New expander.
2091 ("*return"): Emit return via thunk and rename it to ...
2092 ("*return<mode>"): ... this one.
2093 * config/s390/s390.opt: Add new options and and enum for the
2096 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2098 * lra-constraints.c (match_reload): Unconditionally use
2099 gen_lowpart_SUBREG, rather than selecting between that
2100 and equivalent gen_rtx_SUBREG code.
2102 2018-02-08 Richard Biener <rguenther@suse.de>
2104 PR tree-optimization/84233
2105 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2106 changed flag instead of boguously re-using phi_inserted.
2108 2018-02-08 Martin Jambor <mjambor@suse.cz>
2110 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2111 static local variables.
2113 2018-02-08 Richard Biener <rguenther@suse.de>
2115 PR tree-optimization/84278
2116 * tree-vect-stmts.c (vectorizable_store): When looking for
2117 smaller vector types to perform grouped strided loads/stores
2118 make sure the mode is supported by the target.
2119 (vectorizable_load): Likewise.
2121 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2123 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2124 Increase LDP/STP opportunities by adding adjacent callee-saves.
2126 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2128 PR rtl-optimization/84068
2129 PR rtl-optimization/83459
2130 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2132 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2134 PR tree-optimization/84224
2135 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2136 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2139 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2142 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2143 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2144 count, remove one USE.
2146 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2148 * doc/install.texi (Configuration): Document the
2149 --with-long-double-format={ibm,ieee} PowerPC configuration
2153 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2154 Convert from define_expand to be define_insn_and_split. Rework
2155 float/double/_Float128 conversions to QI/HI/SImode to work with
2156 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2157 conversions to QI/HImode types did a store and then a load to
2158 truncate the value. For conversions to VSX registers, don't split
2159 the insn, instead emit the code directly. Use the code iterator
2160 any_fix to combine signed and unsigned conversions.
2161 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2162 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2163 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2164 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2165 (fix_<mode>di2_hw): Likewise.
2166 (fixuns_<mode>di2_hw): Likewise.
2167 (fix_<mode>si2_hw): Likewise.
2168 (fixuns_<mode>si2_hw): Likewise.
2169 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2170 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2171 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2172 fix<uns>_trunc<SFDF:mode>si2_p8.
2173 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2175 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2176 (fix<uns>_<mode>_mem): Likewise.
2177 (fctiw<u>z_<mode>_mem): Likewise.
2178 (fix<uns>_<mode>_mem): Likewise.
2179 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2180 the register allocator from doing a direct move to the GPRs to do
2181 a store, and instead use the ISA 3.0 store byte/half-word from
2182 vector register instruction. For IEEE 128-bit floating point,
2183 also optimize stores of 32-bit ints.
2184 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2186 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2188 * genextract.c (push_pathstr_operand): New function to support
2190 (walk_rtx): Call push_pathstr_operand.
2191 (print_path): Support [a-zA-Z].
2193 2018-02-07 Richard Biener <rguenther@suse.de>
2195 PR tree-optimization/84037
2196 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2197 (cse_and_gimplify_to_preheader): Declare.
2198 (vect_get_place_in_interleaving_chain): Likewise.
2199 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2201 (_loop_vec_info::~_loop_vec_info): Delete it.
2202 (cse_and_gimplify_to_preheader): New function.
2203 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2204 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2205 (vectorizable_load): Likewise. For grouped stores always base
2206 the IV on the first element.
2207 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2208 condition before gimplifying.
2210 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2212 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2213 *DIV_EXPR and *MOD_EXPR.
2215 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2218 * config/i386/i386.c (ix86_option_override_internal): Mask out
2219 the CF_SET bit when checking -fcf-protection.
2221 2018-02-07 Tom de Vries <tom@codesourcery.com>
2224 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2227 2018-02-07 Richard Biener <rguenther@suse.de>
2229 PR tree-optimization/84204
2230 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2233 PR tree-optimization/84205
2234 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2235 special-case isl_ast_op_zdiv_r.
2237 PR tree-optimization/84223
2238 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2239 Only add conditions from within the region.
2240 (gather_bbs::after_dom_children): Adjust.
2242 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2245 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2246 * config/avr/avr.md: Only post-reload split REG-REG moves if
2247 either register is GENERAL_REG_P.
2249 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2251 PR tree-optimization/84235
2252 * tree-ssa-scopedtables.c
2253 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2254 if the subtraction is performed in floating point type where NaNs are
2255 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2256 build 1. Formatting fix.
2258 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2261 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2262 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2263 and skip it regardless of bb boundaries. Use CALL_P macro,
2264 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2265 unnecessarily, formatting fix.
2267 2018-02-06 Michael Collison <michael.collison@arm.com>
2269 * config/arm/thumb2.md:
2270 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2271 (*thumb_mov_notscc): Ditto.
2273 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2276 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2277 unsigned_fix, not "s".
2279 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2281 * configure.ac (gcc_fn_eh_frame_ro): New function.
2282 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2283 correct .eh_frame permissions.
2284 * configure: Regenerate.
2286 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2288 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2291 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2293 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2294 Display warning message for -mno-speculate-indirect-jumps.
2296 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2298 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2300 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2302 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2304 PR tree-optimization/84225
2305 * tree-eh.c (find_trapping_overflow): Only call
2306 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2308 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2311 * config/i386/i386.c: Reimplement the check of possible options
2312 -mibt/-mshstk conbination. Change error messages.
2313 * doc/invoke.texi: Fix a typo: remove extra '='.
2315 2018-02-06 Marek Polacek <polacek@redhat.com>
2317 PR tree-optimization/84228
2318 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2320 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2323 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2324 emitted arch directives.
2325 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2326 __ARM_FEATURE_COPROC before changing architectures.
2328 2018-02-06 Richard Biener <rguenther@suse.de>
2330 * config/i386/i386.c (print_reg): Fix typo.
2331 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2333 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2335 * configure: Regenerate.
2337 2018-02-05 Martin Sebor <msebor@redhat.com>
2339 PR tree-optimization/83369
2340 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2343 2018-02-05 Martin Liska <mliska@suse.cz>
2345 * doc/invoke.texi: Cherry-pick upstream r323995.
2347 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2349 * ira.c (ira_init_register_move_cost): Adjust comment.
2351 2018-02-05 Martin Liska <mliska@suse.cz>
2353 PR gcov-profile/84137
2354 * doc/gcov.texi: Fix typo in documentation.
2356 2018-02-05 Martin Liska <mliska@suse.cz>
2358 PR gcov-profile/83879
2359 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2360 using dlopen functionality.
2362 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2364 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2365 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2366 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2367 _mm_maskz_range_round_ss): New intrinsics.
2368 (__builtin_ia32_rangesd128_round)
2369 (__builtin_ia32_rangess128_round): Remove.
2370 (__builtin_ia32_rangesd128_mask_round,
2371 __builtin_ia32_rangess128_mask_round): New builtins.
2372 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2373 __builtin_ia32_rangess128_round): Remove.
2374 (__builtin_ia32_rangesd128_mask_round,
2375 __builtin_ia32_rangess128_mask_round): New builtins.
2376 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2377 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2378 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2379 "<round_saeonly_constraint>")): Changed to ...
2380 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2381 "<round_saeonly_scalar_constraint>")): ... this.
2382 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2383 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2384 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2385 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2386 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2388 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2390 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2392 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2393 Remove all values except native, 8540 and 8548.
2395 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2397 * config/i386/i386.c (ix86_output_function_return): Pass
2398 INVALID_REGNUM, instead of -1, as invalid register number to
2399 indirect_thunk_name and output_indirect_thunk.
2401 2018-02-02 Julia Koval <julia.koval@intel.com>
2403 * config.gcc: Add -march=icelake.
2404 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2405 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2406 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2407 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2408 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2409 (processor_target_table): Add icelake.
2410 (ix86_option_override_internal): Handle new PTAs.
2411 (get_builtin_code_for_version): Handle icelake.
2412 (M_INTEL_COREI7_ICELAKE): New.
2413 (fold_builtin_cpu): Handle icelake.
2414 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2415 * doc/invoke.texi: Add -march=icelake.
2417 2018-02-02 Julia Koval <julia.koval@intel.com>
2419 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2420 to wide_int_bitmask.
2421 * wide-int-bitmask.h: New.
2423 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2426 * config/i386/i386.md: Replace Pmode with word_mode in
2427 builtin_setjmp_setup and builtin_longjmp to support x32.
2429 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2433 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2435 (rs6000_supported_cpu_names): New static variable.
2436 (linux_cpu_translation_table): Likewise.
2437 (elf_platform) <cpu>: Define new static variable and use it.
2438 Translate kernel AT_PLATFORM name to canonical name if needed.
2439 Error if platform name is unknown.
2441 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2444 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2446 2018-02-01 Jeff Law <law@redhat.com>
2449 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2450 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2451 the scratch if RELEASE_VIA_POP is false.
2452 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2453 If we have to save a temporary register, decrement SIZE appropriately.
2454 Pass new arguments to release_scratch_register_on_entry.
2455 (ix86_adjust_stack_and_probe): Likewise.
2456 (ix86_emit_probe_stack_range): Pass new arguments to
2457 release_scratch_register_on_entry.
2459 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2461 PR rtl-optimization/84157
2462 * combine.c (change_zero_ext): Use REG_P predicate in
2463 front of HARD_REGISTER_P predicate.
2465 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2467 * config/avr/avr.c (avr_option_override): Move disabling of
2468 -fdelete-null-pointer-checks to...
2469 * common/config/avr/avr-common.c (avr_option_optimization_table):
2472 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2474 PR tree-optimization/81635
2475 * tree-data-ref.c (split_constant_offset_1): For types that
2476 wrap on overflow, try to use range info to prove that wrapping
2479 2018-02-01 Renlin Li <renlin.li@arm.com>
2482 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2484 (aarch64_register_move_cost): Likewise.
2485 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2487 (REG_CLASS_NAMES): Likewise.
2488 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2489 TAILCALL_ADDR_REGS. Remove IP registers.
2490 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2492 2018-02-01 Richard Biener <rguenther@suse.de>
2494 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2495 for specifying RPO order and allow NULL for that.
2496 * domwalk.c (dom_walker::dom_walker): Likewise.
2497 (dom_walker::walk): Handle NULL RPO order.
2498 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2500 (rewrite_update_dom_walker): Likewise.
2501 (mark_def_dom_walker): Likewise.
2503 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2505 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2506 (aarch64_maybe_expand_sve_subreg_move): Declare.
2507 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2508 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2510 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2511 that are semantically a reverse operation.
2512 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2513 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2514 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2516 (aarch64_can_change_mode_class): For big-endian, forbid changes
2517 between two SVE modes if they have different element sizes.
2519 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2521 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2522 the TImode handling for big-endian targets.
2524 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2526 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2527 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2529 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2530 Remove BSWAP handing for big-endian targets and use the form of
2531 LD1RQ appropariate for the mode.
2533 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2535 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2536 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2539 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2542 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2543 check for operands that need to go through aarch64_sve_reload_be.
2545 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2547 PR tree-optimization/81661
2548 PR tree-optimization/84117
2549 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2550 * tree-eh.c: Include gimplify.h.
2551 (find_trapping_overflow, replace_trapping_overflow,
2552 rewrite_to_non_trapping_overflow): New functions.
2553 * tree-vect-loop.c: Include tree-eh.h.
2554 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2555 * tree-data-ref.c: Include tree-eh.h.
2556 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2558 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2560 PR rtl-optimization/84123
2561 * combine.c (change_zero_ext): Check if hard register satisfies
2562 can_change_dest_mode before calling gen_lowpart_SUBREG.
2564 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2567 * ira.c (ira_init_register_move_cost): Remove assert.
2569 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2571 PR rtl-optimization/84071
2572 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2573 * doc/tm.texi: Regenerate.
2575 2018-01-31 Richard Biener <rguenther@suse.de>
2577 PR tree-optimization/84132
2578 * tree-data-ref.c (analyze_miv_subscript): Properly
2579 check whether evolution_function_is_affine_multivariate_p
2580 before calling gcd_of_steps_may_divide_p.
2582 2018-01-31 Julia Koval <julia.koval@intel.com>
2585 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2586 * config/i386/i386.md (rdpid_rex64) New.
2587 (rdpid): Make 32bit only.
2589 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2592 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2593 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2595 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2598 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2600 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2602 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2604 PR rtl-optimization/84071
2605 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2606 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2608 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2610 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2611 (arc_attribute_table): Add 'aux' attribute.
2612 (arc_in_small_data_p): Consider aux like variables.
2613 (arc_is_aux_reg_p): New function.
2614 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2615 (arc_get_aux_arg): New function.
2616 (prepare_move_operands): Handle aux-register access.
2617 (arc_handle_aux_attribute): New function.
2618 * doc/extend.texi (ARC Variable attributes): Add subsection.
2620 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2622 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2623 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2624 (arc_attribute_table): Add 'uncached' attribute.
2625 (arc_print_operand): Print '.di' flag for uncached memory
2627 (arc_in_small_data_p): Do not consider for small data the uncached
2629 (arc_is_uncached_mem_p): New function.
2630 * config/arc/predicates.md (compact_store_memory_operand): Check
2631 for uncached memory accesses.
2632 (nonvol_nonimm_operand): Likewise.
2633 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2635 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2638 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2639 falign-loops=): Add Optimization flag.
2641 2018-01-30 Jeff Law <law@redhat.com>
2644 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2645 INT_REGISTERS_SAVED. Check it prior to calling
2646 get_scratch_register_on_entry.
2647 (ix86_adjust_stack_and_probe): Similarly.
2648 (ix86_emit_probe_stack_range): Similarly.
2649 (ix86_expand_prologue): Corresponding changes.
2651 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2654 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2655 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2657 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2660 * lra-constraints.c (curr_insn_transform): Process AND in the
2663 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2665 PR rtl-optimization/83986
2666 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2667 dependence against last_pending_memory_flush in addition to
2670 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2672 PR tree-optimization/81611
2673 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2676 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2679 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2682 2018-01-30 Richard Biener <rguenther@suse.de>
2683 Jakub Jelinek <jakub@redhat.com>
2685 PR tree-optimization/84111
2686 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2687 inner loops added during recursion, as they don't have up-to-date
2690 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2693 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2694 (can_inline_edge_by_limits_p): ... here.
2695 (can_early_inline_edge_p, check_callers,
2696 update_caller_keys, update_callee_keys, recursive_inlining,
2697 add_new_edges_to_heap, speculation_useful_p,
2698 inline_small_functions,
2699 inline_small_functions, flatten_function,
2700 inline_to_all_callers_1): Update.
2702 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2704 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2707 2018-01-30 Richard Biener <rguenther@suse.de>
2709 PR tree-optimization/83008
2710 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2711 invariant and constant vector uses in stmts when they need
2714 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2717 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2718 * configure: Regenerate.
2720 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2722 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2724 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2725 Use gen_rtx_REG rather than gen_lowpart.
2727 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2729 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2730 rather than 0 when creating partial subregs.
2732 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2734 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2737 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2740 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2741 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2742 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2743 flags. This restores the settings used before the 2017-07-24.
2744 Turning off pre increment/decrement/modify allows IVOPTS to
2745 optimize DF/SF loops where the index is an int.
2747 2018-01-29 Richard Biener <rguenther@suse.de>
2748 Kelvin Nilsen <kelvin@gcc.gnu.org>
2751 * tree-vect-stmts.c (vectorizable_call): Don't call
2752 targetm.vectorize_builtin_md_vectorized_function if callee is
2755 2018-01-22 Carl Love <cel@us.ibm.com>
2757 * doc/extend.tex: Fix typo in second arg in
2758 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2760 2018-01-29 Richard Biener <rguenther@suse.de>
2762 PR tree-optimization/84086
2763 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2764 (flush_ssaname_freelist): When SSA names were released reset
2765 the SCEV hash table.
2767 2018-01-29 Richard Biener <rguenther@suse.de>
2769 PR tree-optimization/84057
2770 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2771 removed paths when removing edges.
2773 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2775 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2776 -mfunction-return=@var{choice}.
2778 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2781 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2783 (layout::print_source_line): Likewise.
2784 (test_get_line_width_without_trailing_whitespace): Add test cases.
2786 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2789 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2792 2018-01-26 Jim Wilson <jimw@sifive.com>
2794 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2796 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2799 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2801 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2802 and CMP + SUB-immediate -> SUBS.
2804 2018-01-26 Martin Sebor <msebor@redhat.com>
2806 PR tree-optimization/83896
2807 * tree-ssa-strlen.c (get_string_len): Rename...
2808 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2809 Avoid assuming length is constant.
2810 (handle_char_store): Use HOST_WIDE_INT for string length.
2812 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2815 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2816 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2818 2018-01-26 Richard Biener <rguenther@suse.de>
2820 PR rtl-optimization/84003
2821 * dse.c (record_store): Only record redundant stores when
2822 the earlier store aliases at least all accesses the later one does.
2824 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2826 PR rtl-optimization/83985
2827 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2828 REG_CFA_RESTORE insns.
2829 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2830 REG_CFA_RESTORE insns here.
2833 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2834 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2836 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2838 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2839 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2840 (arc_init): Likewise.
2841 (arc_override_options): Likewise.
2842 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2844 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2846 * config/arc/arc.h (TARGET_DBNZ): Define.
2847 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2848 properly set the tune attribute.
2849 (dbnz): Use TARGET_DBNZ guard.
2850 * config/arc/arc.opt (mtune): Add core3 option.
2852 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2854 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2855 recognize new pic like addresses.
2856 (arc_delegitimize_address): Clean up.
2858 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2860 * config/arc/arc-arches.def: Option mrf16 valid for all
2862 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2863 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2864 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2865 * config/arc/arc-tables.opt: Regenerate.
2866 * config/arc/arc.c (arc_conditional_register_usage): Handle
2867 reduced register file case.
2868 (arc_file_start): Set must have build attributes.
2869 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2871 * config/arc/arc.opt (mrf16): Add new option.
2872 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2873 * config/arc/genmultilib.awk: Handle new mrf16 option.
2874 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2875 * config/arc/t-multilib: Regenerate.
2876 * doc/invoke.texi (ARC Options): Document mrf16 option.
2878 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2880 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2881 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2882 (arc_attribute_table): Add 'secure_call' attribute.
2883 (arc_print_operand): Print secure call operand.
2884 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2886 (arc_is_secure_call_p): New function. * config/arc/arc.md
2887 (call_i): Add support for sjli instruction.
2888 (call_value_i): Likewise.
2889 * config/arc/constraints.md (Csc): New constraint.
2891 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2892 John Eric Martin <John.Martin@emmicro-us.com>
2894 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2895 * config/arc/arc.c (_arc_jli_section): New struct.
2896 (arc_jli_section): New type.
2897 (rc_jli_sections): New static variable.
2898 (arc_handle_jli_attribute): New function.
2899 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2900 (arc_file_end): New function.
2901 (TARGET_ASM_FILE_END): Define.
2902 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2903 (arc_add_jli_section): New function.
2904 (jli_call_scan): Likewise.
2905 (arc_reorg): Call jli_call_scan.
2906 (arc_output_addsi): Remove 'S' from printing asm operand.
2907 (arc_is_jli_call_p): New function.
2908 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2910 (movhi_insn): Likewise.
2911 (movsi_insn): Likewise.
2912 (movsi_set_cc_insn): Likewise.
2913 (loadqi_update): Likewise.
2914 (load_zeroextendqisi_update): Likewise.
2915 (load_signextendqisi_update): Likewise.
2916 (loadhi_update): Likewise.
2917 (load_zeroextendhisi_update): Likewise.
2918 (load_signextendhisi_update): Likewise.
2919 (loadsi_update): Likewise.
2920 (loadsf_update): Likewise.
2921 (movsicc_insn): Likewise.
2922 (bset_insn): Likewise.
2923 (bxor_insn): Likewise.
2924 (bclr_insn): Likewise.
2925 (bmsk_insn): Likewise.
2926 (bicsi3_insn): Likewise.
2927 (cmpsi_cc_c_insn): Likewise.
2928 (movsi_ne): Likewise.
2929 (movsi_cond_exec): Likewise.
2930 (clrsbsi2): Likewise.
2941 (call_i): Remove 'S' asm letter, add jli instruction.
2942 (call_value_i): Likewise.
2943 * config/arc/arc.op (mjli-always): New option.
2944 * config/arc/constraints.md (Cji): New constraint.
2945 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2947 (subsf3_fpx): Likewise.
2948 (mulsf3_fpx): Likewise.
2949 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2951 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2953 * doc/invoke.texi (ARC): Document mjli-always option.
2955 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2957 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2958 avoid addition with 0 and use incw and decw where possible.
2960 2018-01-26 Richard Biener <rguenther@suse.de>
2962 PR tree-optimization/81082
2963 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2964 association if it requires casting to unsigned.
2965 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2966 from fold_plusminus_mult_expr to catch important cases late when
2967 range info is available.
2969 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2971 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2972 * configure.ac (hidden_linkonce): New test.
2973 * configure: Regenerate.
2974 * config.in: Regenerate.
2976 2018-01-26 Julia Koval <julia.koval@intel.com>
2978 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2979 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2980 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2981 _mm_mask_bitshuffle_epi64_mask): Fix type.
2982 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2983 USI_FTYPE_V4DI_V4DI_USI): Remove.
2984 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2985 __builtin_ia32_vpshufbitqmb256_mask,
2986 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2987 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2988 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2990 2018-01-26 Alan Modra <amodra@gmail.com>
2993 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2994 UNSPEC_VBPERMQ. Sort other unspecs.
2996 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2998 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
3000 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
3003 * predict.c (drop_profile): Do not push/pop cfun; update also
3005 (handle_missing_profiles): Fix logic looking for zero profiles.
3007 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3010 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3011 on functions with #pragma omp declare simd or functions with simd
3013 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3014 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3015 Remove trailing \n from warning_at calls.
3017 2018-01-25 Tom de Vries <tom@codesourcery.com>
3020 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3021 for neutered workers.
3023 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3026 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3027 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3029 2018-01-24 Jeff Law <law@redhat.com>
3032 * i386.c (get_probe_interval): Move to earlier point.
3033 (ix86_compute_frame_layout): If -fstack-clash-protection and
3034 the frame is larger than the probe interval, then use pushes
3035 to save registers rather than reg->mem moves.
3036 (ix86_expand_prologue): Remove conditional for int_registers_saved
3039 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3042 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3043 min/max for never referenced object.
3045 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3048 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3050 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3051 attributes from DECL_ATTRIBUTES (decl) without affecting
3052 DECL_ATTRIBUTES (current_function_decl).
3053 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3054 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3056 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3058 PR tree-optimization/83979
3059 * fold-const.c (fold_comparison): Use constant_boolean_node
3060 instead of boolean_{true,false}_node.
3062 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3064 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3067 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3069 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3070 Simplify the clause that sets the length attribute.
3071 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3072 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3073 clause that sets the length attribute.
3074 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3076 2018-01-24 Tom de Vries <tom@codesourcery.com>
3079 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3080 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3081 Add strict parameter.
3082 (prevent_branch_around_nothing): Insert dummy insn between branch to
3083 label and label with no ptx insn inbetween.
3084 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3086 2018-01-24 Tom de Vries <tom@codesourcery.com>
3089 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3090 for neutered threads in warp.
3091 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3093 2018-01-24 Richard Biener <rguenther@suse.de>
3095 PR tree-optimization/83176
3096 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3099 2018-01-24 Richard Biener <rguenther@suse.de>
3101 PR tree-optimization/82819
3102 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3103 code generating pluses that are no-ops in the target precision.
3105 2018-01-24 Richard Biener <rguenther@suse.de>
3108 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3110 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3112 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3113 to merge probabilities.
3114 * predict.c (probably_never_executed): Also mark as cold functions
3115 with global 0 profile and guessed local profile.
3116 * profile-count.c (profile_probability::combine_with_count): New
3118 * profile-count.h (profile_probability::operator*,
3119 profile_probability::operator*=, profile_probability::operator/,
3120 profile_probability::operator/=): Reduce precision to adjusted
3121 and set value to guessed on contradictory divisions.
3122 (profile_probability::combine_with_freq): Remove.
3123 (profile_probability::combine_wiht_count): Declare.
3124 (profile_count::force_nonzero):: Set to adjusted.
3125 (profile_count::probability_in):: Set quality to adjusted.
3126 * tree-ssa-tail-merge.c (replace_block_by): Use
3129 2018-01-23 Andrew Waterman <andrew@sifive.com>
3130 Jim Wilson <jimw@sifive.com>
3132 * config/riscv/riscv.c (riscv_stack_boundary): New.
3133 (riscv_option_override): Set riscv_stack_boundary. Handle
3134 riscv_preferred_stack_boundary_arg.
3135 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3136 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3137 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3138 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3139 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3140 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3142 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3145 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3146 of struct ix86_frame.
3147 (ix86_expand_epilogue): Likewise. Add a local variable for
3148 the reg_save_offset field in struct ix86_frame.
3150 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3152 PR tree-optimization/82604
3153 * tree-loop-distribution.c (enum partition_kind): New enum item
3154 PKIND_PARTIAL_MEMSET.
3155 (partition_builtin_p): Support above new enum item.
3156 (generate_code_for_partition): Ditto.
3157 (compute_access_range): Differentiate cases that equality can be
3158 proven at all loops, the innermost loops or no loops.
3159 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3160 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3161 (finalize_partitions, distribute_loop): Don't fuse partition of
3162 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3163 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3166 2018-01-23 Martin Liska <mliska@suse.cz>
3168 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3169 order to ignore the predictor.
3170 (PRED_POLYMORPHIC_CALL): Likewise.
3171 (PRED_RECURSIVE_CALL): Likewise.
3173 2018-01-23 Martin Liska <mliska@suse.cz>
3175 * tree-profile.c (tree_profiling): Print function header to
3176 aware reader which function we are working on.
3177 * value-prof.c (gimple_find_values_to_profile): Do not print
3178 not interesting value histograms.
3180 2018-01-23 Martin Liska <mliska@suse.cz>
3182 * profile-count.h (enum profile_quality): Add
3183 profile_uninitialized as the first value. Do not number values
3184 as they are zero based.
3185 (profile_count::verify): Update sanity check.
3186 (profile_probability::verify): Likewise.
3188 2018-01-23 Nathan Sidwell <nathan@acm.org>
3190 * doc/invoke.texi (ffor-scope): Deprecate.
3192 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3194 PR tree-optimization/83510
3195 * domwalk.c (set_all_edges_as_executable): New function.
3196 (dom_walker::dom_walker): Convert bool param
3197 "skip_unreachable_blocks" to enum reachability. Move setup of
3198 edge flags to set_all_edges_as_executable and only do it when
3199 reachability is REACHABLE_BLOCKS.
3200 * domwalk.h (enum dom_walker::reachability): New enum.
3201 (dom_walker::dom_walker): Convert bool param
3202 "skip_unreachable_blocks" to enum reachability.
3203 (set_all_edges_as_executable): New decl.
3204 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3205 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3207 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3208 but converting true to REACHABLE_BLOCKS.
3209 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3211 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3212 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3213 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3215 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3216 if check_all_array_refs will be called.
3218 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3220 * tree.c (selftest::test_location_wrappers): Add more test
3223 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3225 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3226 (selftest::test_bit_in_range): Likewise.
3228 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3231 * doc/sourcebuild.texi (vect_float): Say that the selector
3232 only describes the situation when -funsafe-math-optimizations is on.
3233 (vect_float_strict): Document.
3235 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3237 PR tree-optimization/83965
3238 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3239 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3240 instead of checking only for a reduction.
3241 (vect_recog_widen_sum_pattern): Likewise.
3243 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3245 * predict.c (probably_never_executed): Only use precise profile info.
3246 (compute_function_frequency): Skip after inlining hack since we now
3247 have quality checking.
3249 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3251 * profile-count.h (profile_probability::very_unlikely,
3252 profile_probability::unlikely, profile_probability::even): Set
3253 precision to guessed.
3255 2018-01-23 Richard Biener <rguenther@suse.de>
3257 PR tree-optimization/83963
3258 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3259 Properly terminate dominator walk when crossing the exit edge not
3260 when visiting its source block.
3262 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3265 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3266 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3268 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3270 PR tree-optimization/83957
3271 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3272 semicolon after for body surrounded by braces.
3274 PR tree-optimization/83081
3275 * profile-count.h (profile_probability::split): New method.
3276 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3277 Use profile_probability::split.
3278 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3279 when splitting a single conditional jump into 2.
3281 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3283 PR tree-optimization/69452
3284 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3287 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3289 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3290 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3291 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3293 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3295 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3296 * config/rl78/rl78.md (movdi): New define_expand.
3297 * config/rl78/rl78.c (rl78_split_movdi): New function.
3299 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3302 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3304 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3305 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3306 128-bit to produce an UNSPEC move to get the double word with the
3307 signbit and then a shift directly to do signbit.
3308 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3309 implementation with a new version that just does either a direct
3310 move or a regular move. Move memory interface to separate insns.
3311 Move insns so they are next to the expander.
3312 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3313 with signbit move. Split big and little endian case.
3314 (signbit<mode>2_dm_mem_le): Likewise.
3315 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3316 (signbit<mode>2_dm2): Likewise.
3318 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3320 * config/rl78/rl78.md (anddi3): New define_expand.
3322 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3324 * config/rl78/rl78.md (umindi3): New define_expand.
3326 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3328 * config/rl78/rl78.md (smindi3): New define_expand.
3330 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3332 * config/rl78/rl78.md (smaxdi3): New define_expand.
3334 2018-01-22 Carl Love <cel@us.ibm.com>
3336 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3337 LVX_V1TI): Add macro expansion.
3338 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3339 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3340 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3341 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3342 Change check to determine if the instruction is a byte reversing
3343 entry. Fix typo in comment.
3344 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3345 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3346 Add def_builtin calls for new builtins.
3347 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3348 Add define_insn expansion.
3350 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3352 * config/rl78/rl78.md (umaxdi3): New define_expand.
3354 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3356 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3357 for non-QImode registers.
3359 2018-01-22 Richard Biener <rguenther@suse.de>
3361 PR tree-optimization/83963
3362 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3363 including the loop exit block.
3364 (scop_detection::merge_sese): Likewise.
3365 (scop_detection::add_scop): Do it here instead.
3367 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3369 * doc/sourcebuild.texi (arm_softfloat): Document.
3371 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3374 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3375 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3376 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3378 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3379 David Edelsohn <dje.gcc@gmail.com>
3382 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3383 Change "crset eq" to "crset 2".
3384 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3385 (*call_indirect_aix<mode>_nospec): Likewise.
3386 (*call_value_indirect_aix<mode>_nospec): Likewise.
3387 (*call_indirect_elfv2<mode>_nospec): Likewise.
3388 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3389 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3390 change assembly output from . to $.
3391 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3392 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3393 (*tablejump<mode>_internal1_nospec): Likewise.
3395 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3398 * config/sh/sh_optimize_sett_clrt.cc:
3399 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3401 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3403 PR tree-optimization/83940
3404 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3405 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3406 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3407 use it to pass back the definition type.
3408 (vect_check_store_rhs): Likewise rhs_dt_out.
3409 (vect_build_gather_load_calls): Add a mask_dt argument and use
3410 it instead of a call to vect_is_simple_use.
3411 (vectorizable_store): Update calls to vect_check_load_store_mask
3412 and vect_check_store_rhs. Use the dt returned by the latter instead
3413 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3414 instead of calls to vect_is_simple_use. Pass the scalar rather
3415 than the vector operand to vect_is_simple_use when handling
3416 second and subsequent copies of an rhs value.
3417 (vectorizable_load): Update calls to vect_check_load_store_mask
3418 and vect_build_gather_load_calls. Use the cached mask_dt and
3419 gs_info.offset_dt instead of calls to vect_is_simple_use.
3421 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3424 * tree-emutls.c: Include gimplify.h.
3425 (lower_emutls_2): New function.
3426 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3427 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3428 it before further processing.
3431 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3432 UINTVAL (trueop1) instead of INTVAL (op1).
3434 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3438 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3439 INCOMING_FRAME_SP_OFFSET if not defined.
3440 (scan_trace): Add ENTRY argument. If true and
3441 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3442 emit a note to adjust the CFA offset.
3443 (create_cfi_notes): Adjust scan_trace callers.
3444 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3445 INCOMING_FRAME_SP_OFFSET in the CIE.
3446 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3447 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3449 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3450 * doc/tm.texi: Regenerated.
3452 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3454 PR rtl-optimization/83147
3455 * lra-constraints.c (remove_inheritance_pseudos): Use
3456 lra_substitute_pseudo_within_insn.
3458 2018-01-19 Tom de Vries <tom@codesourcery.com>
3459 Cesar Philippidis <cesar@codesourcery.com>
3462 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3464 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3467 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3468 spaces for function labels.
3470 2018-01-19 Martin Liska <mliska@suse.cz>
3472 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3473 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3474 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3475 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3476 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3477 (PRED_CONST_RETURN): Change from 69 to 65.
3478 (PRED_NULL_RETURN): Change from 91 to 71.
3479 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3480 (PRED_LOOP_GUARD): Change from 66 to 73.
3482 2018-01-19 Martin Liska <mliska@suse.cz>
3484 * predict.c (predict_insn_def): Add new assert.
3485 (struct branch_predictor): Change type to signed integer.
3486 (test_prediction_value_range): Amend test to cover
3488 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3489 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3490 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3491 (PRED_LOOP_IV_COMPARE): Likewise.
3492 * predict.h (PROB_UNINITIALIZED): Define new constant.
3494 2018-01-19 Martin Liska <mliska@suse.cz>
3496 * predict.c (dump_prediction): Add new format for
3497 analyze_brprob.py script which is enabled with -details
3499 * profile-count.h (precise_p): New function.
3501 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3503 PR tree-optimization/83922
3504 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3505 there are no statements that need masking.
3506 (vect_active_double_reduction_p): New function.
3507 (vect_analyze_loop_operations): Use it when handling phis that
3508 are not in the loop header.
3510 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3512 PR tree-optimization/83914
3513 * tree-vect-loop.c (vectorizable_induction): Don't convert
3514 init_expr or apply the peeling adjustment for inductions
3515 that are nested within the vectorized loop.
3517 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3519 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3522 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3526 * function.h (gimplify_parameters): Add gimple_seq * argument.
3527 * function.c: Include gimple.h and options.h.
3528 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3529 for the added local temporaries if needed.
3530 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3531 if there are any parameter cleanups, wrap whole body into a
3532 try/finally with the cleanups.
3534 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3537 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3538 Use GET_MODE_CLASS for scalar floating point.
3540 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3544 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3545 Fix call of call_cgraph_insertion_hooks.
3547 2018-01-18 Martin Sebor <msebor@redhat.com>
3549 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3551 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3554 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3557 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3560 * common.opt: (-ffile-prefix-map): New option.
3561 * opts.c (common_handle_option): Defer it.
3562 * opts-global.c (handle_common_deferred_options): Handle it.
3563 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3564 * file-prefix-map.h: New file.
3565 (remap_debug_filename, add_debug_prefix_map): ...here.
3566 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3567 * final.c (debug_prefix_map, add_debug_prefix_map
3568 remap_debug_filename): Move to...
3569 * file-prefix-map.c: New file.
3570 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3571 generalize, get rid of alloca(), use strrchr() instead of strchr().
3572 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3573 Implement in terms of add_prefix_map().
3574 (remap_macro_filename, remap_debug_filename): Implement in term of
3576 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3577 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3578 * dbxout.c: Include file-prefix-map.h.
3579 * varasm.c: Likewise.
3580 * vmsdbgout.c: Likewise.
3581 * xcoffout.c: Likewise.
3582 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3583 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3584 * doc/invoke.texi (-ffile-prefix-map): Document.
3585 (-fdebug-prefix-map): Update description.
3587 2018-01-18 Martin Liska <mliska@suse.cz>
3589 * config/i386/i386.c (indirect_thunk_name): Document that also
3591 (output_indirect_thunk): Document why both instructions
3592 (pause and lfence) are generated.
3594 2018-01-18 Richard Biener <rguenther@suse.de>
3596 PR tree-optimization/83887
3597 * graphite-scop-detection.c
3598 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3599 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3600 (scop_detection::merge_sese): Re-implement with a flood-fill
3601 algorithm that properly finds a SESE region if it exists.
3603 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3606 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3607 pointer_diff optimizations use view_convert instead of convert.
3609 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3611 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3612 Generate different code for -mno-speculate-indirect-jumps.
3613 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3614 (*call_indirect_aix<mode>): Disable for
3615 -mno-speculate-indirect-jumps.
3616 (*call_indirect_aix<mode>_nospec): New define_insn.
3617 (*call_value_indirect_aix<mode>): Disable for
3618 -mno-speculate-indirect-jumps.
3619 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3620 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3621 -mno-speculate-indirect-jumps.
3622 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3624 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3626 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3627 long double type, set the flags for noting the default long double
3628 type, even if we don't pass or return a long double type.
3630 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3633 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3636 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3638 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3639 support for merge[hl].
3640 (fold_mergehl_helper): New helper function.
3641 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3642 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3643 (altivec_vmrglw_direct): Add xxmrglw insn.
3645 2018-01-17 Andrew Waterman <andrew@sifive.com>
3647 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3648 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3650 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3653 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3654 call the lto_location_cache before reading the
3655 DECL_SOURCE_LOCATION of the types.
3657 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3658 Richard Sandiford <richard.sandiford@linaro.org>
3660 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3661 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3662 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3663 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3664 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3666 * config/aarch64/constraints.md (aarch64_movti_operand):
3668 * config/aarch64/predicates.md (Uti): Add new constraint.
3670 2018-01-17 Carl Love <cel@us.ibm.com>
3672 * config/rs6000/vsx.md (define_expand xl_len_r,
3673 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3674 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3676 (define_expand, define_insn): Move the shift left from the
3677 define_insn to the define_expand for lxvl and stxvl instructions.
3678 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3679 and XL_LEN_R definitions to PURE.
3681 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3683 * config/i386/i386.c (indirect_thunk_name): Declare regno
3684 as unsigned int. Compare regno with INVALID_REGNUM.
3685 (output_indirect_thunk): Ditto.
3686 (output_indirect_thunk_function): Ditto.
3687 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3688 in the call to output_indirect_thunk_function.
3690 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3693 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3694 rather than the size of inner_type to determine the stack slot size
3695 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3697 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3700 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3703 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3705 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3706 endian Linux systems to optionally enable multilibs for selecting
3707 the long double type if the user configured an explicit type.
3708 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3709 have no long double multilibs if not defined.
3710 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3711 warn if the user used -mabi={ieee,ibm}longdouble and we built
3712 multilibs for long double.
3713 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3714 appropriate multilib option.
3715 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3717 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3718 for building long double multilibs.
3719 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3721 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3723 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3726 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3728 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3731 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3734 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3737 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3739 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3740 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3742 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3744 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3745 different rtl trees depending on TARGET_64BIT.
3746 (rs6000_gen_lvx): Likewise.
3748 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3750 * config/visium/visium.md (nop): Tweak comment.
3751 (hazard_nop): Likewise.
3753 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3755 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3756 -mspeculate-indirect-jumps.
3757 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3758 for -mno-speculate-indirect-jumps.
3759 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3760 (*call_value_indirect_elfv2<mode>): Disable for
3761 -mno-speculate-indirect-jumps.
3762 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3763 (indirect_jump): Emit different RTL for
3764 -mno-speculate-indirect-jumps.
3765 (*indirect_jump<mode>): Disable for
3766 -mno-speculate-indirect-jumps.
3767 (*indirect_jump<mode>_nospec): New define_insn.
3768 (tablejump): Emit different RTL for
3769 -mno-speculate-indirect-jumps.
3770 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3771 (tablejumpsi_nospec): New define_expand.
3772 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3773 (tablejumpdi_nospec): New define_expand.
3774 (*tablejump<mode>_internal1): Disable for
3775 -mno-speculate-indirect-jumps.
3776 (*tablejump<mode>_internal1_nospec): New define_insn.
3777 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3780 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3782 * caller-save.c (insert_save): Drop unnecessary parameter. All
3785 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3786 Richard Biener <rguenth@suse.de>
3789 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3790 return early, inline manually is_gimple_sizepos. Make sure if we
3791 call gimplify_expr we don't end up with a gimple constant.
3792 * tree.c (variably_modified_type_p): Don't return true for
3793 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3794 * gimplify.h (is_gimple_sizepos): Remove.
3796 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3798 PR tree-optimization/83857
3799 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3800 vectorizable_live_operation for pure SLP statements.
3801 (vectorizable_live_operation): Handle PHIs.
3803 2018-01-16 Richard Biener <rguenther@suse.de>
3805 PR tree-optimization/83867
3806 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3807 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3809 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3812 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3813 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3814 If off is not INTEGER_CST, issue a may not be aligned warning
3815 rather than isn't aligned. Use isn%'t rather than isn't.
3816 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3818 <case MULT_EXPR>: Improve the case when bottom and one of the
3819 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3820 operand, in that case check if the other operand is multiple of
3821 bottom divided by the INTEGER_CST operand.
3823 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3826 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3827 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3828 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3829 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3830 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3831 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3832 (pa_function_arg_size): New function.
3834 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3836 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3837 in a separate statement.
3839 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3841 PR tree-optimization/83847
3842 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3843 group gathers and scatters.
3845 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3847 PR rtl-optimization/86620
3848 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3850 PR rtl-optimization/83213
3851 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3852 to last if both are JUMP_INSNs.
3854 PR tree-optimization/83843
3855 * gimple-ssa-store-merging.c
3856 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3857 store_immediate_info for bswap/nop orig_stores.
3859 2018-01-15 Andrew Waterman <andrew@sifive.com>
3861 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3863 <UDIV>: Increase cost if !TARGET_DIV.
3865 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3867 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3868 (define_attr "cr_logical_3op"): New.
3869 (cceq_ior_compare): Adjust.
3870 (cceq_ior_compare_complement): Adjust.
3871 (*cceq_rev_compare): Adjust.
3872 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3873 (is_cracked_insn): Adjust.
3874 (insn_must_be_first_in_group): Adjust.
3875 * config/rs6000/40x.md: Adjust.
3876 * config/rs6000/440.md: Adjust.
3877 * config/rs6000/476.md: Adjust.
3878 * config/rs6000/601.md: Adjust.
3879 * config/rs6000/603.md: Adjust.
3880 * config/rs6000/6xx.md: Adjust.
3881 * config/rs6000/7450.md: Adjust.
3882 * config/rs6000/7xx.md: Adjust.
3883 * config/rs6000/8540.md: Adjust.
3884 * config/rs6000/cell.md: Adjust.
3885 * config/rs6000/e300c2c3.md: Adjust.
3886 * config/rs6000/e500mc.md: Adjust.
3887 * config/rs6000/e500mc64.md: Adjust.
3888 * config/rs6000/e5500.md: Adjust.
3889 * config/rs6000/e6500.md: Adjust.
3890 * config/rs6000/mpc.md: Adjust.
3891 * config/rs6000/power4.md: Adjust.
3892 * config/rs6000/power5.md: Adjust.
3893 * config/rs6000/power6.md: Adjust.
3894 * config/rs6000/power7.md: Adjust.
3895 * config/rs6000/power8.md: Adjust.
3896 * config/rs6000/power9.md: Adjust.
3897 * config/rs6000/rs64.md: Adjust.
3898 * config/rs6000/titan.md: Adjust.
3900 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3902 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3903 ix86_indirect_branch_register logic.
3905 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3907 * config/i386/constraints.md (Bs): Update
3908 ix86_indirect_branch_register check. Don't check
3909 ix86_indirect_branch_register with GOT_memory_operand.
3911 * config/i386/predicates.md (GOT_memory_operand): Don't check
3912 ix86_indirect_branch_register here.
3913 (GOT32_symbol_operand): Likewise.
3915 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3917 * config/i386/predicates.md (constant_call_address_operand):
3918 Rewrite ix86_indirect_branch_register logic.
3919 (sibcall_insn_operand): Likewise.
3921 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3923 * config/i386/constraints.md (Bs): Replace
3924 ix86_indirect_branch_thunk_register with
3925 ix86_indirect_branch_register.
3927 * config/i386/i386.md (indirect_jump): Likewise.
3928 (tablejump): Likewise.
3929 (*sibcall_memory): Likewise.
3930 (*sibcall_value_memory): Likewise.
3931 Peepholes of indirect call and jump via memory: Likewise.
3932 * config/i386/i386.opt: Likewise.
3933 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3934 (GOT_memory_operand): Likewise.
3935 (call_insn_operand): Likewise.
3936 (sibcall_insn_operand): Likewise.
3937 (GOT32_symbol_operand): Likewise.
3939 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3942 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3943 type rather than type addr's type points to.
3944 (expand_omp_atomic_mutex): Likewise.
3945 (expand_omp_atomic): Likewise.
3947 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3950 * config/i386/i386.c (output_indirect_thunk_function): Use
3951 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3952 for __x86_return_thunk.
3954 2018-01-15 Richard Biener <rguenther@suse.de>
3957 * expmed.c (extract_bit_field_1): Fix typo.
3959 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3962 * config/arm/iterators.md (VF): New mode iterator.
3963 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3964 Remove integer-related logic from pattern.
3965 (neon_vabd<mode>_3): Likewise.
3967 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3970 * common.opt (fstrict-overflow): No longer an alias.
3971 (fwrapv-pointer): New option.
3972 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3973 also for pointer types based on flag_wrapv_pointer.
3974 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3975 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3976 opts->x_flag_wrapv got set.
3977 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3978 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3979 POINTER_TYPE_OVERFLOW_UNDEFINED.
3980 * match.pd: Likewise in address comparison pattern.
3981 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3983 2018-01-15 Richard Biener <rguenther@suse.de>
3986 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3987 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3988 Reset type names to their identifier if their TYPE_DECL doesn't
3989 have linkage (and thus is used for ODR and devirt).
3990 (save_debug_info_for_decl): Remove.
3991 (save_debug_info_for_type): Likewise.
3992 (add_tree_to_fld_list): Adjust.
3993 * tree-pretty-print.c (dump_generic_node): Make dumping of
3994 type names more robust.
3996 2018-01-15 Richard Biener <rguenther@suse.de>
3998 * BASE-VER: Bump to 8.0.1.
4000 2018-01-14 Martin Sebor <msebor@redhat.com>
4003 * builtins.c (check_access): Avoid warning when the no-warning bit
4006 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4008 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4009 * ira-color (allocno_hard_regs_compare): Likewise.
4011 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4014 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4015 Use .pushsection/.popsection.
4017 2018-01-14 Martin Sebor <msebor@redhat.com>
4020 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4022 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4024 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4025 entry from extra_headers.
4026 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4027 extra_headers, make the list bitwise identical to the i?86-*-* one.
4029 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4031 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4032 -mcmodel=large with -mindirect-branch=thunk,
4033 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4034 -mfunction-return=thunk-extern.
4035 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4036 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4037 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4039 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4041 * config/i386/i386.c (print_reg): Print the name of the full
4042 integer register without '%'.
4043 (ix86_print_operand): Handle 'V'.
4044 * doc/extend.texi: Document 'V' modifier.
4046 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4048 * config/i386/constraints.md (Bs): Disallow memory operand for
4049 -mindirect-branch-register.
4051 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4052 (GOT_memory_operand): Likewise.
4053 (call_insn_operand): Likewise.
4054 (sibcall_insn_operand): Likewise.
4055 (GOT32_symbol_operand): Likewise.
4056 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4057 for -mindirect-branch-register.
4058 (tablejump): Likewise.
4059 (*sibcall_memory): Likewise.
4060 (*sibcall_value_memory): Likewise.
4061 Disallow peepholes of indirect call and jump via memory for
4062 -mindirect-branch-register.
4063 (*call_pop): Replace m with Bw.
4064 (*call_value_pop): Likewise.
4065 (*sibcall_pop_memory): Replace m with Bs.
4066 * config/i386/i386.opt (mindirect-branch-register): New option.
4067 * doc/invoke.texi: Document -mindirect-branch-register option.
4069 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4071 * config/i386/i386-protos.h (ix86_output_function_return): New.
4072 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4073 set function_return_type.
4074 (indirect_thunk_name): Add ret_p to indicate thunk for function
4076 (output_indirect_thunk_function): Pass false to
4077 indirect_thunk_name.
4078 (ix86_output_indirect_branch_via_reg): Likewise.
4079 (ix86_output_indirect_branch_via_push): Likewise.
4080 (output_indirect_thunk_function): Create alias for function
4081 return thunk if regno < 0.
4082 (ix86_output_function_return): New function.
4083 (ix86_handle_fndecl_attribute): Handle function_return.
4084 (ix86_attribute_table): Add function_return.
4085 * config/i386/i386.h (machine_function): Add
4086 function_return_type.
4087 * config/i386/i386.md (simple_return_internal): Use
4088 ix86_output_function_return.
4089 (simple_return_internal_long): Likewise.
4090 * config/i386/i386.opt (mfunction-return=): New option.
4091 (indirect_branch): Mention -mfunction-return=.
4092 * doc/extend.texi: Document function_return function attribute.
4093 * doc/invoke.texi: Document -mfunction-return= option.
4095 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4097 * config/i386/i386-opts.h (indirect_branch): New.
4098 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4099 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4100 with local indirect jump when converting indirect call and jump.
4101 (ix86_set_indirect_branch_type): New.
4102 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4103 (indirectlabelno): New.
4104 (indirect_thunk_needed): Likewise.
4105 (indirect_thunk_bnd_needed): Likewise.
4106 (indirect_thunks_used): Likewise.
4107 (indirect_thunks_bnd_used): Likewise.
4108 (INDIRECT_LABEL): Likewise.
4109 (indirect_thunk_name): Likewise.
4110 (output_indirect_thunk): Likewise.
4111 (output_indirect_thunk_function): Likewise.
4112 (ix86_output_indirect_branch_via_reg): Likewise.
4113 (ix86_output_indirect_branch_via_push): Likewise.
4114 (ix86_output_indirect_branch): Likewise.
4115 (ix86_output_indirect_jmp): Likewise.
4116 (ix86_code_end): Call output_indirect_thunk_function if needed.
4117 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4119 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4120 (ix86_attribute_table): Add indirect_branch.
4121 * config/i386/i386.h (machine_function): Add indirect_branch_type
4122 and has_local_indirect_jump.
4123 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4125 (tablejump): Likewise.
4126 (*indirect_jump): Use ix86_output_indirect_jmp.
4127 (*tablejump_1): Likewise.
4128 (simple_return_indirect_internal): Likewise.
4129 * config/i386/i386.opt (mindirect-branch=): New option.
4130 (indirect_branch): New.
4133 (thunk-inline): Likewise.
4134 (thunk-extern): Likewise.
4135 * doc/extend.texi: Document indirect_branch function attribute.
4136 * doc/invoke.texi: Document -mindirect-branch= option.
4138 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4141 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4143 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4145 * ipa-inline.c (want_inline_small_function_p): Return false if
4146 inlining has already failed with CIF_FINAL_ERROR.
4147 (update_caller_keys): Call want_inline_small_function_p before
4149 (update_callee_keys): Likewise.
4151 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4153 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4155 (rs6000_quadword_masked_address_p): Likewise.
4156 (quad_aligned_load_p): Likewise.
4157 (quad_aligned_store_p): Likewise.
4158 (const_load_sequence_p): Add comment to describe the outer-most loop.
4159 (mimic_memory_attributes_and_flags): New function.
4160 (rs6000_gen_stvx): Likewise.
4161 (replace_swapped_aligned_store): Likewise.
4162 (rs6000_gen_lvx): Likewise.
4163 (replace_swapped_aligned_load): Likewise.
4164 (replace_swapped_load_constant): Capitalize argument name in
4165 comment describing this function.
4166 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4167 and stores that access quad-word aligned addresses and replace
4168 with stvx or lvx instructions when appropriate.
4169 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4170 New function prototype.
4171 (rs6000_quadword_masked_address_p): Likewise.
4172 (rs6000_gen_lvx): Likewise.
4173 (rs6000_gen_stvx): Likewise.
4174 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4175 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4176 when memory address is aligned.
4177 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4178 this split to select lvx instruction when memory address is aligned.
4179 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4180 instruction when memory address is aligned.
4181 (*vsx_le_perm_load_v16qi): Likewise.
4182 (four unnamed splitters): Modify to select the stvx instruction
4183 when memory is aligned.
4185 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4187 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4188 which appears in the queue multiple times.
4190 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4191 Alan Hayward <alan.hayward@arm.com>
4192 David Sherwood <david.sherwood@arm.com>
4194 * tree-vectorizer.h (vec_lower_bound): New structure.
4195 (_loop_vec_info): Add check_nonzero and lower_bounds.
4196 (LOOP_VINFO_CHECK_NONZERO): New macro.
4197 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4198 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4199 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4200 fields. Make seg_len the distance travelled, not including the
4202 (dr_direction_indicator): Declare.
4203 (dr_zero_step_indicator): Likewise.
4204 (dr_known_forward_stride_p): Likewise.
4205 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4207 (runtime_alias_check_p): Allow runtime alias checks with
4209 (operator ==): Compare access_size and align.
4210 (prune_runtime_alias_test_list): Rework for new distinction between
4211 the access_size and seg_len.
4212 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4214 (get_segment_min_max): New function.
4215 (create_intersect_range_checks): Use it.
4216 (dr_step_indicator): New function.
4217 (dr_direction_indicator): Likewise.
4218 (dr_zero_step_indicator): Likewise.
4219 (dr_known_forward_stride_p): Likewise.
4220 * tree-loop-distribution.c (data_ref_segment_size): Return
4221 DR_STEP * (niters - 1).
4222 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4224 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4225 (vect_preserves_scalar_order_p): New function, split out from...
4226 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4227 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4228 (vect_vfa_access_size): New function.
4229 (vect_vfa_align): Likewise.
4230 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4231 (dump_lower_bound): New function.
4232 (vect_check_lower_bound): Likewise.
4233 (vect_small_gap_p): Likewise.
4234 (vectorizable_with_step_bound_p): Likewise.
4235 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4236 depencies if the vectorization factor is 1. Convert the checks
4237 for nonzero steps into checks on the bounds of DR_STEP. Try using
4238 a bunds check for variable steps if the minimum required step is
4239 relatively small. Update calls to the dr_with_seg_len
4240 constructor and to vect_compile_time_alias.
4241 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4243 (vect_loop_versioning): Call it.
4244 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4246 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4248 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4249 Alan Hayward <alan.hayward@arm.com>
4250 David Sherwood <david.sherwood@arm.com>
4252 * doc/sourcebuild.texi (vect_scatter_store): Document.
4253 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4255 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4257 * genopinit.c (main): Add supports_vec_scatter_store and
4258 supports_vec_scatter_store_cached to target_optabs.
4259 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4260 IFN_MASK_SCATTER_STORE.
4261 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4263 * internal-fn.h (internal_store_fn_p): Declare.
4264 (internal_fn_stored_value_index): Likewise.
4265 * internal-fn.c (scatter_store_direct): New macro.
4266 (expand_scatter_store_optab_fn): New function.
4267 (direct_scatter_store_optab_supported_p): New macro.
4268 (internal_store_fn_p): New function.
4269 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4270 IFN_MASK_SCATTER_STORE.
4271 (internal_fn_mask_index): Likewise.
4272 (internal_fn_stored_value_index): New function.
4273 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4275 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4276 * optabs-query.c (supports_vec_scatter_store_p): New function.
4277 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4278 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4279 true for scatter stores.
4280 (vect_gather_scatter_fn_p): Handle scatter stores too.
4281 (vect_check_gather_scatter): Consider using scatter stores if
4282 supports_vec_scatter_store_p.
4283 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4285 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4286 internal_fn_stored_value_index.
4287 (check_load_store_masking): Handle scatter stores too.
4288 (vect_get_store_rhs): Make public.
4289 (vectorizable_call): Use internal_store_fn_p.
4290 (vectorizable_store): Handle scatter store internal functions.
4291 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4292 when deciding whether the end of the group has been reached.
4293 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4294 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4295 (mask_scatter_store<mode>): New insns.
4297 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4298 Alan Hayward <alan.hayward@arm.com>
4299 David Sherwood <david.sherwood@arm.com>
4301 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4302 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4303 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4305 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4306 Use vect_truncate_gather_scatter_offset if we can't treat the
4307 operation as a normal gather load or scatter store.
4308 (get_group_load_store_type): Take the gather_scatter_info
4309 as argument. Try using a gather load or scatter store for
4310 single-element groups.
4311 (get_load_store_type): Update calls to get_group_load_store_type
4312 and vect_use_strided_gather_scatters_p.
4314 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4315 Alan Hayward <alan.hayward@arm.com>
4316 David Sherwood <david.sherwood@arm.com>
4318 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4319 optional tree argument.
4320 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4322 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4323 but continue to use the current value as a fallback.
4324 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4325 to compare the updates.
4326 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4327 (get_load_store_type): Use it when handling a strided access.
4328 (vect_get_strided_load_store_ops): New function.
4329 (vect_get_data_ptr_increment): Likewise.
4330 (vectorizable_load): Handle strided gather loads. Always pass
4331 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4333 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4334 Alan Hayward <alan.hayward@arm.com>
4335 David Sherwood <david.sherwood@arm.com>
4337 * doc/md.texi (gather_load@var{m}): Document.
4338 (mask_gather_load@var{m}): Likewise.
4339 * genopinit.c (main): Add supports_vec_gather_load and
4340 supports_vec_gather_load_cached to target_optabs.
4341 * optabs-tree.c (init_tree_optimization_optabs): Use
4342 ggc_cleared_alloc to allocate target_optabs.
4343 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4344 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4346 * internal-fn.h (internal_load_fn_p): Declare.
4347 (internal_gather_scatter_fn_p): Likewise.
4348 (internal_fn_mask_index): Likewise.
4349 (internal_gather_scatter_fn_supported_p): Likewise.
4350 * internal-fn.c (gather_load_direct): New macro.
4351 (expand_gather_load_optab_fn): New function.
4352 (direct_gather_load_optab_supported_p): New macro.
4353 (direct_internal_fn_optab): New function.
4354 (internal_load_fn_p): Likewise.
4355 (internal_gather_scatter_fn_p): Likewise.
4356 (internal_fn_mask_index): Likewise.
4357 (internal_gather_scatter_fn_supported_p): Likewise.
4358 * optabs-query.c (supports_at_least_one_mode_p): New function.
4359 (supports_vec_gather_load_p): Likewise.
4360 * optabs-query.h (supports_vec_gather_load_p): Declare.
4361 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4362 and memory_type field.
4363 (NUM_PATTERNS): Bump to 15.
4364 * tree-vect-data-refs.c: Include internal-fn.h.
4365 (vect_gather_scatter_fn_p): New function.
4366 (vect_describe_gather_scatter_call): Likewise.
4367 (vect_check_gather_scatter): Try using internal functions for
4368 gather loads. Recognize existing calls to a gather load function.
4369 (vect_analyze_data_refs): Consider using gather loads if
4370 supports_vec_gather_load_p.
4371 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4372 (vect_get_gather_scatter_offset_type): Likewise.
4373 (vect_convert_mask_for_vectype): Likewise.
4374 (vect_add_conversion_to_patterm): Likewise.
4375 (vect_try_gather_scatter_pattern): Likewise.
4376 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4377 (vect_vect_recog_func_ptrs): Add it.
4378 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4379 internal_fn_mask_index and internal_gather_scatter_fn_p.
4380 (check_load_store_masking): Take the gather_scatter_info as an
4381 argument and handle gather loads.
4382 (vect_get_gather_scatter_ops): New function.
4383 (vectorizable_call): Check internal_load_fn_p.
4384 (vectorizable_load): Likewise. Handle gather load internal
4386 (vectorizable_store): Update call to check_load_store_masking.
4387 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4388 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4389 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4390 (aarch64_gather_scale_operand_d): New predicates.
4391 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4392 (mask_gather_load<mode>): New insns.
4394 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4395 Alan Hayward <alan.hayward@arm.com>
4396 David Sherwood <david.sherwood@arm.com>
4398 * optabs.def (fold_left_plus_optab): New optab.
4399 * doc/md.texi (fold_left_plus_@var{m}): Document.
4400 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4401 * internal-fn.c (fold_left_direct): Define.
4402 (expand_fold_left_optab_fn): Likewise.
4403 (direct_fold_left_optab_supported_p): Likewise.
4404 * fold-const-call.c (fold_const_fold_left): New function.
4405 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4406 * tree-parloops.c (valid_reduction_p): New function.
4407 (gather_scalar_reductions): Use it.
4408 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4409 (vect_finish_replace_stmt): Declare.
4410 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4411 (needs_fold_left_reduction_p): New function, split out from...
4412 (vect_is_simple_reduction): ...here. Accept reductions that
4413 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4414 (vect_force_simple_reduction): Also store the reduction type in
4415 the assignment's STMT_VINFO_REDUC_TYPE.
4416 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4417 (merge_with_identity): New function.
4418 (vect_expand_fold_left): Likewise.
4419 (vectorize_fold_left_reduction): Likewise.
4420 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4421 scalar phi in place for it. Check for target support and reject
4422 cases that would reassociate the operation. Defer the transform
4423 phase to vectorize_fold_left_reduction.
4424 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4425 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4426 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4428 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4430 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4431 call to ifc_temp_var.
4433 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4434 Alan Hayward <alan.hayward@arm.com>
4435 David Sherwood <david.sherwood@arm.com>
4437 * target.def (legitimize_address_displacement): Take the original
4438 offset as a poly_int.
4439 * targhooks.h (default_legitimize_address_displacement): Update
4441 * targhooks.c (default_legitimize_address_displacement): Likewise.
4442 * doc/tm.texi: Regenerate.
4443 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4444 as an argument, moving assert of ad->disp == ad->disp_term to...
4445 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4446 Try calling targetm.legitimize_address_displacement before expanding
4447 the address rather than afterwards, and adjust for the new interface.
4448 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4449 Match the new hook interface. Handle SVE addresses.
4450 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4453 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4455 * Makefile.in (OBJS): Add early-remat.o.
4456 * target.def (select_early_remat_modes): New hook.
4457 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4458 * doc/tm.texi: Regenerate.
4459 * targhooks.h (default_select_early_remat_modes): Declare.
4460 * targhooks.c (default_select_early_remat_modes): New function.
4461 * timevar.def (TV_EARLY_REMAT): New timevar.
4462 * passes.def (pass_early_remat): New pass.
4463 * tree-pass.h (make_pass_early_remat): Declare.
4464 * early-remat.c: New file.
4465 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4467 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4469 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4470 Alan Hayward <alan.hayward@arm.com>
4471 David Sherwood <david.sherwood@arm.com>
4473 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4474 vfm1 with a bound_epilog parameter.
4475 (vect_do_peeling): Update calls accordingly, and move the prologue
4476 call earlier in the function. Treat the base bound_epilog as 0 for
4477 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4478 this base when peeling for gaps.
4479 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4480 with fully-masked loops.
4481 (vect_estimate_min_profitable_iters): Handle the single peeled
4482 iteration in that case.
4484 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4485 Alan Hayward <alan.hayward@arm.com>
4486 David Sherwood <david.sherwood@arm.com>
4488 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4489 single-element interleaving even if the size is not a power of 2.
4490 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4491 accesses for single-element interleaving if the group size is
4494 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4495 Alan Hayward <alan.hayward@arm.com>
4496 David Sherwood <david.sherwood@arm.com>
4498 * doc/md.texi (fold_extract_last_@var{m}): Document.
4499 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4500 * optabs.def (fold_extract_last_optab): New optab.
4501 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4502 * internal-fn.c (fold_extract_direct): New macro.
4503 (expand_fold_extract_optab_fn): Likewise.
4504 (direct_fold_extract_optab_supported_p): Likewise.
4505 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4506 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4507 EXTRACT_LAST_REDUCTION.
4508 (get_initial_def_for_reduction): Do not create an initial vector
4509 for EXTRACT_LAST_REDUCTION reductions.
4510 (vectorizable_reduction): Leave the scalar phi in place for
4511 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4512 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4513 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4514 transform phase to vectorizable_condition.
4515 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4517 (vect_finish_stmt_generation): ...here.
4518 (vect_finish_replace_stmt): New function.
4519 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4520 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4522 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4524 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4525 Alan Hayward <alan.hayward@arm.com>
4526 David Sherwood <david.sherwood@arm.com>
4528 * doc/md.texi (extract_last_@var{m}): Document.
4529 * optabs.def (extract_last_optab): New optab.
4530 * internal-fn.def (EXTRACT_LAST): New internal function.
4531 * internal-fn.c (cond_unary_direct): New macro.
4532 (expand_cond_unary_optab_fn): Likewise.
4533 (direct_cond_unary_optab_supported_p): Likewise.
4534 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4535 loops using EXTRACT_LAST.
4536 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4537 (extract_last_<mode>): ...this optab.
4538 (vec_extract<mode><Vel>): Update accordingly.
4540 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4541 Alan Hayward <alan.hayward@arm.com>
4542 David Sherwood <david.sherwood@arm.com>
4544 * target.def (empty_mask_is_expensive): New hook.
4545 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4546 * doc/tm.texi: Regenerate.
4547 * targhooks.h (default_empty_mask_is_expensive): Declare.
4548 * targhooks.c (default_empty_mask_is_expensive): New function.
4549 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4550 if the target says that empty masks are expensive.
4551 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4553 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4555 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4556 Alan Hayward <alan.hayward@arm.com>
4557 David Sherwood <david.sherwood@arm.com>
4559 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4560 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4561 (vect_use_loop_mask_for_alignment_p): New function.
4562 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4563 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4564 niters_skip argument. Make sure that the first niters_skip elements
4565 of the first iteration are inactive.
4566 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4567 Update call to vect_set_loop_masks_directly.
4568 (get_misalign_in_elems): New function, split out from...
4569 (vect_gen_prolog_loop_niters): ...here.
4570 (vect_update_init_of_dr): Take a code argument that specifies whether
4571 the adjustment should be added or subtracted.
4572 (vect_update_init_of_drs): Likewise.
4573 (vect_prepare_for_masked_peels): New function.
4574 (vect_do_peeling): Skip prologue peeling if we're using a mask
4575 instead. Update call to vect_update_inits_of_drs.
4576 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4578 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4579 alignment. Do not include the number of peeled iterations in
4580 the minimum threshold in that case.
4581 (vectorizable_induction): Adjust the start value down by
4582 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4583 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4584 Take the number of skipped iterations into account when calculating
4586 * tree-vect-stmts.c (vect_gen_while_not): New function.
4588 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4589 Alan Hayward <alan.hayward@arm.com>
4590 David Sherwood <david.sherwood@arm.com>
4592 * doc/sourcebuild.texi (vect_fully_masked): Document.
4593 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4595 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4597 (vect_analyze_loop_2): ...here. Don't check the vectorization
4598 factor against the number of loop iterations if the loop is
4601 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4602 Alan Hayward <alan.hayward@arm.com>
4603 David Sherwood <david.sherwood@arm.com>
4605 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4606 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4607 (dump_groups): Update accordingly.
4608 (iv_use::mem_type): New member variable.
4609 (address_p): New function.
4610 (record_use): Add a mem_type argument and initialize the new
4612 (record_group_use): Add a mem_type argument. Use address_p.
4613 Remove obsolete null checks of base_object. Update call to record_use.
4614 (find_interesting_uses_op): Update call to record_group_use.
4615 (find_interesting_uses_cond): Likewise.
4616 (find_interesting_uses_address): Likewise.
4617 (get_mem_type_for_internal_fn): New function.
4618 (find_address_like_use): Likewise.
4619 (find_interesting_uses_stmt): Try find_address_like_use before
4620 calling find_interesting_uses_op.
4621 (addr_offset_valid_p): Use the iv mem_type field as the type
4622 of the addressed memory.
4623 (add_autoinc_candidates): Likewise.
4624 (get_address_cost): Likewise.
4625 (split_small_address_groups_p): Use address_p.
4626 (split_address_groups): Likewise.
4627 (add_iv_candidate_for_use): Likewise.
4628 (autoinc_possible_for_pair): Likewise.
4629 (rewrite_groups): Likewise.
4630 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4631 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4632 (get_alias_ptr_type_for_ptr_address): New function.
4633 (rewrite_use_address): Rewrite address uses in calls that were
4634 identified by find_address_like_use.
4636 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4637 Alan Hayward <alan.hayward@arm.com>
4638 David Sherwood <david.sherwood@arm.com>
4640 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4642 * gimple-expr.h (is_gimple_addressable: Likewise.
4643 * gimple-expr.c (is_gimple_address): Likewise.
4644 * internal-fn.c (expand_call_mem_ref): New function.
4645 (expand_mask_load_optab_fn): Use it.
4646 (expand_mask_store_optab_fn): Likewise.
4648 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4649 Alan Hayward <alan.hayward@arm.com>
4650 David Sherwood <david.sherwood@arm.com>
4652 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4653 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4654 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4655 (cond_umax@var{mode}): Document.
4656 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4657 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4658 (cond_umin_optab, cond_umax_optab): New optabs.
4659 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4660 (COND_IOR, COND_XOR): New internal functions.
4661 * internal-fn.h (get_conditional_internal_fn): Declare.
4662 * internal-fn.c (cond_binary_direct): New macro.
4663 (expand_cond_binary_optab_fn): Likewise.
4664 (direct_cond_binary_optab_supported_p): Likewise.
4665 (get_conditional_internal_fn): New function.
4666 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4667 Cope with reduction statements that are vectorized as calls rather
4669 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4670 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4671 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4672 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4673 (UNSPEC_COND_EOR): New unspecs.
4674 (optab): Add mappings for them.
4675 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4676 (sve_int_op, sve_fp_op): New int attributes.
4678 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4679 Alan Hayward <alan.hayward@arm.com>
4680 David Sherwood <david.sherwood@arm.com>
4682 * optabs.def (while_ult_optab): New optab.
4683 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4684 * internal-fn.def (WHILE_ULT): New internal function.
4685 * internal-fn.h (direct_internal_fn_supported_p): New override
4686 that takes two types as argument.
4687 * internal-fn.c (while_direct): New macro.
4688 (expand_while_optab_fn): New function.
4689 (convert_optab_supported_p): Likewise.
4690 (direct_while_optab_supported_p): New macro.
4691 * wide-int.h (wi::udiv_ceil): New function.
4692 * tree-vectorizer.h (rgroup_masks): New structure.
4693 (vec_loop_masks): New typedef.
4694 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4696 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4697 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4698 (vect_max_vf): New function.
4699 (slpeel_make_loop_iterate_ntimes): Delete.
4700 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4701 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4702 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4703 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4704 internal-fn.h, stor-layout.h and optabs-query.h.
4705 (vect_set_loop_mask): New function.
4706 (add_preheader_seq): Likewise.
4707 (add_header_seq): Likewise.
4708 (interleave_supported_p): Likewise.
4709 (vect_maybe_permute_loop_masks): Likewise.
4710 (vect_set_loop_masks_directly): Likewise.
4711 (vect_set_loop_condition_masked): Likewise.
4712 (vect_set_loop_condition_unmasked): New function, split out from
4713 slpeel_make_loop_iterate_ntimes.
4714 (slpeel_make_loop_iterate_ntimes): Rename to..
4715 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4716 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4717 (vect_do_peeling): Update call accordingly.
4718 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4720 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4721 mask_compare_type, can_fully_mask_p and fully_masked_p.
4722 (release_vec_loop_masks): New function.
4723 (_loop_vec_info): Use it to free the loop masks.
4724 (can_produce_all_loop_masks_p): New function.
4725 (vect_get_max_nscalars_per_iter): Likewise.
4726 (vect_verify_full_masking): Likewise.
4727 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4728 retries, and free the mask rgroups before retrying. Check loop-wide
4729 reasons for disallowing fully-masked loops. Make the final decision
4730 about whether use a fully-masked loop or not.
4731 (vect_estimate_min_profitable_iters): Do not assume that peeling
4732 for the number of iterations will be needed for fully-masked loops.
4733 (vectorizable_reduction): Disable fully-masked loops.
4734 (vectorizable_live_operation): Likewise.
4735 (vect_halve_mask_nunits): New function.
4736 (vect_double_mask_nunits): Likewise.
4737 (vect_record_loop_mask): Likewise.
4738 (vect_get_loop_mask): Likewise.
4739 (vect_transform_loop): Handle the case in which the final loop
4740 iteration might handle a partial vector. Call vect_set_loop_condition
4741 instead of slpeel_make_loop_iterate_ntimes.
4742 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4743 (check_load_store_masking): New function.
4744 (prepare_load_store_mask): Likewise.
4745 (vectorizable_store): Handle fully-masked loops.
4746 (vectorizable_load): Likewise.
4747 (supportable_widening_operation): Use vect_halve_mask_nunits for
4749 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4750 (vect_gen_while): New function.
4751 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4752 (aarch64_uqdec<mode>): New insn.
4754 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4755 Alan Hayward <alan.hayward@arm.com>
4756 David Sherwood <david.sherwood@arm.com>
4758 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4759 (reduc_xor_scal_optab): New optabs.
4760 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4761 (reduc_xor_scal_@var{m}): Document.
4762 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4763 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4765 * fold-const-call.c (fold_const_call): Handle them.
4766 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4767 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4768 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4769 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4770 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4771 (UNSPEC_XORV): New unspecs.
4772 (optab): Add entries for them.
4773 (BITWISEV): New int iterator.
4774 (bit_reduc_op): New int attributes.
4776 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4777 Alan Hayward <alan.hayward@arm.com>
4778 David Sherwood <david.sherwood@arm.com>
4780 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4781 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4782 * optabs.def (vec_shl_insert_optab): New optab.
4783 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4784 (duplicate_and_interleave): Likewise.
4785 * tree-vect-loop.c: Include internal-fn.h.
4786 (neutral_op_for_slp_reduction): New function, split out from
4787 get_initial_defs_for_reduction.
4788 (get_initial_def_for_reduction): Handle option 2 for variable-length
4789 vectors by loading the neutral value into a vector and then shifting
4790 the initial value into element 0.
4791 (get_initial_defs_for_reduction): Replace the code argument with
4792 the neutral value calculated by neutral_op_for_slp_reduction.
4793 Use gimple_build_vector for constant-length vectors.
4794 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4795 but the first group_size elements have a neutral value.
4796 Use duplicate_and_interleave otherwise.
4797 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4798 Update call to get_initial_defs_for_reduction. Handle SLP
4799 reductions for variable-length vectors by creating one vector
4800 result for each scalar result, with the elements associated
4801 with other scalar results stubbed out with the neutral value.
4802 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4803 Require IFN_VEC_SHL_INSERT for double reductions on
4804 variable-length vectors, or SLP reductions that have
4805 a neutral value. Require can_duplicate_and_interleave_p
4806 support for variable-length unchained SLP reductions if there
4807 is no neutral value, such as for MIN/MAX reductions. Also require
4808 the number of vector elements to be a multiple of the number of
4809 SLP statements when doing variable-length unchained SLP reductions.
4810 Update call to vect_create_epilog_for_reduction.
4811 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4812 and remove initial values.
4813 (duplicate_and_interleave): Make public.
4814 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4815 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4817 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4818 Alan Hayward <alan.hayward@arm.com>
4819 David Sherwood <david.sherwood@arm.com>
4821 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4822 (can_duplicate_and_interleave_p): New function.
4823 (vect_get_and_check_slp_defs): Take the vector of statements
4824 rather than just the current one. Remove excess parentheses.
4825 Restriction rejectinon of vect_constant_def and vect_external_def
4826 for variable-length vectors to boolean types, or types for which
4827 can_duplicate_and_interleave_p is false.
4828 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4829 (duplicate_and_interleave): New function.
4830 (vect_get_constant_vectors): Use gimple_build_vector for
4831 constant-length vectors and suitable variable-length constant
4832 vectors. Use duplicate_and_interleave for other variable-length
4833 vectors. Don't defer the update when inserting new statements.
4835 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4836 Alan Hayward <alan.hayward@arm.com>
4837 David Sherwood <david.sherwood@arm.com>
4839 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4840 min_profitable_iters doesn't go negative.
4842 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4843 Alan Hayward <alan.hayward@arm.com>
4844 David Sherwood <david.sherwood@arm.com>
4846 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4847 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4848 * optabs.def (vec_mask_load_lanes_optab): New optab.
4849 (vec_mask_store_lanes_optab): Likewise.
4850 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4851 (MASK_STORE_LANES): Likewise.
4852 * internal-fn.c (mask_load_lanes_direct): New macro.
4853 (mask_store_lanes_direct): Likewise.
4854 (expand_mask_load_optab_fn): Handle masked operations.
4855 (expand_mask_load_lanes_optab_fn): New macro.
4856 (expand_mask_store_optab_fn): Handle masked operations.
4857 (expand_mask_store_lanes_optab_fn): New macro.
4858 (direct_mask_load_lanes_optab_supported_p): Likewise.
4859 (direct_mask_store_lanes_optab_supported_p): Likewise.
4860 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4862 (vect_load_lanes_supported): Likewise.
4863 * tree-vect-data-refs.c (strip_conversion): New function.
4864 (can_group_stmts_p): Likewise.
4865 (vect_analyze_data_ref_accesses): Use it instead of checking
4866 for a pair of assignments.
4867 (vect_store_lanes_supported): Take a masked_p parameter.
4868 (vect_load_lanes_supported): Likewise.
4869 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4870 vect_store_lanes_supported and vect_load_lanes_supported.
4871 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4872 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4873 parameter. Don't allow gaps for masked accesses.
4874 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4875 and vect_load_lanes_supported.
4876 (get_load_store_type): Take a masked_p parameter and update
4877 call to get_group_load_store_type.
4878 (vectorizable_store): Update call to get_load_store_type.
4879 Handle IFN_MASK_STORE_LANES.
4880 (vectorizable_load): Update call to get_load_store_type.
4881 Handle IFN_MASK_LOAD_LANES.
4883 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4884 Alan Hayward <alan.hayward@arm.com>
4885 David Sherwood <david.sherwood@arm.com>
4887 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4889 * config/aarch64/aarch64-protos.h
4890 (aarch64_sve_struct_memory_operand_p): Declare.
4891 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4892 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4893 (VPRED, vpred): Handle SVE structure modes.
4894 * config/aarch64/constraints.md (Utx): New constraint.
4895 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4896 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4897 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4898 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4899 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4900 structure modes. Split into pieces after RA.
4901 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4902 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4904 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4905 SVE structure modes.
4906 (aarch64_classify_address): Likewise.
4907 (sizetochar): Move earlier in file.
4908 (aarch64_print_operand): Handle SVE register lists.
4909 (aarch64_array_mode): New function.
4910 (aarch64_sve_struct_memory_operand_p): Likewise.
4911 (TARGET_ARRAY_MODE): Redefine.
4913 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4914 Alan Hayward <alan.hayward@arm.com>
4915 David Sherwood <david.sherwood@arm.com>
4917 * target.def (array_mode): New target hook.
4918 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4919 * doc/tm.texi: Regenerate.
4920 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4921 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4922 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4924 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4927 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4928 Alan Hayward <alan.hayward@arm.com>
4929 David Sherwood <david.sherwood@arm.com>
4931 * fold-const.c (fold_binary_loc): Check the argument types
4932 rather than the result type when testing for a vector operation.
4934 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4936 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4937 * doc/tm.texi: Regenerate.
4939 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4940 Alan Hayward <alan.hayward@arm.com>
4941 David Sherwood <david.sherwood@arm.com>
4943 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4944 (sve): Document new AArch64 extension.
4945 * doc/md.texi (w): Extend the description of the AArch64
4946 constraint to include SVE vectors.
4947 (Upl, Upa): Document new AArch64 predicate constraints.
4948 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4950 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4951 (msve-vector-bits=): New option.
4952 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4953 SVE when these are disabled.
4954 (sve): New extension.
4955 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4956 modes. Adjust their number of units based on aarch64_sve_vg.
4957 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4958 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4959 aarch64_addr_query_type.
4960 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4961 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4962 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4963 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4964 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4965 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4966 (aarch64_simd_imm_zero_p): Delete.
4967 (aarch64_check_zero_based_sve_index_immediate): Declare.
4968 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4969 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4970 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4971 (aarch64_sve_float_mul_immediate_p): Likewise.
4972 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4974 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4975 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4976 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4977 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4978 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4979 (aarch64_regmode_natural_size): Likewise.
4980 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4981 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4983 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4984 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4985 for VG and the SVE predicate registers.
4986 (V_ALIASES): Add a "z"-prefixed alias.
4987 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4988 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4989 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4990 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4991 (REG_CLASS_NAMES): Add entries for them.
4992 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4993 and the predicate registers.
4994 (aarch64_sve_vg): Declare.
4995 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4996 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4997 (REGMODE_NATURAL_SIZE): Define.
4998 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
5000 * config/aarch64/aarch64.c: Include cfgrtl.h.
5001 (simd_immediate_info): Add a constructor for series vectors,
5002 and an associated step field.
5003 (aarch64_sve_vg): New variable.
5004 (aarch64_dbx_register_number): Handle VG and the predicate registers.
5005 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5006 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5007 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5008 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5009 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5010 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5011 (aarch64_get_mask_mode): New functions.
5012 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5013 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5014 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5015 predicate modes and predicate registers. Explicitly restrict
5016 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5017 to store a vector mode if it is recognized by
5018 aarch64_classify_vector_mode.
5019 (aarch64_regmode_natural_size): New function.
5020 (aarch64_hard_regno_caller_save_mode): Return the original mode
5022 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5023 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5024 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5025 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5027 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5028 does not overlap dest if the function is frame-related. Handle
5030 (aarch64_split_add_offset): New function.
5031 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5032 them aarch64_add_offset.
5033 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5034 and update call to aarch64_sub_sp.
5035 (aarch64_add_cfa_expression): New function.
5036 (aarch64_expand_prologue): Pass extra temporary registers to the
5037 functions above. Handle the case in which we need to emit new
5038 DW_CFA_expressions for registers that were originally saved
5039 relative to the stack pointer, but now have to be expressed
5040 relative to the frame pointer.
5041 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5043 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5044 IP0 and IP1 values for SVE frames.
5045 (aarch64_expand_vec_series): New function.
5046 (aarch64_expand_sve_widened_duplicate): Likewise.
5047 (aarch64_expand_sve_const_vector): Likewise.
5048 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5049 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5050 into the register, rather than emitting a SET directly.
5051 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5052 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5053 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5054 (offset_9bit_signed_scaled_p): New functions.
5055 (aarch64_replicate_bitmask_imm): New function.
5056 (aarch64_bitmask_imm): Use it.
5057 (aarch64_cannot_force_const_mem): Reject expressions involving
5058 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5059 (aarch64_classify_index): Handle SVE indices, by requiring
5060 a plain register index with a scale that matches the element size.
5061 (aarch64_classify_address): Handle SVE addresses. Assert that
5062 the mode of the address is VOIDmode or an integer mode.
5063 Update call to aarch64_classify_symbol.
5064 (aarch64_classify_symbolic_expression): Update call to
5065 aarch64_classify_symbol.
5066 (aarch64_const_vec_all_in_range_p): New function.
5067 (aarch64_print_vector_float_operand): Likewise.
5068 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5069 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5070 and the FP immediates 1.0 and 0.5.
5071 (aarch64_print_address_internal): Handle SVE addresses.
5072 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5073 (aarch64_regno_regclass): Handle predicate registers.
5074 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5076 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5077 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5078 (aarch64_convert_sve_vector_bits): New function.
5079 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5080 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5082 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5083 Handle SVE vector and predicate modes. Accept VL-based constants
5084 that need only one temporary register, and VL offsets that require
5085 no temporary registers.
5086 (aarch64_conditional_register_usage): Mark the predicate registers
5087 as fixed if SVE isn't available.
5088 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5089 Return true for SVE vector and predicate modes.
5090 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5091 rather than an unsigned int. Handle SVE modes.
5092 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5094 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5096 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5097 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5098 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5099 (aarch64_sve_float_mul_immediate_p): New functions.
5100 (aarch64_sve_valid_immediate): New function.
5101 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5102 Explicitly reject structure modes. Check for INDEX constants.
5103 Handle PTRUE and PFALSE constants.
5104 (aarch64_check_zero_based_sve_index_immediate): New function.
5105 (aarch64_simd_imm_zero_p): Delete.
5106 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5107 vector modes. Accept constants in the range of CNT[BHWD].
5108 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5109 ask for an Advanced SIMD mode.
5110 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5111 (aarch64_simd_vector_alignment): Handle SVE predicates.
5112 (aarch64_vectorize_preferred_vector_alignment): New function.
5113 (aarch64_simd_vector_alignment_reachable): Use it instead of
5115 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5116 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5118 (MAX_VECT_LEN): Delete.
5119 (expand_vec_perm_d): Add a vec_flags field.
5120 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5121 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5122 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5124 (aarch64_evpc_rev): Rename to...
5125 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5126 (aarch64_evpc_rev_global): New function.
5127 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5128 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5130 (aarch64_evpc_sve_tbl): New function.
5131 (aarch64_expand_vec_perm_const_1): Update after rename of
5132 aarch64_evpc_rev. Handle SVE permutes too, trying
5133 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5134 than aarch64_evpc_tbl.
5135 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5136 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5137 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5138 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5139 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5140 (aarch64_expand_sve_vcond): New functions.
5141 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5142 of aarch64_vector_mode_p.
5143 (aarch64_dwarf_poly_indeterminate_value): New function.
5144 (aarch64_compute_pressure_classes): Likewise.
5145 (aarch64_can_change_mode_class): Likewise.
5146 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5147 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5148 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5149 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5150 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5151 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5152 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5153 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5155 (Dn, Dl, Dr): Accept const as well as const_vector.
5156 (Dz): Likewise. Compare against CONST0_RTX.
5157 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5158 of "vector" where appropriate.
5159 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5160 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5161 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5162 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5163 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5164 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5165 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5166 (v_int_equiv): Extend to SVE modes.
5167 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5169 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5170 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5171 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5172 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5173 (SVE_COND_FP_CMP): New int iterators.
5174 (perm_hilo): Handle the new unpack unspecs.
5175 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5177 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5178 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5179 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5180 (aarch64_equality_operator, aarch64_constant_vector_operand)
5181 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5182 (aarch64_sve_nonimmediate_operand): Likewise.
5183 (aarch64_sve_general_operand): Likewise.
5184 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5185 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5186 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5187 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5188 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5189 (aarch64_sve_float_arith_immediate): Likewise.
5190 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5191 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5192 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5193 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5194 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5195 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5196 (aarch64_sve_float_arith_operand): Likewise.
5197 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5198 (aarch64_sve_float_mul_operand): Likewise.
5199 (aarch64_sve_vec_perm_operand): Likewise.
5200 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5201 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5202 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5203 as well as const_vector.
5204 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5205 in file. Use CONST0_RTX and CONSTM1_RTX.
5206 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5207 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5208 Use aarch64_simd_imm_zero.
5209 * config/aarch64/aarch64-sve.md: New file.
5210 * config/aarch64/aarch64.md: Include it.
5211 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5212 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5213 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5214 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5215 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5216 (sve): New attribute.
5217 (enabled): Disable instructions with the sve attribute unless
5219 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5220 aarch64_expand_mov_immediate.
5221 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5222 CNT[BHSD] immediates.
5223 (movti): Split CONST_POLY_INT moves into two halves.
5224 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5225 Split additions that need a temporary here if the destination
5226 is the stack pointer.
5227 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5228 (*add<mode>3_poly_1): New instruction.
5229 (set_clobber_cc): New expander.
5231 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5233 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5234 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5235 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5236 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5237 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5238 Change innermode from fixed_mode_size to machine_mode.
5239 (simplify_subreg): Update call accordingly. Handle a constant-sized
5240 subreg of a variable-length CONST_VECTOR.
5242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5243 Alan Hayward <alan.hayward@arm.com>
5244 David Sherwood <david.sherwood@arm.com>
5246 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5247 (add_offset_to_base): New function, split out from...
5248 (create_mem_ref): ...here. When handling a scale other than 1,
5249 check first whether the address is valid without the offset.
5250 Add it into the base if so, leaving the index and scale as-is.
5252 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5255 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5256 fold_for_warn before checking if arg2 is INTEGER_CST.
5258 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5260 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5261 (store_multiple_operation): Delete.
5262 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5263 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5264 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5265 guarded by TARGET_STRING.
5266 (rs6000_output_load_multiple): Delete.
5267 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5268 OPTION_MASK_STRING / TARGET_STRING handling.
5269 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5270 (const rs6000_opt_masks) <"string">: Change mask to 0.
5271 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5272 (MASK_STRING): Delete.
5273 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5275 (load_multiple): Delete.
5282 (store_multiple): Delete.
5289 (movmemsi_8reg): Delete.
5290 (corresponding unnamed define_insn): Delete.
5291 (movmemsi_6reg): Delete.
5292 (corresponding unnamed define_insn): Delete.
5293 (movmemsi_4reg): Delete.
5294 (corresponding unnamed define_insn): Delete.
5295 (movmemsi_2reg): Delete.
5296 (corresponding unnamed define_insn): Delete.
5297 (movmemsi_1reg): Delete.
5298 (corresponding unnamed define_insn): Delete.
5299 * config/rs6000/rs6000.opt (mno-string): New.
5300 (mstring): Replace by deprecation warning stub.
5301 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5303 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5305 * regrename.c (regrename_do_replace): If replacing the same
5306 reg multiple times, try to reuse last created gen_raw_REG.
5309 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5310 main to workaround a bug in GDB.
5312 2018-01-12 Tom de Vries <tom@codesourcery.com>
5315 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5317 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5319 PR rtl-optimization/80481
5320 * ira-color.c (get_cap_member): New function.
5321 (allocnos_conflict_by_live_ranges_p): Use it.
5322 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5323 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5325 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5328 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5329 (*saddl_se_1): Ditto.
5331 (*ssubl_se_1): Ditto.
5333 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5335 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5336 rather than wi::to_widest for DR_INITs.
5337 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5338 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5339 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5341 (vect_analyze_group_access_1): Note that here.
5343 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5345 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5346 polynomial type sizes.
5348 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5350 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5351 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5352 (gimple_add_tmp_var): Likewise.
5354 2018-01-12 Martin Liska <mliska@suse.cz>
5356 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5357 (gimple_alloc_sizes): Likewise.
5358 (dump_gimple_statistics): Use PRIu64 in printf format.
5359 * gimple.h: Change uint64_t to int.
5361 2018-01-12 Martin Liska <mliska@suse.cz>
5363 * tree-core.h: Use uint64_t instead of int.
5364 * tree.c (tree_node_counts): Likewise.
5365 (tree_node_sizes): Likewise.
5366 (dump_tree_statistics): Use PRIu64 in printf format.
5368 2018-01-12 Martin Liska <mliska@suse.cz>
5370 * Makefile.in: As qsort_chk is implemented in vec.c, add
5371 vec.o to linkage of gencfn-macros.
5372 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5373 passing the info to record_node_allocation_statistics.
5374 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5376 * ggc-common.c (struct ggc_usage): Add operator== and use
5377 it in operator< and compare function.
5378 * mem-stats.h (struct mem_usage): Likewise.
5379 * vec.c (struct vec_usage): Remove operator< and compare
5380 function. Can be simply inherited.
5382 2018-01-12 Martin Jambor <mjambor@suse.cz>
5385 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5386 * tree-ssa-math-opts.c: Include domwalk.h.
5387 (convert_mult_to_fma_1): New function.
5388 (fma_transformation_info): New type.
5389 (fma_deferring_state): Likewise.
5390 (cancel_fma_deferring): New function.
5391 (result_of_phi): Likewise.
5392 (last_fma_candidate_feeds_initial_phi): Likewise.
5393 (convert_mult_to_fma): Added deferring logic, split actual
5394 transformation to convert_mult_to_fma_1.
5395 (math_opts_dom_walker): New type.
5396 (math_opts_dom_walker::after_dom_children): New method, body moved
5397 here from pass_optimize_widening_mul::execute, added deferring logic
5399 (pass_optimize_widening_mul::execute): Moved most of code to
5400 math_opts_dom_walker::after_dom_children.
5401 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5402 * config/i386/i386.c (ix86_option_override_internal): Added
5403 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5405 2018-01-12 Richard Biener <rguenther@suse.de>
5408 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5409 inline instance vars.
5411 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5414 * config/rx/rx.c (rx_is_restricted_memory_address):
5417 2018-01-12 Richard Biener <rguenther@suse.de>
5419 PR tree-optimization/80846
5420 * target.def (split_reduction): New target hook.
5421 * targhooks.c (default_split_reduction): New function.
5422 * targhooks.h (default_split_reduction): Declare.
5423 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5424 target requests first reduce vectors by combining low and high
5426 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5427 (get_vectype_for_scalar_type_and_size): Export.
5428 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5429 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5430 * doc/tm.texi: Regenerate.
5431 * config/i386/i386.c (ix86_split_reduction): Implement
5432 TARGET_VECTORIZE_SPLIT_REDUCTION.
5434 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5437 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5438 in PIC mode except for TARGET_VXWORKS_RTP.
5439 * config/sparc/sparc.c: Include cfgrtl.h.
5440 (TARGET_INIT_PIC_REG): Define.
5441 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5442 (sparc_pic_register_p): New predicate.
5443 (sparc_legitimate_address_p): Use it.
5444 (sparc_legitimize_pic_address): Likewise.
5445 (sparc_delegitimize_address): Likewise.
5446 (sparc_mode_dependent_address_p): Likewise.
5447 (gen_load_pcrel_sym): Remove 4th parameter.
5448 (load_got_register): Adjust call to above. Remove obsolete stuff.
5449 (sparc_expand_prologue): Do not call load_got_register here.
5450 (sparc_flat_expand_prologue): Likewise.
5451 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5452 (sparc_use_pseudo_pic_reg): New function.
5453 (sparc_init_pic_reg): Likewise.
5454 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5455 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5457 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5459 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5460 Add item for branch_cost.
5462 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5464 PR rtl-optimization/83565
5465 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5466 not extend the result to a larger mode for rotate operations.
5467 (num_sign_bit_copies1): Likewise.
5469 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5472 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5474 Use values-Xc.o for -pedantic.
5475 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5477 2018-01-12 Martin Liska <mliska@suse.cz>
5480 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5482 (possible_polymorphic_call_targets): Use it.
5483 (ipa_devirt): Likewise.
5485 2018-01-12 Martin Liska <mliska@suse.cz>
5487 * profile-count.h (enum profile_quality): Use 0 as invalid
5488 enum value of profile_quality.
5490 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5492 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5493 -mext-string options.
5495 2018-01-12 Richard Biener <rguenther@suse.de>
5497 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5498 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5499 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5501 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5503 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5505 * configure.ac (--with-long-double-format): Add support for the
5506 configuration option to change the default long double format on
5508 * config.gcc (powerpc*-linux*-*): Likewise.
5509 * configure: Regenerate.
5510 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5511 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5512 used without modification.
5514 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5516 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5517 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5518 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5519 MISC_BUILTIN_SPEC_BARRIER.
5520 (rs6000_init_builtins): Likewise.
5521 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5523 (speculation_barrier): New define_insn.
5524 * doc/extend.texi: Document __builtin_speculation_barrier.
5526 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5529 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5530 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5531 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5533 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5534 integral modes instead of "ss" and "sd".
5535 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5536 vectors with 32-bit and 64-bit elements.
5537 (vecdupssescalarmodesuffix): New mode attribute.
5538 (vec_dup<mode>): Use it.
5540 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5543 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5544 frame if argument is passed on stack.
5546 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5549 * ree.c (combine_reaching_defs): Optimize also
5550 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5551 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5553 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5556 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5558 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5561 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5562 after they are computed.
5564 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5566 PR tree-optimization/83695
5567 * gimple-loop-linterchange.cc
5568 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5569 reset cached scev information after interchange.
5570 (pass_linterchange::execute): Remove call to scev_reset_htab.
5572 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5574 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5575 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5576 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5577 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5578 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5579 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5580 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5581 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5582 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5583 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5584 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5585 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5586 (V_lane_reg): Likewise.
5587 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5589 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5590 (vfmal_lane_low<mode>_intrinsic,
5591 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5592 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5593 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5594 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5595 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5596 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5598 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5600 * config/arm/arm-cpus.in (fp16fml): New feature.
5601 (ALL_SIMD): Add fp16fml.
5602 (armv8.2-a): Add fp16fml as an option.
5603 (armv8.3-a): Likewise.
5604 (armv8.4-a): Add fp16fml as part of fp16.
5605 * config/arm/arm.h (TARGET_FP16FML): Define.
5606 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5608 * config/arm/arm-modes.def (V2HF): Define.
5609 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5610 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5611 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5612 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5613 vfmsl_low, vfmsl_high): New set of builtins.
5614 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5615 (vfml_op): New code attribute.
5616 (VFMLHALVES): New int iterator.
5617 (VFML, VFMLSEL): New mode attributes.
5618 (V_reg): Define mapping for V2HF.
5619 (V_hi, V_lo): New mode attributes.
5620 (VF_constraint): Likewise.
5621 (vfml_half, vfml_half_selector): New int attributes.
5622 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5624 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5625 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5627 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5628 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5629 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5630 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5632 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5633 Document new effective target and option set.
5635 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5637 * config/arm/arm-cpus.in (armv8_4): New feature.
5638 (ARMv8_4a): New fgroup.
5639 (armv8.4-a): New arch.
5640 * config/arm/arm-tables.opt: Regenerate.
5641 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5642 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5643 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5644 Add matching rules for -march=armv8.4-a and extensions.
5645 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5647 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5650 * config/rx/rx.md (BW): New mode attribute.
5651 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5653 2018-01-11 Richard Biener <rguenther@suse.de>
5655 PR tree-optimization/83435
5656 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5657 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5658 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5660 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5661 Alan Hayward <alan.hayward@arm.com>
5662 David Sherwood <david.sherwood@arm.com>
5664 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5666 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5667 (aarch64_print_address_internal): Use it to check for a zero offset.
5669 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5670 Alan Hayward <alan.hayward@arm.com>
5671 David Sherwood <david.sherwood@arm.com>
5673 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5674 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5675 Return a poly_int64 rather than a HOST_WIDE_INT.
5676 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5677 rather than a HOST_WIDE_INT.
5678 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5679 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5680 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5681 final_offset from HOST_WIDE_INT to poly_int64.
5682 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5683 to_constant when getting the number of units in an Advanced SIMD
5685 (aarch64_builtin_vectorized_function): Check for a constant number
5687 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5689 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5690 attribute instead of GET_MODE_NUNITS.
5691 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5692 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5693 GET_MODE_SIZE for fixed-size registers.
5694 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5695 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5696 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5697 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5698 (aarch64_print_operand, aarch64_print_address_internal)
5699 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5700 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5701 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5702 Handle polynomial GET_MODE_SIZE.
5703 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5704 wider than SImode without modification.
5705 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5706 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5707 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5708 passing and returning SVE modes.
5709 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5710 rather than GEN_INT.
5711 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5712 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5713 (aarch64_allocate_and_probe_stack_space): Likewise.
5714 (aarch64_layout_frame): Cope with polynomial offsets.
5715 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5716 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5718 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5719 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5720 poly_int64 rather than a HOST_WIDE_INT.
5721 (aarch64_get_separate_components, aarch64_process_components)
5722 (aarch64_expand_prologue, aarch64_expand_epilogue)
5723 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5724 (aarch64_anchor_offset): New function, split out from...
5725 (aarch64_legitimize_address): ...here.
5726 (aarch64_builtin_vectorization_cost): Handle polynomial
5727 TYPE_VECTOR_SUBPARTS.
5728 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5730 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5731 number of elements from the PARALLEL rather than the mode.
5732 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5733 rather than GET_MODE_BITSIZE.
5734 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5735 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5736 (aarch64_expand_vec_perm_const_1): Handle polynomial
5737 d->perm.length () and d->perm elements.
5738 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5739 Apply to_constant to d->perm elements.
5740 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5741 polynomial CONST_VECTOR_NUNITS.
5742 (aarch64_move_pointer): Take amount as a poly_int64 rather
5744 (aarch64_progress_pointer): Avoid temporary variable.
5745 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5746 the mode attribute instead of GET_MODE.
5748 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5749 Alan Hayward <alan.hayward@arm.com>
5750 David Sherwood <david.sherwood@arm.com>
5752 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5753 x exists before using it.
5754 (aarch64_add_constant_internal): Rename to...
5755 (aarch64_add_offset_1): ...this. Replace regnum with separate
5756 src and dest rtxes. Handle the case in which they're different,
5757 including when the offset is zero. Replace scratchreg with an rtx.
5758 Use 2 additions if there is no spare register into which we can
5759 move a 16-bit constant.
5760 (aarch64_add_constant): Delete.
5761 (aarch64_add_offset): Replace reg with separate src and dest
5762 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5763 Use aarch64_add_offset_1.
5764 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5765 an rtx rather than an int. Take the delta as a poly_int64
5766 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5767 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5768 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5769 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5770 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5772 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5773 aarch64_add_constant.
5775 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5777 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5778 Use scalar_float_mode.
5780 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5782 * config/aarch64/aarch64-simd.md
5783 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5784 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5785 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5786 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5787 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5788 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5789 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5790 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5791 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5792 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5794 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5797 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5798 targ_options->x_arm_arch_string is non NULL.
5800 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5802 * config/aarch64/aarch64.h
5803 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5805 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5808 * expmed.c (emit_store_flag_force): Swap if const op0
5809 and change VOIDmode to mode of op0.
5811 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5813 PR rtl-optimization/83761
5814 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5815 than bytes to mode_for_size.
5817 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5820 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5821 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5824 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5827 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5828 when in layout mode.
5829 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5830 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5833 2018-01-10 Michael Collison <michael.collison@arm.com>
5835 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5836 * config/aarch64/aarch64-option-extension.def: Add
5837 AARCH64_OPT_EXTENSION of 'fp16fml'.
5838 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5839 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5840 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5841 * config/aarch64/constraints.md (Ui7): New constraint.
5842 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5843 (VFMLA_SEL_W): Ditto.
5846 (VFMLA16_LOW): New int iterator.
5847 (VFMLA16_HIGH): Ditto.
5848 (UNSPEC_FMLAL): New unspec.
5849 (UNSPEC_FMLSL): Ditto.
5850 (UNSPEC_FMLAL2): Ditto.
5851 (UNSPEC_FMLSL2): Ditto.
5852 (f16mac): New code attribute.
5853 * config/aarch64/aarch64-simd-builtins.def
5854 (aarch64_fmlal_lowv2sf): Ditto.
5855 (aarch64_fmlsl_lowv2sf): Ditto.
5856 (aarch64_fmlalq_lowv4sf): Ditto.
5857 (aarch64_fmlslq_lowv4sf): Ditto.
5858 (aarch64_fmlal_highv2sf): Ditto.
5859 (aarch64_fmlsl_highv2sf): Ditto.
5860 (aarch64_fmlalq_highv4sf): Ditto.
5861 (aarch64_fmlslq_highv4sf): Ditto.
5862 (aarch64_fmlal_lane_lowv2sf): Ditto.
5863 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5864 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5865 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5866 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5867 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5868 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5869 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5870 (aarch64_fmlal_lane_highv2sf): Ditto.
5871 (aarch64_fmlsl_lane_highv2sf): Ditto.
5872 (aarch64_fmlal_laneq_highv2sf): Ditto.
5873 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5874 (aarch64_fmlalq_lane_highv4sf): Ditto.
5875 (aarch64_fmlsl_lane_highv4sf): Ditto.
5876 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5877 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5878 * config/aarch64/aarch64-simd.md:
5879 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5880 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5881 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5882 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5883 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5884 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5885 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5886 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5887 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5888 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5889 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5890 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5891 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5892 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5893 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5894 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5895 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5896 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5897 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5898 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5899 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5900 (vfmlsl_low_u32): Ditto.
5901 (vfmlalq_low_u32): Ditto.
5902 (vfmlslq_low_u32): Ditto.
5903 (vfmlal_high_u32): Ditto.
5904 (vfmlsl_high_u32): Ditto.
5905 (vfmlalq_high_u32): Ditto.
5906 (vfmlslq_high_u32): Ditto.
5907 (vfmlal_lane_low_u32): Ditto.
5908 (vfmlsl_lane_low_u32): Ditto.
5909 (vfmlal_laneq_low_u32): Ditto.
5910 (vfmlsl_laneq_low_u32): Ditto.
5911 (vfmlalq_lane_low_u32): Ditto.
5912 (vfmlslq_lane_low_u32): Ditto.
5913 (vfmlalq_laneq_low_u32): Ditto.
5914 (vfmlslq_laneq_low_u32): Ditto.
5915 (vfmlal_lane_high_u32): Ditto.
5916 (vfmlsl_lane_high_u32): Ditto.
5917 (vfmlal_laneq_high_u32): Ditto.
5918 (vfmlsl_laneq_high_u32): Ditto.
5919 (vfmlalq_lane_high_u32): Ditto.
5920 (vfmlslq_lane_high_u32): Ditto.
5921 (vfmlalq_laneq_high_u32): Ditto.
5922 (vfmlslq_laneq_high_u32): Ditto.
5923 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5924 (AARCH64_FL_FOR_ARCH8_4): New.
5925 (AARCH64_ISA_F16FML): New ISA flag.
5926 (TARGET_F16FML): New feature flag for fp16fml.
5927 (doc/invoke.texi): Document new fp16fml option.
5929 2018-01-10 Michael Collison <michael.collison@arm.com>
5931 * config/aarch64/aarch64-builtins.c:
5932 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5933 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5934 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5935 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5936 (AARCH64_ISA_SHA3): New ISA flag.
5937 (TARGET_SHA3): New feature flag for sha3.
5938 * config/aarch64/iterators.md (sha512_op): New int attribute.
5939 (CRYPTO_SHA512): New int iterator.
5940 (UNSPEC_SHA512H): New unspec.
5941 (UNSPEC_SHA512H2): Ditto.
5942 (UNSPEC_SHA512SU0): Ditto.
5943 (UNSPEC_SHA512SU1): Ditto.
5944 * config/aarch64/aarch64-simd-builtins.def
5945 (aarch64_crypto_sha512hqv2di): New builtin.
5946 (aarch64_crypto_sha512h2qv2di): Ditto.
5947 (aarch64_crypto_sha512su0qv2di): Ditto.
5948 (aarch64_crypto_sha512su1qv2di): Ditto.
5949 (aarch64_eor3qv8hi): Ditto.
5950 (aarch64_rax1qv2di): Ditto.
5951 (aarch64_xarqv2di): Ditto.
5952 (aarch64_bcaxqv8hi): Ditto.
5953 * config/aarch64/aarch64-simd.md:
5954 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5955 (aarch64_crypto_sha512su0qv2di): Ditto.
5956 (aarch64_crypto_sha512su1qv2di): Ditto.
5957 (aarch64_eor3qv8hi): Ditto.
5958 (aarch64_rax1qv2di): Ditto.
5959 (aarch64_xarqv2di): Ditto.
5960 (aarch64_bcaxqv8hi): Ditto.
5961 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5962 (vsha512h2q_u64): Ditto.
5963 (vsha512su0q_u64): Ditto.
5964 (vsha512su1q_u64): Ditto.
5965 (veor3q_u16): Ditto.
5966 (vrax1q_u64): Ditto.
5968 (vbcaxq_u16): Ditto.
5969 * config/arm/types.md (crypto_sha512): New type attribute.
5970 (crypto_sha3): Ditto.
5971 (doc/invoke.texi): Document new sha3 option.
5973 2018-01-10 Michael Collison <michael.collison@arm.com>
5975 * config/aarch64/aarch64-builtins.c:
5976 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5977 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5978 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5979 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5980 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5981 (AARCH64_ISA_SM4): New ISA flag.
5982 (TARGET_SM4): New feature flag for sm4.
5983 * config/aarch64/aarch64-simd-builtins.def
5984 (aarch64_sm3ss1qv4si): Ditto.
5985 (aarch64_sm3tt1aq4si): Ditto.
5986 (aarch64_sm3tt1bq4si): Ditto.
5987 (aarch64_sm3tt2aq4si): Ditto.
5988 (aarch64_sm3tt2bq4si): Ditto.
5989 (aarch64_sm3partw1qv4si): Ditto.
5990 (aarch64_sm3partw2qv4si): Ditto.
5991 (aarch64_sm4eqv4si): Ditto.
5992 (aarch64_sm4ekeyqv4si): Ditto.
5993 * config/aarch64/aarch64-simd.md:
5994 (aarch64_sm3ss1qv4si): Ditto.
5995 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5996 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5997 (aarch64_sm4eqv4si): Ditto.
5998 (aarch64_sm4ekeyqv4si): Ditto.
5999 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
6000 (sm3part_op): Ditto.
6001 (CRYPTO_SM3TT): Ditto.
6002 (CRYPTO_SM3PART): Ditto.
6003 (UNSPEC_SM3SS1): New unspec.
6004 (UNSPEC_SM3TT1A): Ditto.
6005 (UNSPEC_SM3TT1B): Ditto.
6006 (UNSPEC_SM3TT2A): Ditto.
6007 (UNSPEC_SM3TT2B): Ditto.
6008 (UNSPEC_SM3PARTW1): Ditto.
6009 (UNSPEC_SM3PARTW2): Ditto.
6010 (UNSPEC_SM4E): Ditto.
6011 (UNSPEC_SM4EKEY): Ditto.
6012 * config/aarch64/constraints.md (Ui2): New constraint.
6013 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6014 * config/arm/types.md (crypto_sm3): New type attribute.
6015 (crypto_sm4): Ditto.
6016 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6017 (vsm3tt1aq_u32): Ditto.
6018 (vsm3tt1bq_u32): Ditto.
6019 (vsm3tt2aq_u32): Ditto.
6020 (vsm3tt2bq_u32): Ditto.
6021 (vsm3partw1q_u32): Ditto.
6022 (vsm3partw2q_u32): Ditto.
6023 (vsm4eq_u32): Ditto.
6024 (vsm4ekeyq_u32): Ditto.
6025 (doc/invoke.texi): Document new sm4 option.
6027 2018-01-10 Michael Collison <michael.collison@arm.com>
6029 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6030 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6031 (AARCH64_FL_FOR_ARCH8_4): New.
6032 (AARCH64_FL_V8_4): New flag.
6033 (doc/invoke.texi): Document new armv8.4-a option.
6035 2018-01-10 Michael Collison <michael.collison@arm.com>
6037 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6038 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6039 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6040 * config/aarch64/aarch64-option-extension.def: Add
6041 AARCH64_OPT_EXTENSION of 'sha2'.
6042 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6043 (crypto): Disable sha2 and aes if crypto disabled.
6044 (crypto): Enable aes and sha2 if enabled.
6045 (simd): Disable sha2 and aes if simd disabled.
6046 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6048 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6049 (TARGET_SHA2): New feature flag for sha2.
6050 (TARGET_AES): New feature flag for aes.
6051 * config/aarch64/aarch64-simd.md:
6052 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6053 conditional on TARGET_AES.
6054 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6055 (aarch64_crypto_sha1hsi): Make pattern conditional
6057 (aarch64_crypto_sha1hv4si): Ditto.
6058 (aarch64_be_crypto_sha1hv4si): Ditto.
6059 (aarch64_crypto_sha1su1v4si): Ditto.
6060 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6061 (aarch64_crypto_sha1su0v4si): Ditto.
6062 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6063 (aarch64_crypto_sha256su0v4si): Ditto.
6064 (aarch64_crypto_sha256su1v4si): Ditto.
6065 (doc/invoke.texi): Document new aes and sha2 options.
6067 2018-01-10 Martin Sebor <msebor@redhat.com>
6069 PR tree-optimization/83781
6070 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6073 2018-01-11 Martin Sebor <msebor@gmail.com>
6074 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6076 PR tree-optimization/83501
6077 PR tree-optimization/81703
6079 * tree-ssa-strlen.c (get_string_cst): Rename...
6080 (get_string_len): ...to this. Handle global constants.
6081 (handle_char_store): Adjust.
6083 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6084 Jim Wilson <jimw@sifive.com>
6086 * config/riscv/riscv-protos.h (riscv_output_return): New.
6087 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6088 (riscv_attribute_table, riscv_output_return),
6089 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6090 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6091 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6093 (riscv_expand_prologue): Add early return for naked function.
6094 (riscv_expand_epilogue): Likewise.
6095 (riscv_function_ok_for_sibcall): Return false for naked function.
6096 (riscv_set_current_function): New.
6097 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6098 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6099 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6100 * doc/extend.texi (RISC-V Function Attributes): New.
6102 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6104 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6105 check for 128-bit long double before checking TCmode.
6106 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6107 128-bit long doubles before checking TFmode or TCmode.
6108 (FLOAT128_IBM_P): Likewise.
6110 2018-01-10 Martin Sebor <msebor@redhat.com>
6112 PR tree-optimization/83671
6113 * builtins.c (c_strlen): Unconditionally return zero for the empty
6115 Use -Warray-bounds for warnings.
6116 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6117 for non-constant array indices with COMPONENT_REF, arrays of
6118 arrays, and pointers to arrays.
6119 (gimple_fold_builtin_strlen): Determine and set length range for
6120 non-constant character arrays.
6122 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6125 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6128 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6130 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6132 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6135 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6136 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6137 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6138 indexed_or_indirect_operand predicate.
6139 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6140 (*vsx_le_perm_load_v8hi): Likewise.
6141 (*vsx_le_perm_load_v16qi): Likewise.
6142 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6143 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6144 (*vsx_le_perm_store_v8hi): Likewise.
6145 (*vsx_le_perm_store_v16qi): Likewise.
6146 (eight unnamed splitters): Likewise.
6148 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6150 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6151 * config/rs6000/emmintrin.h: Likewise.
6152 * config/rs6000/mmintrin.h: Likewise.
6153 * config/rs6000/xmmintrin.h: Likewise.
6155 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6158 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6160 * tree.c (tree_nop_conversion): Return true for location wrapper
6162 (maybe_wrap_with_location): New function.
6163 (selftest::check_strip_nops): New function.
6164 (selftest::test_location_wrappers): New function.
6165 (selftest::tree_c_tests): Call it.
6166 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6167 (maybe_wrap_with_location): New decl.
6168 (EXPR_LOCATION_WRAPPER_P): New macro.
6169 (location_wrapper_p): New inline function.
6170 (tree_strip_any_location_wrapper): New inline function.
6172 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6175 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6176 stack_realign_offset for the largest alignment of stack slot
6178 (ix86_find_max_used_stack_alignment): New function.
6179 (ix86_finalize_stack_frame_flags): Use it. Set
6180 max_used_stack_alignment if we don't realign stack.
6181 * config/i386/i386.h (machine_function): Add
6182 max_used_stack_alignment.
6184 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6186 * config/arm/arm.opt (-mbranch-cost): New option.
6187 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6190 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6193 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6194 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6196 2018-01-10 Richard Biener <rguenther@suse.de>
6199 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6200 early out so it also covers the case where we have a non-NULL
6203 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6205 PR tree-optimization/83753
6206 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6207 for non-strided grouped accesses if the number of elements is 1.
6209 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6212 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6213 * i386.h (TARGET_USE_GATHER): Define.
6214 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6216 2018-01-10 Martin Liska <mliska@suse.cz>
6219 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6220 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6222 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6223 CLEANUP_NO_PARTITIONING is not set.
6225 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6227 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6228 for vectors, as a partial revert of r254296.
6229 * rtl.h (const_vec_p): Delete.
6230 (const_vec_duplicate_p): Don't test for vector CONSTs.
6231 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6232 * expmed.c (make_tree): Likewise.
6235 * common.md (E, F): Use CONSTANT_P instead of checking for
6237 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6238 checking for CONST_VECTOR.
6240 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6243 * predict.c (force_edge_cold): Handle in more sane way edges
6246 2018-01-09 Carl Love <cel@us.ibm.com>
6248 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6250 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6251 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6252 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6253 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6254 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6255 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6256 * config/rs6000/rs6000-protos.h: Add extern defition for
6257 rs6000_generate_float2_double_code.
6258 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6260 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6261 (float2_v2df): Add define_expand.
6263 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6266 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6267 op_mode in the force_to_mode call.
6269 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6271 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6272 instead of checking each element individually.
6273 (aarch64_evpc_uzp): Likewise.
6274 (aarch64_evpc_zip): Likewise.
6275 (aarch64_evpc_ext): Likewise.
6276 (aarch64_evpc_rev): Likewise.
6277 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6278 instead of checking each element individually. Return true without
6280 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6281 whether all selected elements come from the same input, instead of
6282 checking each element individually. Remove calls to gen_rtx_REG,
6283 start_sequence and end_sequence and instead assert that no rtl is
6286 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6288 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6289 order of HIGH and CONST checks.
6291 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6293 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6294 if the destination isn't an SSA_NAME.
6296 2018-01-09 Richard Biener <rguenther@suse.de>
6298 PR tree-optimization/83668
6299 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6301 (canonicalize_loop_form): ... here, renamed from ...
6302 (canonicalize_loop_closed_ssa_form): ... this and amended to
6303 swap successor edges for loop exit blocks to make us use
6304 the RPO order we need for initial schedule generation.
6306 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6308 PR tree-optimization/64811
6309 * match.pd: When optimizing comparisons with Inf, avoid
6310 introducing or losing exceptions from comparisons with NaN.
6312 2018-01-09 Martin Liska <mliska@suse.cz>
6315 * asan.c (shadow_mem_size): Add gcc_assert.
6317 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6319 Don't save registers in main().
6322 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6323 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6324 * config/avr/avr.c (avr_set_current_function): Don't error if
6325 naked, OS_task or OS_main are specified at the same time.
6326 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6328 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6330 * common/config/avr/avr-common.c (avr_option_optimization_table):
6331 Switch on -mmain-is-OS_task for optimizing compilations.
6333 2018-01-09 Richard Biener <rguenther@suse.de>
6335 PR tree-optimization/83572
6336 * graphite.c: Include cfganal.h.
6337 (graphite_transform_loops): Connect infinite loops to exit
6338 and remove fake edges at the end.
6340 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6342 * ipa-inline.c (edge_badness): Revert accidental checkin.
6344 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6347 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6348 symbols; not inline clones.
6350 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6353 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6354 hard registers. Formatting fixes.
6356 PR preprocessor/83722
6357 * gcc.c (try_generate_repro): Pass
6358 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6359 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6362 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6363 Kito Cheng <kito.cheng@gmail.com>
6365 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6366 (riscv_leaf_function_p): Delete.
6367 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6369 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6371 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6373 (do_ifelse): New function.
6374 (do_isel): New function.
6375 (do_sub3): New function.
6376 (do_add3): New function.
6377 (do_load_mask_compare): New function.
6378 (do_overlap_load_compare): New function.
6379 (expand_compare_loop): New function.
6380 (expand_block_compare): Call expand_compare_loop() when appropriate.
6381 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6383 (-mblock-compare-inline-loop-limit): New option.
6385 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6388 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6389 Reverse order of second and third operands in first alternative.
6390 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6391 of first and second elements in UNSPEC_VPERMR vector.
6392 (altivec_expand_vec_perm_le): Likewise.
6394 2018-01-08 Jeff Law <law@redhat.com>
6396 PR rtl-optimizatin/81308
6397 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6398 (process_switch): If group_case_labels makes a change, then set
6400 (pass_convert_switch::execute): If a switch is converted, then
6401 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6403 PR rtl-optimization/81308
6404 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6407 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6409 PR target/83663 - Revert r255946
6410 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6411 generation for cases where splatting a value is not useful.
6412 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6413 across a vec_duplicate and a paradoxical subreg forming a vector
6414 mode to a vec_concat.
6416 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6418 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6419 -march=armv8.3-a variants.
6420 * config/arm/t-multilib: Likewise.
6421 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6423 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6425 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6427 (cceq_ior_compare_complement): Give it a name so I can use it, and
6428 change boolean_or_operator predicate to boolean_operator so it can
6429 be used to generate a crand.
6430 (eqne): New code iterator.
6431 (bd/bd_neg): New code_attrs.
6432 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6433 a single define_insn.
6434 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6435 decrement (bdnzt/bdnzf/bdzt/bdzf).
6436 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6437 with the new names of the branch decrement patterns, and added the
6438 names of the branch decrement conditional patterns.
6440 2018-01-08 Richard Biener <rguenther@suse.de>
6442 PR tree-optimization/83563
6443 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6446 2018-01-08 Richard Biener <rguenther@suse.de>
6449 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6451 2018-01-08 Richard Biener <rguenther@suse.de>
6453 PR tree-optimization/83685
6454 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6455 references to abnormals.
6457 2018-01-08 Richard Biener <rguenther@suse.de>
6460 * dwarf2out.c (output_indirect_strings): Handle empty
6461 skeleton_debug_str_hash.
6462 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6464 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6466 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6467 (emit_store_direct): Likewise.
6468 (arc_trampoline_adjust_address): Likewise.
6469 (arc_asm_trampoline_template): New function.
6470 (arc_initialize_trampoline): Use asm_trampoline_template.
6471 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6472 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6473 * config/arc/arc.md (flush_icache): Delete pattern.
6475 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6477 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6478 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6481 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6484 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6485 by not USED_FOR_TARGET.
6486 (make_pass_resolve_sw_modes): Likewise.
6488 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6490 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6493 2018-01-08 Richard Biener <rguenther@suse.de>
6496 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6498 2018-01-08 Richard Biener <rguenther@suse.de>
6501 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6503 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6506 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6507 basic blocks with a small number of successors.
6508 (convert_control_dep_chain_into_preds): Improve handling of
6510 (dump_predicates): Split apart into...
6511 (dump_pred_chain): ...here...
6512 (dump_pred_info): ...and here.
6513 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6514 (can_chain_union_be_invalidated_p): Improve check for invalidation
6516 (uninit_uses_cannot_happen): Avoid unnecessary if
6517 convert_control_dep_chain_into_preds yielded nothing.
6519 2018-01-06 Martin Sebor <msebor@redhat.com>
6521 PR tree-optimization/83640
6522 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6523 subtracting negative offset from size.
6524 (builtin_access::overlap): Adjust offset bounds of the access to fall
6525 within the size of the object if possible.
6527 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6529 PR rtl-optimization/83699
6530 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6531 extract_bit_field_as_subreg to cases in which the extracted
6532 value is also a vector.
6534 * lra-constraints.c (process_alt_operands): Test for the equivalence
6535 substitutions when detecting a possible reload cycle.
6537 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6540 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6541 by default if flag_selective_schedling{,2}. Formatting fixes.
6543 PR rtl-optimization/83682
6544 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6545 if it has non-VECTOR_MODE element mode.
6546 (vec_duplicate_p): Likewise.
6549 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6550 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6552 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6555 * config/i386/i386-builtin.def
6556 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6557 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6558 Require also OPTION_MASK_ISA_AVX512F in addition to
6559 OPTION_MASK_ISA_GFNI.
6560 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6561 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6562 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6563 to OPTION_MASK_ISA_GFNI.
6564 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6565 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6566 OPTION_MASK_ISA_AVX512BW.
6567 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6568 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6569 addition to OPTION_MASK_ISA_GFNI.
6570 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6571 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6572 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6573 to OPTION_MASK_ISA_GFNI.
6574 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6575 a requirement for all ISAs rather than any of them with a few
6577 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6579 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6580 bitmasks to be enabled with 3 exceptions, instead of requiring any
6581 enabled ISA with lots of exceptions.
6582 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6583 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6584 Change avx512bw in isa attribute to avx512f.
6585 * config/i386/sgxintrin.h: Add license boilerplate.
6586 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6587 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6588 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6589 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6591 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6592 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6593 temporarily sse2 rather than sse if not enabled already.
6596 * config/i386/sse.md (VI248_VLBW): Rename to ...
6597 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6598 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6599 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6600 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6601 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6602 mode iterator instead of VI248_VLBW.
6604 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6606 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6607 (record_modified): Skip clobbers; add debug output.
6608 (param_change_prob): Use sreal frequencies.
6610 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6612 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6613 punt for user-aligned variables.
6615 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6617 * tree-chrec.c (chrec_contains_symbols): Return true for
6620 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6623 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6624 of (x|y) == x for BICS pattern.
6626 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6628 PR tree-optimization/83605
6629 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6630 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6633 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6635 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6636 * config/epiphany/rtems.h: New file.
6638 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6639 Uros Bizjak <ubizjak@gmail.com>
6642 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6643 QIreg_operand instead of register_operand predicate.
6644 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6645 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6646 comments instead of -fmitigate[-_]rop.
6648 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6651 * cgraphunit.c (symbol_table::compile): Switch to text_section
6652 before calling assembly_start debug hook.
6653 * run-rtl-passes.c (run_rtl_passes): Likewise.
6656 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6658 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6659 range_int_cst_p rather than !symbolic_range_p before calling
6660 extract_range_from_multiplicative_op_1.
6662 2018-01-04 Jeff Law <law@redhat.com>
6664 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6665 redundant test in assertion.
6667 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6669 * doc/rtl.texi: Document machine_mode wrapper classes.
6671 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6673 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6676 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6678 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6679 the VEC_PERM_EXPR fold to fail.
6681 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6684 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6685 to switched_sections.
6687 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6690 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6693 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6696 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6697 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6699 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6702 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6703 is BLKmode and bitpos not zero or mode change is needed.
6705 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6708 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6711 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6714 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6715 instead of MULT rtx. Update all corresponding splitters.
6717 (*ssub<modesuffix>): Ditto.
6719 (*cmp_sadd_di): Update split patterns.
6720 (*cmp_sadd_si): Ditto.
6721 (*cmp_sadd_sidi): Ditto.
6722 (*cmp_ssub_di): Ditto.
6723 (*cmp_ssub_si): Ditto.
6724 (*cmp_ssub_sidi): Ditto.
6725 * config/alpha/predicates.md (const23_operand): New predicate.
6726 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6727 Look for ASHIFT, not MULT inner operand.
6728 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6730 2018-01-04 Martin Liska <mliska@suse.cz>
6732 PR gcov-profile/83669
6733 * gcov.c (output_intermediate_file): Add version to intermediate
6735 * doc/gcov.texi: Document new field 'version' in intermediate
6736 file format. Fix location of '-k' option of gcov command.
6738 2018-01-04 Martin Liska <mliska@suse.cz>
6741 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6743 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6745 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6747 2018-01-03 Martin Sebor <msebor@redhat.com>
6749 PR tree-optimization/83655
6750 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6751 checking calls with invalid arguments.
6753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6755 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6756 (vectorizable_mask_load_store): Delete.
6757 (vectorizable_call): Return false for masked loads and stores.
6758 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6759 instead of gimple_assign_rhs1.
6760 (vectorizable_load): Handle IFN_MASK_LOAD.
6761 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6765 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6767 (vectorizable_mask_load_store): ...here.
6768 (vectorizable_load): ...and here.
6770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6772 * tree-vect-stmts.c (vect_build_all_ones_mask)
6773 (vect_build_zero_merge_argument): New functions, split out from...
6774 (vectorizable_load): ...here.
6776 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6778 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6780 (vectorizable_mask_load_store): ...here.
6781 (vectorizable_store): ...and here.
6783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6785 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6787 (vectorizable_mask_load_store): ...here.
6789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6791 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6792 (vect_model_store_cost): Take a vec_load_store_type instead of a
6794 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6795 (vect_model_store_cost): Take a vec_load_store_type instead of a
6797 (vectorizable_mask_load_store): Update accordingly.
6798 (vectorizable_store): Likewise.
6799 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6803 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6804 IFN_MASK_LOAD calls here rather than...
6805 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6808 Alan Hayward <alan.hayward@arm.com>
6809 David Sherwood <david.sherwood@arm.com>
6811 * expmed.c (extract_bit_field_1): For vector extracts,
6812 fall back to extract_bit_field_as_subreg if vec_extract
6815 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6816 Alan Hayward <alan.hayward@arm.com>
6817 David Sherwood <david.sherwood@arm.com>
6819 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6820 they are variable or constant sized.
6821 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6822 slots for constant-sized data.
6824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6825 Alan Hayward <alan.hayward@arm.com>
6826 David Sherwood <david.sherwood@arm.com>
6828 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6829 handling COND_EXPRs with boolean comparisons, try to find a better
6830 basis for the mask type than the boolean itself.
6832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6834 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6835 is calculated and how it can be overridden.
6836 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6837 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6839 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6843 Alan Hayward <alan.hayward@arm.com>
6844 David Sherwood <david.sherwood@arm.com>
6846 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6847 Remove the mode argument.
6848 (aarch64_simd_valid_immediate): Remove the mode and inverse
6850 * config/aarch64/iterators.md (bitsize): New iterator.
6851 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6852 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6853 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6854 aarch64_simd_valid_immediate.
6855 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6856 (aarch64_reg_or_bic_imm): Likewise.
6857 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6858 with an insn_type enum and msl with a modifier_type enum.
6859 Replace element_width with a scalar_mode. Change the shift
6860 to unsigned int. Add constructors for scalar_float_mode and
6861 scalar_int_mode elements.
6862 (aarch64_vect_float_const_representable_p): Delete.
6863 (aarch64_can_const_movi_rtx_p)
6864 (aarch64_simd_scalar_immediate_valid_for_move)
6865 (aarch64_simd_make_constant): Update call to
6866 aarch64_simd_valid_immediate.
6867 (aarch64_advsimd_valid_immediate_hs): New function.
6868 (aarch64_advsimd_valid_immediate): Likewise.
6869 (aarch64_simd_valid_immediate): Remove mode and inverse
6870 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6871 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6872 and aarch64_float_const_representable_p on the result.
6873 (aarch64_output_simd_mov_immediate): Remove mode argument.
6874 Update call to aarch64_simd_valid_immediate and use of
6875 simd_immediate_info.
6876 (aarch64_output_scalar_simd_mov_immediate): Update call
6879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6880 Alan Hayward <alan.hayward@arm.com>
6881 David Sherwood <david.sherwood@arm.com>
6883 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6884 (mode_nunits): Likewise CONST_MODE_NUNITS.
6885 * machmode.def (ADJUST_NUNITS): Document.
6886 * genmodes.c (mode_data::need_nunits_adj): New field.
6887 (blank_mode): Update accordingly.
6888 (adj_nunits): New variable.
6889 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6891 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6892 listed in adj_nunits.
6893 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6894 listed in adj_nunits. Don't emit case statements for such modes.
6895 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6896 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6897 nothing if adj_nunits is nonnull.
6898 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6899 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6900 (emit_mode_fbit): Update use of print_maybe_const_decl.
6901 (emit_move_size): Likewise. Treat the array as non-const
6903 (emit_mode_adjustments): Handle adj_nunits.
6905 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6907 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6908 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6909 (VECTOR_MODES): Use it.
6910 (make_vector_modes): Take the prefix as an argument.
6912 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6913 Alan Hayward <alan.hayward@arm.com>
6914 David Sherwood <david.sherwood@arm.com>
6916 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6917 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6918 for MODE_VECTOR_BOOL.
6919 * machmode.def (VECTOR_BOOL_MODE): Document.
6920 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6921 (make_vector_bool_mode): New function.
6922 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6924 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6925 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6927 * stor-layout.c (int_mode_for_mode): Likewise.
6928 * tree.c (build_vector_type_for_mode): Likewise.
6929 * varasm.c (output_constant_pool_2): Likewise.
6930 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6931 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6932 for MODE_VECTOR_BOOL.
6933 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6934 of mode class checks.
6935 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6936 instead of a list of mode class checks.
6937 (expand_vector_scalar_condition): Likewise.
6938 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6940 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6941 Alan Hayward <alan.hayward@arm.com>
6942 David Sherwood <david.sherwood@arm.com>
6944 * machmode.h (mode_size): Change from unsigned short to
6946 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6947 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6948 or if measurement_type is not polynomial.
6949 (fixed_size_mode::includes_p): Check for constant-sized modes.
6950 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6951 return a poly_uint16 rather than an unsigned short.
6952 (emit_mode_size): Change the type of mode_size from unsigned short
6953 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6954 (emit_mode_adjustments): Cope with polynomial vector sizes.
6955 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6957 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6959 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6960 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6961 * caller-save.c (setup_save_areas): Likewise.
6962 (replace_reg_with_saved_mem): Likewise.
6963 * calls.c (emit_library_call_value_1): Likewise.
6964 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6965 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6966 (gen_lowpart_for_combine): Likewise.
6967 * convert.c (convert_to_integer_1): Likewise.
6968 * cse.c (equiv_constant, cse_insn): Likewise.
6969 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6970 (cselib_subst_to_values): Likewise.
6971 * dce.c (word_dce_process_block): Likewise.
6972 * df-problems.c (df_word_lr_mark_ref): Likewise.
6973 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6974 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6975 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6976 (rtl_for_decl_location): Likewise.
6977 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6978 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6979 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6980 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6981 (expand_expr_real_1): Likewise.
6982 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6983 (pad_below): Likewise.
6984 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6985 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6986 * ira.c (get_subreg_tracking_sizes): Likewise.
6987 * ira-build.c (ira_create_allocno_objects): Likewise.
6988 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6989 (ira_sort_regnos_for_alter_reg): Likewise.
6990 * ira-costs.c (record_operand_costs): Likewise.
6991 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6992 (resolve_simple_move): Likewise.
6993 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6994 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6995 (lra_constraints): Likewise.
6996 (CONST_POOL_OK_P): Reject variable-sized modes.
6997 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6998 (add_pseudo_to_slot, lra_spill): Likewise.
6999 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7000 * optabs-query.c (get_best_extraction_insn): Likewise.
7001 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7002 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
7003 (expand_mult_highpart, valid_multiword_target_p): Likewise.
7004 * recog.c (offsettable_address_addr_space_p): Likewise.
7005 * regcprop.c (maybe_mode_change): Likewise.
7006 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7007 * regrename.c (build_def_use): Likewise.
7008 * regstat.c (dump_reg_info): Likewise.
7009 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7010 (find_reloads, find_reloads_subreg_address): Likewise.
7011 * reload1.c (eliminate_regs_1): Likewise.
7012 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7013 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7014 (simplify_binary_operation_1, simplify_subreg): Likewise.
7015 * targhooks.c (default_function_arg_padding): Likewise.
7016 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7017 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7018 (verify_gimple_assign_ternary): Likewise.
7019 * tree-inline.c (estimate_move_cost): Likewise.
7020 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7021 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7022 (get_address_cost_ainc): Likewise.
7023 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7024 (vect_supportable_dr_alignment): Likewise.
7025 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7026 (vectorizable_reduction): Likewise.
7027 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7028 (vectorizable_operation, vectorizable_load): Likewise.
7029 * tree.c (build_same_sized_truth_vector_type): Likewise.
7030 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7031 * var-tracking.c (emit_note_insn_var_location): Likewise.
7032 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7033 (ADDR_VEC_ALIGN): Likewise.
7035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7036 Alan Hayward <alan.hayward@arm.com>
7037 David Sherwood <david.sherwood@arm.com>
7039 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7041 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7042 or if measurement_type is polynomial.
7043 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7044 * combine.c (make_extraction): Likewise.
7045 * dse.c (find_shift_sequence): Likewise.
7046 * dwarf2out.c (mem_loc_descriptor): Likewise.
7047 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7048 (extract_bit_field, extract_low_bits): Likewise.
7049 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7050 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7051 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7052 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7053 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7054 * reload.c (find_reloads): Likewise.
7055 * reload1.c (alter_reg): Likewise.
7056 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7057 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7058 * tree-if-conv.c (predicate_mem_writes): Likewise.
7059 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7060 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7061 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7062 * valtrack.c (dead_debug_insert_temp): Likewise.
7063 * varasm.c (mergeable_constant_section): Likewise.
7064 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7067 Alan Hayward <alan.hayward@arm.com>
7068 David Sherwood <david.sherwood@arm.com>
7070 * expr.c (expand_assignment): Cope with polynomial mode sizes
7071 when assigning to a CONCAT.
7073 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7074 Alan Hayward <alan.hayward@arm.com>
7075 David Sherwood <david.sherwood@arm.com>
7077 * machmode.h (mode_precision): Change from unsigned short to
7079 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7081 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7082 or if measurement_type is not polynomial.
7083 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7084 in which the mode is already known to be a scalar_int_mode.
7085 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7086 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7088 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7089 for GET_MODE_PRECISION.
7090 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7091 for GET_MODE_PRECISION.
7092 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7094 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7095 (expand_field_assignment, make_extraction): Likewise.
7096 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7097 (get_last_value): Likewise.
7098 * convert.c (convert_to_integer_1): Likewise.
7099 * cse.c (cse_insn): Likewise.
7100 * expr.c (expand_expr_real_1): Likewise.
7101 * lra-constraints.c (simplify_operand_subreg): Likewise.
7102 * optabs-query.c (can_atomic_load_p): Likewise.
7103 * optabs.c (expand_atomic_load): Likewise.
7104 (expand_atomic_store): Likewise.
7105 * ree.c (combine_reaching_defs): Likewise.
7106 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7107 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7108 * tree.h (type_has_mode_precision_p): Likewise.
7109 * ubsan.c (instrument_si_overflow): Likewise.
7111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7112 Alan Hayward <alan.hayward@arm.com>
7113 David Sherwood <david.sherwood@arm.com>
7115 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7116 polynomial numbers of units.
7117 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7118 (valid_vector_subparts_p): New function.
7119 (build_vector_type): Remove temporary shim and take the number
7120 of units as a poly_uint64 rather than an int.
7121 (build_opaque_vector_type): Take the number of units as a
7122 poly_uint64 rather than an int.
7123 * tree.c (build_vector_from_ctor): Handle polynomial
7124 TYPE_VECTOR_SUBPARTS.
7125 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7126 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7127 (build_vector_from_val): If the number of units is variable,
7128 use build_vec_duplicate_cst for constant operands and
7129 VEC_DUPLICATE_EXPR otherwise.
7130 (make_vector_type): Remove temporary is_constant ().
7131 (build_vector_type, build_opaque_vector_type): Take the number of
7132 units as a poly_uint64 rather than an int.
7133 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7135 * cfgexpand.c (expand_debug_expr): Likewise.
7136 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7137 (store_constructor, expand_expr_real_1): Likewise.
7138 (const_scalar_mask_from_tree): Likewise.
7139 * fold-const-call.c (fold_const_reduction): Likewise.
7140 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7141 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7142 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7143 (fold_relational_const): Likewise.
7144 (native_interpret_vector): Likewise. Change the size from an
7145 int to an unsigned int.
7146 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7147 TYPE_VECTOR_SUBPARTS.
7148 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7149 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7150 duplicating a non-constant operand into a variable-length vector.
7151 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7152 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7153 * ipa-icf.c (sem_variable::equals): Likewise.
7154 * match.pd: Likewise.
7155 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7156 * print-tree.c (print_node): Likewise.
7157 * stor-layout.c (layout_type): Likewise.
7158 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7159 * tree-cfg.c (verify_gimple_comparison): Likewise.
7160 (verify_gimple_assign_binary): Likewise.
7161 (verify_gimple_assign_ternary): Likewise.
7162 (verify_gimple_assign_single): Likewise.
7163 * tree-pretty-print.c (dump_generic_node): Likewise.
7164 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7165 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7166 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7167 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7168 (vect_shift_permute_load_chain): Likewise.
7169 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7170 (expand_vector_condition, optimize_vector_constructor): Likewise.
7171 (lower_vec_perm, get_compute_type): Likewise.
7172 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7173 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7174 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7175 (vect_recog_mask_conversion_pattern): Likewise.
7176 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7177 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7178 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7179 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7180 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7181 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7182 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7183 (supportable_widening_operation): Likewise.
7184 (supportable_narrowing_operation): Likewise.
7185 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7187 * varasm.c (output_constant): Likewise.
7189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7190 Alan Hayward <alan.hayward@arm.com>
7191 David Sherwood <david.sherwood@arm.com>
7193 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7194 so that both the length == 3 and length != 3 cases set up their
7195 own permute vectors. Add comments explaining why we know the
7196 number of elements is constant.
7197 (vect_permute_load_chain): Likewise.
7199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7200 Alan Hayward <alan.hayward@arm.com>
7201 David Sherwood <david.sherwood@arm.com>
7203 * machmode.h (mode_nunits): Change from unsigned char to
7205 (ONLY_FIXED_SIZE_MODES): New macro.
7206 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7207 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7208 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7210 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7211 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7212 or if measurement_type is not polynomial.
7213 * genmodes.c (ZERO_COEFFS): New macro.
7214 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7216 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7217 Use ZERO_COEFFS when emitting initializers.
7218 * data-streamer.h (bp_pack_poly_value): New function.
7219 (bp_unpack_poly_value): Likewise.
7220 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7221 for GET_MODE_NUNITS.
7222 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7223 for GET_MODE_NUNITS.
7224 * tree.c (make_vector_type): Remove temporary shim and make
7225 the real function take the number of units as a poly_uint64
7227 (build_vector_type_for_mode): Handle polynomial nunits.
7228 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7229 * emit-rtl.c (const_vec_series_p_1): Likewise.
7230 (gen_rtx_CONST_VECTOR): Likewise.
7231 * fold-const.c (test_vec_duplicate_folding): Likewise.
7232 * genrecog.c (validate_pattern): Likewise.
7233 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7234 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7235 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7236 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7237 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7238 * rtlanal.c (subreg_get_info): Likewise.
7239 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7240 (vect_grouped_load_supported): Likewise.
7241 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7242 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7243 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7244 (simplify_const_unary_operation, simplify_binary_operation_1)
7245 (simplify_const_binary_operation, simplify_ternary_operation)
7246 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7247 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7248 instead of CONST_VECTOR_NUNITS.
7249 * varasm.c (output_constant_pool_2): Likewise.
7250 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7251 explicit-encoded elements in the XVEC for variable-length vectors.
7253 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7255 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7258 Alan Hayward <alan.hayward@arm.com>
7259 David Sherwood <david.sherwood@arm.com>
7261 * coretypes.h (fixed_size_mode): Declare.
7262 (fixed_size_mode_pod): New typedef.
7263 * builtins.h (target_builtins::x_apply_args_mode)
7264 (target_builtins::x_apply_result_mode): Change type to
7265 fixed_size_mode_pod.
7266 * builtins.c (apply_args_size, apply_result_size, result_vector)
7267 (expand_builtin_apply_args_1, expand_builtin_apply)
7268 (expand_builtin_return): Update accordingly.
7270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7272 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7273 * cselib.c (cselib_hash_rtx): Likewise.
7274 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7275 CONST_VECTOR encoding.
7277 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7278 Jeff Law <law@redhat.com>
7281 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7282 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7283 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7284 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7287 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7288 explicitly probe *sp in a noreturn function if there were any callee
7289 register saves or frame pointer is needed.
7291 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7294 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7295 BLKmode for ternary, binary or unary expressions.
7298 * var-tracking.c (delete_vta_debug_insn): New inline function.
7299 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7300 insns from get_insns () to NULL instead of each bb separately.
7301 Use delete_vta_debug_insn. No longer static.
7302 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7303 delete_vta_debug_insns callers.
7304 * rtl.h (delete_vta_debug_insns): Declare.
7305 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7306 instead of variable_tracking_main.
7308 2018-01-03 Martin Sebor <msebor@redhat.com>
7310 PR tree-optimization/83603
7311 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7312 arguments past the endof the argument list in functions declared
7313 without a prototype.
7314 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7315 Avoid checking when arguments are null.
7317 2018-01-03 Martin Sebor <msebor@redhat.com>
7320 * doc/extend.texi (attribute const): Fix a typo.
7321 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7322 issuing -Wsuggest-attribute for void functions.
7324 2018-01-03 Martin Sebor <msebor@redhat.com>
7326 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7327 offset_int::from instead of wide_int::to_shwi.
7328 (maybe_diag_overlap): Remove assertion.
7329 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7330 * gimple-ssa-sprintf.c (format_directive): Same.
7331 (parse_directive): Same.
7332 (sprintf_dom_walker::compute_format_length): Same.
7333 (try_substitute_return_value): Same.
7335 2018-01-03 Jeff Law <law@redhat.com>
7338 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7339 non-constant residual for zero at runtime and avoid probing in
7340 that case. Reorganize code for trailing problem to mirror handling
7343 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7345 PR tree-optimization/83501
7346 * tree-ssa-strlen.c (get_string_cst): New.
7347 (handle_char_store): Call get_string_cst.
7349 2018-01-03 Martin Liska <mliska@suse.cz>
7351 PR tree-optimization/83593
7352 * tree-ssa-strlen.c: Include tree-cfg.h.
7353 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7354 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7355 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7357 (strlen_dom_walker::before_dom_children): Call
7358 gimple_purge_dead_eh_edges. Dump tranformation with details
7360 (strlen_dom_walker::before_dom_children): Update call by adding
7361 new argument cleanup_eh.
7362 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7364 2018-01-03 Martin Liska <mliska@suse.cz>
7367 * cif-code.def (VARIADIC_THUNK): New enum value.
7368 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7371 2018-01-03 Jan Beulich <jbeulich@suse.com>
7373 * sse.md (mov<mode>_internal): Tighten condition for when to use
7374 vmovdqu<ssescalarsize> for TI and OI modes.
7376 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7378 Update copyright years.
7380 2018-01-03 Martin Liska <mliska@suse.cz>
7383 * ipa-visibility.c (function_and_variable_visibility): Skip
7384 functions with noipa attribure.
7386 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7388 * gcc.c (process_command): Update copyright notice dates.
7389 * gcov-dump.c (print_version): Ditto.
7390 * gcov.c (print_version): Ditto.
7391 * gcov-tool.c (print_version): Ditto.
7392 * gengtype.c (create_file): Ditto.
7393 * doc/cpp.texi: Bump @copying's copyright year.
7394 * doc/cppinternals.texi: Ditto.
7395 * doc/gcc.texi: Ditto.
7396 * doc/gccint.texi: Ditto.
7397 * doc/gcov.texi: Ditto.
7398 * doc/install.texi: Ditto.
7399 * doc/invoke.texi: Ditto.
7401 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7403 * vector-builder.h (vector_builder::m_full_nelts): Change from
7404 unsigned int to poly_uint64.
7405 (vector_builder::full_nelts): Update prototype accordingly.
7406 (vector_builder::new_vector): Likewise.
7407 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7408 (vector_builder::operator ==): Likewise.
7409 (vector_builder::finalize): Likewise.
7410 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7411 Take the number of elements as a poly_uint64 rather than an
7413 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7414 from unsigned int to poly_uint64.
7415 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7416 (vec_perm_indices::new_vector): Likewise.
7417 (vec_perm_indices::length): Likewise.
7418 (vec_perm_indices::nelts_per_input): Likewise.
7419 (vec_perm_indices::input_nelts): Likewise.
7420 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7421 number of elements per input as a poly_uint64 rather than an
7422 unsigned int. Use the original encoding for variable-length
7423 vectors, rather than clamping each individual element.
7424 For the second and subsequent elements in each pattern,
7425 clamp the step and base before clamping their sum.
7426 (vec_perm_indices::series_p): Handle polynomial element counts.
7427 (vec_perm_indices::all_in_range_p): Likewise.
7428 (vec_perm_indices_to_tree): Likewise.
7429 (vec_perm_indices_to_rtx): Likewise.
7430 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7431 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7432 (tree_vector_builder::new_binary_operation): Handle polynomial
7433 element counts. Return false if we need to know the number
7434 of elements at compile time.
7435 * fold-const.c (fold_vec_perm): Punt if the number of elements
7436 isn't known at compile time.
7438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7440 * vec-perm-indices.h (vec_perm_builder): Change element type
7441 from HOST_WIDE_INT to poly_int64.
7442 (vec_perm_indices::element_type): Update accordingly.
7443 (vec_perm_indices::clamp): Handle polynomial element_types.
7444 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7445 (vec_perm_indices::all_in_range_p): Likewise.
7446 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7448 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7449 polynomial vec_perm_indices element types.
7450 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7451 * fold-const.c (fold_vec_perm): Likewise.
7452 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7453 * tree-vect-generic.c (lower_vec_perm): Likewise.
7454 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7455 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7456 element type to HOST_WIDE_INT.
7458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7459 Alan Hayward <alan.hayward@arm.com>
7460 David Sherwood <david.sherwood@arm.com>
7462 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7463 rather than an int. Use plus_constant.
7464 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7465 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7467 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7468 Alan Hayward <alan.hayward@arm.com>
7469 David Sherwood <david.sherwood@arm.com>
7471 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7472 a HOST_WIDE_INT to a poly_int64.
7474 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7475 Alan Hayward <alan.hayward@arm.com>
7476 David Sherwood <david.sherwood@arm.com>
7478 * calls.c (load_register_parameters): Cope with polynomial
7479 mode sizes. Require a constant size for BLKmode parameters
7480 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7481 forces a parameter to be padded at the lsb end in order to
7482 fill a complete number of words, require the parameter size
7483 to be ordered wrt UNITS_PER_WORD.
7485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7486 Alan Hayward <alan.hayward@arm.com>
7487 David Sherwood <david.sherwood@arm.com>
7489 * reload1.c (spill_stack_slot_width): Change element type
7490 from unsigned int to poly_uint64_pod.
7491 (alter_reg): Treat mode sizes as polynomial.
7493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7494 Alan Hayward <alan.hayward@arm.com>
7495 David Sherwood <david.sherwood@arm.com>
7497 * reload.c (complex_word_subreg_p): New function.
7498 (reload_inner_reg_of_subreg, push_reload): Use it.
7500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7501 Alan Hayward <alan.hayward@arm.com>
7502 David Sherwood <david.sherwood@arm.com>
7504 * lra-constraints.c (process_alt_operands): Reject matched
7505 operands whose sizes aren't ordered.
7506 (match_reload): Refer to this check here.
7508 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7509 Alan Hayward <alan.hayward@arm.com>
7510 David Sherwood <david.sherwood@arm.com>
7512 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7513 that the mode size is in the set {1, 2, 4, 8, 16}.
7515 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7516 Alan Hayward <alan.hayward@arm.com>
7517 David Sherwood <david.sherwood@arm.com>
7519 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7520 Use plus_constant instead of gen_rtx_PLUS.
7522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7523 Alan Hayward <alan.hayward@arm.com>
7524 David Sherwood <david.sherwood@arm.com>
7526 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7527 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7528 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7529 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7530 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7531 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7532 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7533 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7534 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7535 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7537 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7538 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7539 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7540 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7541 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7542 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7543 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7544 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7545 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7546 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7548 * expr.c (emit_move_resolve_push): Treat the input and result
7549 of PUSH_ROUNDING as a poly_int64.
7550 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7551 (emit_push_insn): Likewise.
7552 * lra-eliminations.c (mark_not_eliminable): Likewise.
7553 * recog.c (push_operand): Likewise.
7554 * reload1.c (elimination_effects): Likewise.
7555 * rtlanal.c (nonzero_bits1): Likewise.
7556 * calls.c (store_one_arg): Likewise. Require the padding to be
7557 known at compile time.
7559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7560 Alan Hayward <alan.hayward@arm.com>
7561 David Sherwood <david.sherwood@arm.com>
7563 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7564 Use plus_constant instead of gen_rtx_PLUS.
7566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7567 Alan Hayward <alan.hayward@arm.com>
7568 David Sherwood <david.sherwood@arm.com>
7570 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7574 Alan Hayward <alan.hayward@arm.com>
7575 David Sherwood <david.sherwood@arm.com>
7577 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7578 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7579 via stack temporaries. Treat the mode size as polynomial too.
7581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7582 Alan Hayward <alan.hayward@arm.com>
7583 David Sherwood <david.sherwood@arm.com>
7585 * expr.c (expand_expr_real_2): When handling conversions involving
7586 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7587 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7588 as a poly_uint64 too.
7590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7591 Alan Hayward <alan.hayward@arm.com>
7592 David Sherwood <david.sherwood@arm.com>
7594 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7596 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7597 Alan Hayward <alan.hayward@arm.com>
7598 David Sherwood <david.sherwood@arm.com>
7600 * combine.c (can_change_dest_mode): Handle polynomial
7601 REGMODE_NATURAL_SIZE.
7602 * expmed.c (store_bit_field_1): Likewise.
7603 * expr.c (store_constructor): Likewise.
7604 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7605 and polynomial REGMODE_NATURAL_SIZE.
7606 (gen_lowpart_common): Likewise.
7607 * reginfo.c (record_subregs_of_mode): Likewise.
7608 * rtlanal.c (read_modify_subreg_p): Likewise.
7610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7611 Alan Hayward <alan.hayward@arm.com>
7612 David Sherwood <david.sherwood@arm.com>
7614 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7615 numbers of elements.
7617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7618 Alan Hayward <alan.hayward@arm.com>
7619 David Sherwood <david.sherwood@arm.com>
7621 * match.pd: Cope with polynomial numbers of vector elements.
7623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7624 Alan Hayward <alan.hayward@arm.com>
7625 David Sherwood <david.sherwood@arm.com>
7627 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7628 in a POINTER_PLUS_EXPR.
7630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7631 Alan Hayward <alan.hayward@arm.com>
7632 David Sherwood <david.sherwood@arm.com>
7634 * omp-simd-clone.c (simd_clone_subparts): New function.
7635 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7636 (ipa_simd_modify_function_body): Likewise.
7638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7639 Alan Hayward <alan.hayward@arm.com>
7640 David Sherwood <david.sherwood@arm.com>
7642 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7643 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7644 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7645 (expand_vector_condition, vector_element): Likewise.
7646 (subparts_gt): New function.
7647 (get_compute_type): Use subparts_gt.
7648 (count_type_subparts): Delete.
7649 (expand_vector_operations_1): Use subparts_gt instead of
7650 count_type_subparts.
7652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7653 Alan Hayward <alan.hayward@arm.com>
7654 David Sherwood <david.sherwood@arm.com>
7656 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7657 (vect_compile_time_alias): ...this new function. Do the calculation
7658 on poly_ints rather than trees.
7659 (vect_prune_runtime_alias_test_list): Update call accordingly.
7661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7662 Alan Hayward <alan.hayward@arm.com>
7663 David Sherwood <david.sherwood@arm.com>
7665 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7667 (vect_schedule_slp_instance): Likewise.
7669 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7670 Alan Hayward <alan.hayward@arm.com>
7671 David Sherwood <david.sherwood@arm.com>
7673 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7674 constant and extern definitions for variable-length vectors.
7675 (vect_get_constant_vectors): Note that the number of units
7676 is known to be constant.
7678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7679 Alan Hayward <alan.hayward@arm.com>
7680 David Sherwood <david.sherwood@arm.com>
7682 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7683 of units as polynomial. Choose between WIDE and NARROW based
7686 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7687 Alan Hayward <alan.hayward@arm.com>
7688 David Sherwood <david.sherwood@arm.com>
7690 * tree-vect-stmts.c (simd_clone_subparts): New function.
7691 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7694 Alan Hayward <alan.hayward@arm.com>
7695 David Sherwood <david.sherwood@arm.com>
7697 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7698 vectors as polynomial. Use build_index_vector for
7701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7702 Alan Hayward <alan.hayward@arm.com>
7703 David Sherwood <david.sherwood@arm.com>
7705 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7706 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7707 for variable-length vectors.
7708 (vectorizable_mask_load_store): Treat the number of units as
7709 polynomial, asserting that it is constant if the condition has
7710 already been enforced.
7711 (vectorizable_store, vectorizable_load): Likewise.
7713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7714 Alan Hayward <alan.hayward@arm.com>
7715 David Sherwood <david.sherwood@arm.com>
7717 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7718 of units as polynomial. Punt if we can't tell at compile time
7719 which vector contains the final result.
7721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7722 Alan Hayward <alan.hayward@arm.com>
7723 David Sherwood <david.sherwood@arm.com>
7725 * tree-vect-loop.c (vectorizable_induction): Treat the number
7726 of units as polynomial. Punt on SLP inductions. Use an integer
7727 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7728 cast of such a series for variable-length floating-point
7731 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7732 Alan Hayward <alan.hayward@arm.com>
7733 David Sherwood <david.sherwood@arm.com>
7735 * tree.h (build_index_vector): Declare.
7736 * tree.c (build_index_vector): New function.
7737 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7738 of units as polynomial, forcibly converting it to a constant if
7739 vectorizable_reduction has already enforced the condition.
7740 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7741 to create a {1,2,3,...} vector.
7742 (vectorizable_reduction): Treat the number of units as polynomial.
7743 Choose vectype_in based on the largest scalar element size rather
7744 than the smallest number of units. Enforce the restrictions
7747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7748 Alan Hayward <alan.hayward@arm.com>
7749 David Sherwood <david.sherwood@arm.com>
7751 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7752 number of units as polynomial.
7754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7755 Alan Hayward <alan.hayward@arm.com>
7756 David Sherwood <david.sherwood@arm.com>
7758 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7759 * target.def (autovectorize_vector_sizes): Return the vector sizes
7760 by pointer, using vector_sizes rather than a bitmask.
7761 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7762 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7763 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7765 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7766 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7767 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7768 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7769 * omp-general.c (omp_max_vf): Likewise.
7770 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7771 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7772 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7773 * tree-vect-slp.c (vect_slp_bb): Likewise.
7774 * doc/tm.texi: Regenerate.
7775 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7777 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7778 the vector size as a poly_uint64 rather than an unsigned int.
7779 (current_vector_size): Change from an unsigned int to a poly_uint64.
7780 (get_vectype_for_scalar_type): Update accordingly.
7781 * tree.h (build_truth_vector_type): Take the size and number of
7782 units as a poly_uint64 rather than an unsigned int.
7783 (build_vector_type): Add a temporary overload that takes
7784 the number of units as a poly_uint64 rather than an unsigned int.
7785 * tree.c (make_vector_type): Likewise.
7786 (build_truth_vector_type): Take the number of units as a poly_uint64
7787 rather than an unsigned int.
7789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7790 Alan Hayward <alan.hayward@arm.com>
7791 David Sherwood <david.sherwood@arm.com>
7793 * target.def (get_mask_mode): Take the number of units and length
7794 as poly_uint64s rather than unsigned ints.
7795 * targhooks.h (default_get_mask_mode): Update accordingly.
7796 * targhooks.c (default_get_mask_mode): Likewise.
7797 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7798 * doc/tm.texi: Regenerate.
7800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7801 Alan Hayward <alan.hayward@arm.com>
7802 David Sherwood <david.sherwood@arm.com>
7804 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7805 * omp-general.c (omp_max_vf): Likewise.
7806 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7807 (expand_omp_simd): Handle polynomial safelen.
7808 * omp-low.c (omplow_simd_context): Add a default constructor.
7809 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7810 (lower_rec_simd_input_clauses): Update accordingly.
7811 (lower_rec_input_clauses): Likewise.
7813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7814 Alan Hayward <alan.hayward@arm.com>
7815 David Sherwood <david.sherwood@arm.com>
7817 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7818 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7819 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7820 (vect_analyze_slp_cost): Likewise.
7821 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7822 (vect_model_load_cost): Likewise.
7824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7825 Alan Hayward <alan.hayward@arm.com>
7826 David Sherwood <david.sherwood@arm.com>
7828 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7829 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7830 from an unsigned int * to a poly_uint64_pod *.
7831 (calculate_unrolling_factor): New function.
7832 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7835 Alan Hayward <alan.hayward@arm.com>
7836 David Sherwood <david.sherwood@arm.com>
7838 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7839 from an unsigned int to a poly_uint64.
7840 (_loop_vec_info::slp_unrolling_factor): Likewise.
7841 (_loop_vec_info::vectorization_factor): Change from an int
7843 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7844 (vect_get_num_vectors): New function.
7845 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7846 (vect_get_num_copies): Use vect_get_num_vectors.
7847 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7848 to an unsigned int *.
7849 (vect_analyze_data_refs): Change min_vf from an int * to a
7851 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7852 than an unsigned HOST_WIDE_INT.
7853 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7854 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7855 to an unsigned int *.
7856 (vect_analyze_data_ref_dependences): Likewise.
7857 (vect_compute_data_ref_alignment): Handle polynomial vf.
7858 (vect_enhance_data_refs_alignment): Likewise.
7859 (vect_prune_runtime_alias_test_list): Likewise.
7860 (vect_shift_permute_load_chain): Likewise.
7861 (vect_supportable_dr_alignment): Likewise.
7862 (dependence_distance_ge_vf): Take the vectorization factor as a
7863 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7864 (vect_analyze_data_refs): Change min_vf from an int * to a
7866 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7867 vfm1 as a poly_uint64 rather than an int. Make the same change
7868 for the returned bound_scalar.
7869 (vect_gen_vector_loop_niters): Handle polynomial vf.
7870 (vect_do_peeling): Likewise. Update call to
7871 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7872 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7874 * tree-vect-loop.c (vect_determine_vectorization_factor)
7875 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7876 (vect_get_known_peeling_cost): Likewise.
7877 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7878 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7879 (vect_transform_loop): Likewise. Use the lowest possible VF when
7880 updating the upper bounds of the loop.
7881 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7883 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7884 polynomial unroll factors.
7885 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7886 (vect_make_slp_decision): Likewise.
7887 (vect_supported_load_permutation_p): Likewise, and polynomial
7889 (vect_analyze_slp_cost): Handle polynomial vf.
7890 (vect_slp_analyze_node_operations): Likewise.
7891 (vect_slp_analyze_bb_1): Likewise.
7892 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7893 than an unsigned HOST_WIDE_INT.
7894 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7895 (vectorizable_load): Handle polynomial vf.
7896 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7898 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7900 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7901 Alan Hayward <alan.hayward@arm.com>
7902 David Sherwood <david.sherwood@arm.com>
7904 * match.pd: Handle bit operations involving three constants
7905 and try to fold one pair.
7907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7909 * tree-vect-loop-manip.c: Include gimple-fold.h.
7910 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7911 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7912 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7913 Add a path that uses a step of VF instead of 1, but disable it
7915 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7916 and niters_no_overflow parameters. Update calls to
7917 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7918 Create a new SSA name if the latter choses to use a ste other
7919 than zero, and return it via niters_vector_mult_vf_var.
7920 * tree-vect-loop.c (vect_transform_loop): Update calls to
7921 vect_do_peeling, vect_gen_vector_loop_niters and
7922 slpeel_make_loop_iterate_ntimes.
7923 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7924 (vect_gen_vector_loop_niters): Update declarations after above changes.
7926 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7928 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7929 128-bit round to integer instructions.
7930 (ceil<mode>2): Likewise.
7931 (btrunc<mode>2): Likewise.
7932 (round<mode>2): Likewise.
7934 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7936 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7937 unaligned VSX load/store on P8/P9.
7938 (expand_block_clear): Allow the use of unaligned VSX
7939 load/store on P8/P9.
7941 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7943 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7945 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7946 swap associated with both a load and a store.
7948 2018-01-02 Andrew Waterman <andrew@sifive.com>
7950 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7951 * config/riscv/riscv.md (clear_cache): Use it.
7953 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7955 * web.c: Remove out-of-date comment.
7957 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7959 * expr.c (fixup_args_size_notes): Check that any existing
7960 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7961 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7962 (emit_single_push_insn): ...here.
7964 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7966 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7967 (const_vector_encoded_nelts): New function.
7968 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7969 (const_vector_int_elt, const_vector_elt): Declare.
7970 * emit-rtl.c (const_vector_int_elt_1): New function.
7971 (const_vector_elt): Likewise.
7972 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7973 of CONST_VECTOR_ELT.
7975 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7977 * expr.c: Include rtx-vector-builder.h.
7978 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7979 directly on the tree encoding.
7980 (const_vector_from_tree): Likewise.
7981 * optabs.c: Include rtx-vector-builder.h.
7982 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7983 sequence of "u" values.
7984 * vec-perm-indices.c: Include rtx-vector-builder.h.
7985 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7986 directly on the vec_perm_indices encoding.
7988 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7990 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7991 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7992 * rtx-vector-builder.h: New file.
7993 * rtx-vector-builder.c: Likewise.
7994 * rtl.h (rtx_def::u2): Add a const_vector field.
7995 (CONST_VECTOR_NPATTERNS): New macro.
7996 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7997 (CONST_VECTOR_DUPLICATE_P): Likewise.
7998 (CONST_VECTOR_STEPPED_P): Likewise.
7999 (CONST_VECTOR_ENCODED_ELT): Likewise.
8000 (const_vec_duplicate_p): Check for a duplicated vector encoding.
8001 (unwrap_const_vec_duplicate): Likewise.
8002 (const_vec_series_p): Check for a non-duplicated vector encoding.
8003 Say that the function only returns true for integer vectors.
8004 * emit-rtl.c: Include rtx-vector-builder.h.
8005 (gen_const_vec_duplicate_1): Delete.
8006 (gen_const_vector): Call gen_const_vec_duplicate instead of
8007 gen_const_vec_duplicate_1.
8008 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8009 (gen_const_vec_duplicate): Use rtx_vector_builder.
8010 (gen_const_vec_series): Likewise.
8011 (gen_rtx_CONST_VECTOR): Likewise.
8012 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8013 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8014 Build a new vector rather than modifying a CONST_VECTOR in-place.
8015 (handle_special_swappables): Update call accordingly.
8016 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8017 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8018 Build a new vector rather than modifying a CONST_VECTOR in-place.
8019 (handle_special_swappables): Update call accordingly.
8021 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8023 * simplify-rtx.c (simplify_const_binary_operation): Use
8024 CONST_VECTOR_ELT instead of XVECEXP.
8026 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8028 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8029 the selector elements to be different from the data elements
8030 if the selector is a VECTOR_CST.
8031 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8032 ssizetype for the selector.
8034 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8036 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8037 before testing each element individually.
8038 * tree-vect-generic.c (lower_vec_perm): Likewise.
8040 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8042 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8043 * selftest-run-tests.c (selftest::run_tests): Call it.
8044 * vector-builder.h (vector_builder::operator ==): New function.
8045 (vector_builder::operator !=): Likewise.
8046 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8047 (vec_perm_indices::all_from_input_p): New function.
8048 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8049 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8050 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8051 instead of reading the VECTOR_CST directly. Detect whether both
8052 vector inputs are the same before constructing the vec_perm_indices,
8053 and update the number of inputs argument accordingly. Use the
8054 utility functions added above. Only construct sel2 if we need to.
8056 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8058 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8059 the broadcast of the low byte.
8060 (expand_mult_highpart): Use an explicit encoding for the permutes.
8061 * optabs-query.c (can_mult_highpart_p): Likewise.
8062 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8063 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8064 (vectorizable_bswap): Likewise.
8065 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8066 explicit encoding for the power-of-2 permutes.
8067 (vect_permute_store_chain): Likewise.
8068 (vect_grouped_load_supported): Likewise.
8069 (vect_permute_load_chain): Likewise.
8071 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8073 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8074 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8075 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8076 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8077 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8078 (vect_gen_perm_mask_any): Likewise.
8080 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8082 * int-vector-builder.h: New file.
8083 * vec-perm-indices.h: Include int-vector-builder.h.
8084 (vec_perm_indices): Redefine as an int_vector_builder.
8085 (auto_vec_perm_indices): Delete.
8086 (vec_perm_builder): Redefine as a stand-alone class.
8087 (vec_perm_indices::vec_perm_indices): New function.
8088 (vec_perm_indices::clamp): Likewise.
8089 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8090 (vec_perm_indices::new_vector): New function.
8091 (vec_perm_indices::new_expanded_vector): Update for new
8092 vec_perm_indices class.
8093 (vec_perm_indices::rotate_inputs): New function.
8094 (vec_perm_indices::all_in_range_p): Operate directly on the
8095 encoded form, without computing elided elements.
8096 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8097 encoding. Update for new vec_perm_indices class.
8098 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8099 the given vec_perm_builder.
8100 (expand_vec_perm_var): Update vec_perm_builder constructor.
8101 (expand_mult_highpart): Use vec_perm_builder instead of
8102 auto_vec_perm_indices.
8103 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8104 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8105 or double series encoding as appropriate.
8106 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8107 vec_perm_indices instead of auto_vec_perm_indices.
8108 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8109 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8110 (vect_permute_store_chain): Likewise.
8111 (vect_grouped_load_supported): Likewise.
8112 (vect_permute_load_chain): Likewise.
8113 (vect_shift_permute_load_chain): Likewise.
8114 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8115 (vect_transform_slp_perm_load): Likewise.
8116 (vect_schedule_slp_instance): Likewise.
8117 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8118 (vectorizable_mask_load_store): Likewise.
8119 (vectorizable_bswap): Likewise.
8120 (vectorizable_store): Likewise.
8121 (vectorizable_load): Likewise.
8122 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8123 vec_perm_indices instead of auto_vec_perm_indices. Use
8124 tree_to_vec_perm_builder to read the vector from a tree.
8125 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8126 vec_perm_builder instead of a vec_perm_indices.
8127 (have_whole_vector_shift): Use vec_perm_builder and
8128 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8129 truncation to calc_vec_perm_mask_for_shift.
8130 (vect_create_epilog_for_reduction): Likewise.
8131 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8132 from auto_vec_perm_indices to vec_perm_indices.
8133 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8134 instead of changing individual elements.
8135 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8136 the vector in d.perm.
8137 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8138 from auto_vec_perm_indices to vec_perm_indices.
8139 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8140 instead of changing individual elements.
8141 (arm_vectorize_vec_perm_const): Use new_vector to install
8142 the vector in d.perm.
8143 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8144 Update vec_perm_builder constructor.
8145 (rs6000_expand_interleave): Likewise.
8146 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8147 (rs6000_expand_interleave): Likewise.
8149 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8151 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8152 to qimode could truncate the indices.
8153 * optabs.c (expand_vec_perm_var): Likewise.
8155 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8157 * Makefile.in (OBJS): Add vec-perm-indices.o.
8158 * vec-perm-indices.h: New file.
8159 * vec-perm-indices.c: Likewise.
8160 * target.h (vec_perm_indices): Replace with a forward class
8162 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8163 * optabs.h: Include vec-perm-indices.h.
8164 (expand_vec_perm): Delete.
8165 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8166 (expand_vec_perm_const): Declare.
8167 * target.def (vec_perm_const_ok): Replace with...
8168 (vec_perm_const): ...this new hook.
8169 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8170 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8171 * doc/tm.texi: Regenerate.
8172 * optabs.def (vec_perm_const): Delete.
8173 * doc/md.texi (vec_perm_const): Likewise.
8174 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8175 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8176 expand_vec_perm for constant permutation vectors. Assert that
8177 the mode of variable permutation vectors is the integer equivalent
8178 of the mode that is being permuted.
8179 * optabs-query.h (selector_fits_mode_p): Declare.
8180 * optabs-query.c: Include vec-perm-indices.h.
8181 (selector_fits_mode_p): New function.
8182 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8183 is defined, instead of checking whether the vec_perm_const_optab
8184 exists. Use targetm.vectorize.vec_perm_const instead of
8185 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8186 fit in the vector mode before using a variable permute.
8187 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8188 vec_perm_indices instead of an rtx.
8189 (expand_vec_perm): Replace with...
8190 (expand_vec_perm_const): ...this new function. Take the selector
8191 as a vec_perm_indices rather than an rtx. Also take the mode of
8192 the selector. Update call to shift_amt_for_vec_perm_mask.
8193 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8194 Use vec_perm_indices::new_expanded_vector to expand the original
8195 selector into bytes. Check whether the indices fit in the vector
8196 mode before using a variable permute.
8197 (expand_vec_perm_var): Make global.
8198 (expand_mult_highpart): Use expand_vec_perm_const.
8199 * fold-const.c: Includes vec-perm-indices.h.
8200 * tree-ssa-forwprop.c: Likewise.
8201 * tree-vect-data-refs.c: Likewise.
8202 * tree-vect-generic.c: Likewise.
8203 * tree-vect-loop.c: Likewise.
8204 * tree-vect-slp.c: Likewise.
8205 * tree-vect-stmts.c: Likewise.
8206 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8208 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8209 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8210 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8211 (aarch64_vectorize_vec_perm_const): ...this new function.
8212 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8213 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8214 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8215 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8216 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8217 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8218 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8220 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8221 check for NEON modes.
8222 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8223 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8224 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8225 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8227 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8228 the old VEC_PERM_CONST conditions.
8229 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8230 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8231 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8232 (ia64_vectorize_vec_perm_const_ok): Merge into...
8233 (ia64_vectorize_vec_perm_const): ...this new function.
8234 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8235 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8236 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8237 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8238 * config/mips/mips.c (mips_expand_vec_perm_const)
8239 (mips_vectorize_vec_perm_const_ok): Merge into...
8240 (mips_vectorize_vec_perm_const): ...this new function.
8241 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8242 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8243 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8244 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8245 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8246 (rs6000_expand_vec_perm_const): Delete.
8247 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8249 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8250 (altivec_expand_vec_perm_const_le): Take each operand individually.
8251 Operate on constant selectors rather than rtxes.
8252 (altivec_expand_vec_perm_const): Likewise. Update call to
8253 altivec_expand_vec_perm_const_le.
8254 (rs6000_expand_vec_perm_const): Delete.
8255 (rs6000_vectorize_vec_perm_const_ok): Delete.
8256 (rs6000_vectorize_vec_perm_const): New function.
8257 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8258 an element count and rtx array.
8259 (rs6000_expand_extract_even): Update call accordingly.
8260 (rs6000_expand_interleave): Likewise.
8261 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8262 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8263 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8264 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8265 (rs6000_expand_vec_perm_const): Delete.
8266 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8267 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8268 (altivec_expand_vec_perm_const_le): Take each operand individually.
8269 Operate on constant selectors rather than rtxes.
8270 (altivec_expand_vec_perm_const): Likewise. Update call to
8271 altivec_expand_vec_perm_const_le.
8272 (rs6000_expand_vec_perm_const): Delete.
8273 (rs6000_vectorize_vec_perm_const_ok): Delete.
8274 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8275 reference to the SPE evmerge intructions.
8276 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8277 an element count and rtx array.
8278 (rs6000_expand_extract_even): Update call accordingly.
8279 (rs6000_expand_interleave): Likewise.
8280 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8281 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8283 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8285 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8287 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8288 vector mode and that that mode matches the mode of the data
8290 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8291 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8292 directly using expand_vec_perm_1 when forcing selectors into
8294 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8296 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8298 * optabs-query.h (can_vec_perm_p): Delete.
8299 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8300 * optabs-query.c (can_vec_perm_p): Split into...
8301 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8302 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8303 particular selector is valid.
8304 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8305 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8306 (vect_grouped_load_supported): Likewise.
8307 (vect_shift_permute_load_chain): Likewise.
8308 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8309 (vect_transform_slp_perm_load): Likewise.
8310 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8311 (vectorizable_bswap): Likewise.
8312 (vect_gen_perm_mask_checked): Likewise.
8313 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8314 implementations of variable permutation vectors into account
8315 when deciding which selector to use.
8316 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8317 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8318 with a false third argument.
8319 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8320 to test whether the constant selector is valid and can_vec_perm_var_p
8321 to test whether a variable selector is valid.
8323 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8325 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8326 * optabs-query.c (can_vec_perm_p): Likewise.
8327 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8328 instead of vec_perm_indices.
8329 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8330 (vect_gen_perm_mask_checked): Likewise,
8331 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8332 (vect_gen_perm_mask_checked): Likewise,
8334 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8336 * optabs-query.h (qimode_for_vec_perm): Declare.
8337 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8338 (qimode_for_vec_perm): ...this new function.
8339 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8341 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8343 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8344 does not have a conditional at the top.
8346 2018-01-02 Richard Biener <rguenther@suse.de>
8348 * ipa-inline.c (big_speedup_p): Fix expression.
8350 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8353 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8356 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8360 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8361 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8362 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8363 cond_taken_branch_cost 3->4.
8365 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8367 PR tree-optimization/83581
8368 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8369 TODO_cleanup_cfg if any changes have been made.
8372 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8373 convert_modes if target mode has the right side, but different mode
8377 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8378 last argument when extracting from CONCAT. If either from_real or
8379 from_imag is NULL, use expansion through memory. If result is not
8380 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8381 the parts directly to inner mode, if even that fails, use expansion
8385 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8386 check for bswap in mode rather than HImode and use that in expand_unop
8389 Copyright (C) 2018 Free Software Foundation, Inc.
8391 Copying and distribution of this file, with or without modification,
8392 are permitted in any medium without royalty provided the copyright
8393 notice and this notice are preserved.