1 2018-03-07 Jakub Jelinek <jakub@redhat.com>
4 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
6 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
7 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
10 * multiple_target.c: Include tree-inline.h and intl.h.
11 (expand_target_clones): Diagnose and fail if node->definition and
12 !tree_versionable_function_p (node->decl).
14 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
16 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
18 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
19 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
22 2018-03-06 Jakub Jelinek <jakub@redhat.com>
25 * combine.c (try_combine): Use reg_or_subregno instead of handling
26 just paradoxical SUBREGs and REGs.
28 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
30 * config/arc/arc.c (arc_finalize_pic): Remove function.
31 (arc_must_save_register): We use single base PIC register, remove
32 checks to save/restore the PIC register.
33 (arc_expand_prologue): Likewise.
34 * config/arc/arc-protos.h (arc_set_default_type_attributes):
36 (arc_verify_short): Likewise.
37 (arc_attr_type): Likewise.
38 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
39 (walk_stores): Likewise.
40 (arc_address_cost): Make it static.
41 (arc_verify_short): Likewise.
42 (branch_dest): Likewise.
43 (arc_attr_type): Likewise.
44 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
45 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
46 (arc_final_prescan_insn): Remove inserting the nops due to
47 hardware hazards. It is done in reorg step.
48 (insn_length_variant_t): Remove.
49 (insn_length_parameters_t): Likewise.
50 (arc_insn_length_parameters): Likewise.
51 (arc_get_insn_variants): Likewise.
52 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
54 2018-03-06 Jakub Jelinek <jakub@redhat.com>
57 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
60 PR tree-optimization/84687
61 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
63 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
65 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
67 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
68 Rename to ppc_speculation_barrier.
69 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
70 __builtin_ppc_speculation_barrier.
72 2018-03-05 Jakub Jelinek <jakub@redhat.com>
75 * combine.c (combine_simplify_rtx): Don't try to simplify if
76 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
79 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
81 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
82 to 32 bytes when compiling for POWER9.
84 2018-03-05 Jakub Jelinek <jakub@redhat.com>
87 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
88 regparm >= 3 with no arg reg available also for calls with
89 flag_force_indirect_call. Pass decl to ix86_function_regparm.
92 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
94 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
96 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
99 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
101 2018-03-05 Richard Biener <rguenther@suse.de>
103 PR tree-optimization/84486
104 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
105 When inserting a __builtin_assume_aligned call set the LHS
106 SSA name alignment info accordingly.
108 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
110 PR tree-optimization/84114
111 * config/aarch64/aarch64.c (aarch64_reassociation_width)
112 Avoid reassociation of FLOAT_MODE addition.
114 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
116 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
117 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
118 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
119 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
120 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
121 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
122 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
124 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
125 __builtin_ia32_wbinvd): New builtins.
126 (SPECIAL_ARGS2): New.
127 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
128 (SPECIAL_ARGS2): New.
129 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
130 (ix86_valid_target_attribute_inner_p): Ditto.
131 (ix86_init_mmx_sse_builtins): Add special_args2.
132 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
133 TARGET_WBNOINVD_P): New.
134 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
135 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
136 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
137 * config/i386/immintrin.h (_wbinvd): New intrinsic.
138 * config/i386/pconfigintrin.h: New file.
139 * config/i386/wbnoinvdintrin.h: Ditto.
140 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
141 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
143 2018-03-05 Richard Biener <rguenther@suse.de>
145 PR tree-optimization/84670
146 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
148 (BB_VISITED_WITH_VISITED_SUCCS): New define.
149 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
150 (compute_antic_aux): Only assert the number of values in ANTIC_IN
151 doesn't grow if all successors (recursively) were visited at least
154 2018-03-05 Richard Biener <rguenther@suse.de>
156 PR tree-optimization/84650
157 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
158 if executed in the loop pipeline.
160 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
162 * doc/configfiles.texi (Configuration Files): Move info about
163 conditionalizing $target-protos.h to...
164 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
165 differs from $target-protos.h.
167 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
168 Chung-Ju Wu <jasonwucj@gmail.com>
170 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
171 * config/nds32/nds32-multiple.md (setmemsi): Define.
172 * config/nds32/nds32-memory-manipulation.c
173 (nds32_gen_dup_4_byte_to_word_value): New.
174 (emit_setmem_word_loop): New.
175 (emit_setmem_byte_loop): New.
176 (nds32_expand_setmem_loop): New.
177 (nds32_expand_setmem_loop_v3m): New.
178 (nds32_expand_setmem_unroll): New.
179 (nds32_expand_setmem): New.
181 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
182 Chung-Ju Wu <jasonwucj@gmail.com>
184 * config/nds32/nds32-memory-manipulation.c
185 (nds32_emit_load_store): New.
186 (nds32_emit_post_inc_load_store): New.
187 (nds32_emit_mem_move): New.
188 (nds32_emit_mem_move_block): New.
189 (nds32_expand_movmemsi_loop_unknown_size): New.
190 (nds32_expand_movmemsi_loop_known_size): New.
191 (nds32_expand_movmemsi_loop): New.
192 (nds32_expand_movmemsi_unroll): New.
193 (nds32_expand_movmemqi): Rename ...
194 (nds32_expand_movmemsi): ... to this.
195 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
196 (movmemsi): ... to this.
197 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
198 (nds32_expand_movmemsi): ... to this.
200 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
201 Monk Chiang <sh.chiang04@gmail.com>
202 Chung-Ju Wu <jasonwucj@gmail.com>
204 * config/nds32/nds32-protos.h
205 (nds32_expand_load_multiple): New arguments.
206 (nds32_expand_store_multiple): Ditto.
207 (nds32_valid_multiple_load_store): Rename ...
208 (nds32_valid_multiple_load_store_p): ... to this.
209 * config/nds32/nds32-memory-manipulation.c
210 (nds32_expand_load_multiple): Refine implementation.
211 (nds32_expand_store_multiple): Ditto.
212 * config/nds32/nds32-multiple.md
213 (load_multiple): Update nds32_expand_load_multiple interface.
214 (store_multiple): Update nds32_expand_store_multiple interface.
215 * config/nds32/nds32-predicates.c
216 (nds32_valid_multiple_load_store): Rename ...
217 (nds32_valid_multiple_load_store_p): ... to this and refine
219 * config/nds32/predicates.md
220 (nds32_load_multiple_and_update_address_operation): New predicate.
221 (nds32_store_multiple_and_update_address_operation): New predicate.
223 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
224 Chung-Ju Wu <jasonwucj@gmail.com>
226 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
227 (combo): New attribute.
228 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
230 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
232 * config/nds32/nds32.opt: Change -mcmodel= default value.
234 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
235 Monk Chiang <sh.chiang04@gmail.com>
236 Chung-Ju Wu <jasonwucj@gmail.com>
238 * config/nds32/constants.md (unspec_element): New enum.
239 * config/nds32/constraints.md (Umw): New constraint.
240 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
241 * config/nds32/nds32-intrinsic.md: Likewise.
242 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
243 (nds32_valid_smw_lwm_base_p): New.
244 (nds32_output_smw_single_word): New.
245 (nds32_output_lmw_single_word): New.
246 (nds32_expand_unaligned_load): New.
247 (nds32_expand_unaligned_store): New.
248 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
249 (nds32_output_smw_single_word): Declare.
250 (nds32_output_lmw_single_word): Declare.
251 (nds32_expand_unaligned_load): Declare.
252 (nds32_expand_unaligned_store): Declare.
253 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
254 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
255 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
256 NDS32_BUILTIN_UASTORE_DW.
257 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
260 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
261 Kito Cheng <kito.cheng@gmail.com>
262 Chung-Ju Wu <jasonwucj@gmail.com>
264 * config/nds32/nds32-intrinsic.c
265 (nds32_expand_builtin_null_ftype_reg): Delete.
266 (nds32_expand_builtin_reg_ftype_imm): Ditto.
267 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
268 (nds32_read_argument): New.
269 (nds32_legitimize_target): Ditto.
270 (nds32_legitimize_argument): Ditto.
271 (nds32_check_constant_argument): Ditto.
272 (nds32_expand_unop_builtin): Ditto.
273 (nds32_expand_unopimm_builtin): Ditto.
274 (nds32_expand_binop_builtin): Ditto.
275 (nds32_builtin_decl_impl): Ditto.
276 (builtin_description): Ditto.
277 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
278 (nds32_init_builtins_impl): Ditto.
279 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
280 (nds32_builtin_decl): New.
281 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
282 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
284 2018-03-02 Jeff Law <law@redhat.com>
286 * reorg.c (stop_search_p): Handle DEBUG_INSN.
287 (redundant_insn, fill_simple_delay_slots): Likewise.
288 (fill_slots_from_thread): Likewise.
289 * resource.c (mark_referenced_resources): Likewise.
290 (mark_set_resources, find_dead_or_set_registers): Likewise.
292 2018-03-02 Jakub Jelinek <jakub@redhat.com>
294 * substring-locations.h (format_warning_va): Formatting fix for
296 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
298 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
299 * substring-locations.c: Include intl.h.
300 (format_warning_va): Turned into small wrapper around
301 format_warning_n_va, renamed to ...
302 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
303 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
305 (format_warning_at_substring_n): New function.
306 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
307 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
308 format_warning_at_substring with just a shorter name instead of
309 const function pointer.
310 (fmtwarn_n): New function.
311 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
312 appropriate, get rid of all the fmtstr temporaries, move conditionals
313 with G_() wrapped string literals directly into fmtwarn arguments,
314 cast dir.len to (int), formatting fixes.
316 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
318 * doc/invoke.texi: Remove "Cilk Plus" references.
320 2018-03-02 Jakub Jelinek <jakub@redhat.com>
321 Richard Biener <rguenther@suse.de>
324 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
325 for error or warning attributes if CALL_FROM_THUNK_P is set.
328 2018-03-02 Jakub Jelinek <jakub@redhat.com>
331 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
332 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
335 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
336 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
338 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
339 instead of -1U in last predictors element's probability member.
341 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
344 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
345 arguments if they are comparable.
347 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
349 PR tree-optimization/84634
350 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
351 masks and masked_loop_p with a single loop_masks, making sure it's
352 null for bb vectorization.
354 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
356 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
357 (vect_analyze_data_ref_access): Use loop->safe_len rather than
358 loop->force_vectorize to check whether there is no alias.
360 2018-03-02 Jakub Jelinek <jakub@redhat.com>
363 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
365 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
367 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
368 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
369 instead of a loop around prev_real_insn.
370 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
374 * config/i386/i386.c (ix86_print_operand): Use conditional
375 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
378 2018-03-02 Richard Biener <rguenther@suse.de>
380 PR tree-optimization/84427
381 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
382 (bitmap_set_subtract_values): Rewrite to handle multiple
385 (prune_clobbered_mems): Likewise.
386 (phi_translate): Take edge instead of pred/phiblock.
387 (phi_translate_1): Likewise.
388 (phi_translate_set): Likewise. Insert all translated
389 exprs for a value into the set, keeping possibly multiple
390 expressions per value.
391 (compute_antic_aux): Adjust for phi_translate changes.
392 When intersecting union the expressions and prune those
393 not in the final value set, keeping possibly multiple
394 expressions per value. Do not use value-insertion
395 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
396 all expressions. Add verification that the value-sets
397 only shrink during iteration.
398 (compute_partial_antic_aux): Adjust for the phi_translate changes.
399 (do_pre_regular_insertion): Likewise.
400 (do_pre_partial_partial_insertion): Likewise.
402 2018-03-02 Richard Biener <rguenther@suse.de>
405 * config/darwin.c (saved_debug_info_level): New static global.
406 (darwin_asm_lto_start): Disable debug info generation for LTO out.
407 (darwin_asm_lto_end): Restore debug info generation settings.
409 2018-03-01 Martin Liska <mliska@suse.cz>
412 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
415 2018-03-01 Richard Biener <rguenther@suse.de>
418 * dwarf2out.c (gen_variable_die): Properly handle late VLA
419 type annotation with LTO when debug was disabled at compile-time.
421 2018-03-01 Matthew Fortune <mfortune@gmail.com>
423 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
425 (mips_final_postscan_insn): Likewise.
427 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
429 PR rtl-optimization/84528
430 * alias.c (init_alias_target): Add commentary.
431 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
432 a unique base value if the frame pointer is not eliminated
433 to the stack pointer.
435 2018-03-01 Tom de Vries <tom@codesourcery.com>
437 PR rtl-optimization/83327
438 * lra-int.h (hard_regs_spilled_into): Declare.
439 * lra.c (hard_regs_spilled_into): Define.
440 (init_reg_info): Init hard_regs_spilled_into.
441 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
442 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
443 (process_bb_lives): Handle hard_regs_spilled_into.
444 (lra_create_live_ranges_1): Before doing liveness propagation, clear
445 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
447 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
449 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
450 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
451 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
452 * config/rs6000/aix72.h: New file.
454 2018-02-28 Jakub Jelinek <jakub@redhat.com>
456 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
457 instead of warning_at with conditional singular and plural messages
461 * stor-layout.c (update_alignment_for_field): For
462 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
463 && !DECL_PACKED (field), do the alignment update, just use
464 only desired_align instead of MAX (type_align, desired_align)
466 (place_field): Don't do known_align < desired_align handling
467 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
468 is non-NULL, instead do it after rli->prev_field handling and
469 only if not within a bitfield word. For DECL_PACKED (field)
470 use type_align of BITS_PER_UNIT.
472 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
474 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
475 superfluous parentheses and trailing spaces.
477 2018-02-28 Richard Biener <rguenther@suse.de>
479 PR tree-optimization/84584
480 * graphite-scop-detection.c (scop_detection::add_scop): Discard
481 SCoPs with fake exit edge.
483 2018-02-28 Martin Liska <mliska@suse.cz>
486 * timevar.c (timer::print): Fix format to properly print 100%
489 2018-02-28 Richard Biener <rguenther@suse.de>
492 * genmatch.c (capture_info::walk_match): Do not mark
493 captured expressions without operands as expr_p given
494 they act more like predicates and should be subject to
495 "lost tail" side-effect preserving.
497 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
499 PR rtl-optimization/81611
500 * auto-inc-dec.c (attempt_change): Move dead note from
501 mem_insn if it's the next use of regno
502 (find_address): Take address use of reg holding
503 non-incremented value. Add parm to limit search to the named
505 (merge_in_block): Attempt to use a mem insn that is the next
506 use of the original regno.
508 2018-02-27 Martin Sebor <msebor@redhat.com>
511 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
512 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
514 2018-02-27 Martin Sebor <msebor@redhat.com>
517 * diagnostic-core.h (warning_n, error_n, inform_n): Change
518 n argument to unsigned HOST_WIDE_INT.
519 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
520 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
521 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
522 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
524 2018-02-27 Richard Biener <rguenther@suse.de>
526 PR tree-optimization/84512
527 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
528 Do not use the estimate returned from record_stmt_cost for
529 the scalar iteration cost but sum properly using add_stmt_cost.
531 2018-02-27 Richard Biener <rguenther@suse.de>
533 PR tree-optimization/84466
534 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
535 Adjust last change to less strictly validate use operands.
537 2018-02-27 Martin Liska <mliska@suse.cz>
539 PR gcov-profile/84548
540 * gcov.c (process_file): Allow partial overlap and consider it
541 also as group functions.
542 (output_lines): Properly calculate range of lines for a group.
544 2018-02-27 Martin Liska <mliska@suse.cz>
546 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
547 'ggc' suffixes. Change first column width.
548 (timer::print): Fix formatting of the column.
550 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
552 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
553 preserve inline entry blocks for the sake of debug inline
554 entry point markers alone.
555 (remove_unused_locals): Suggest in comments a better place to
556 force the preservation of inline entry blocks that are
557 otherwise unused, but do not preserve them.
559 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
561 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
563 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
566 * config/i386/constraints.md (Bs): Replace
567 ix86_indirect_branch_register with
568 TARGET_INDIRECT_BRANCH_REGISTER.
570 * config/i386/i386.md (indirect_jump): Likewise.
571 (tablejump): Likewise.
572 (*sibcall_memory): Likewise.
573 (*sibcall_value_memory): Likewise.
574 Peepholes of indirect call and jump via memory: Likewise.
575 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
576 (*sibcall_value_GOT_32): Likewise.
577 * config/i386/predicates.md (indirect_branch_operand): Likewise.
578 (GOT_memory_operand): Likewise.
579 (call_insn_operand): Likewise.
580 (sibcall_insn_operand): Likewise.
581 (GOT32_symbol_operand): Likewise.
582 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
584 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
586 PR rtl-optimization/83496
587 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
588 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
589 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
590 redundant insn, if any.
591 (relax_delay_slots): Likewise.
592 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
594 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
596 PR tree-optimization/83965
597 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
598 that grouped statements are part of a reduction chain. Return
599 true if the statement is not marked as a reduction itself but
601 (vect_recog_dot_prod_pattern): Don't check whether the statement
602 is part of a group here.
603 (vect_recog_sad_pattern): Likewise.
604 (vect_recog_widen_sum_pattern): Likewise.
606 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
609 * final.c (rest_of_clean_state): Also look for calls inside sequences.
611 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
614 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
616 (ix86_output_indirect_function_return): New prototype.
617 (ix86_split_simple_return_pop_internal): Likewise.
618 * config/i386/i386.c (indirect_return_via_cx): New.
619 (indirect_return_via_cx_bnd): Likewise.
620 (indirect_thunk_name): Handle return va CX_REG.
621 (output_indirect_thunk_function): Create alias for
622 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
623 (ix86_output_indirect_jmp): Remove the bool argument.
624 (ix86_output_indirect_function_return): New function.
625 (ix86_split_simple_return_pop_internal): Likewise.
626 * config/i386/i386.md (*indirect_jump): Don't pass false
627 to ix86_output_indirect_jmp.
628 (*tablejump_1): Likewise.
629 (simple_return_pop_internal): Change it to define_insn_and_split.
630 Call ix86_split_simple_return_pop_internal to split it for
632 (simple_return_indirect_internal): Call
633 ix86_output_indirect_function_return instead of
634 ix86_output_indirect_jmp.
636 2018-02-26 Jakub Jelinek <jakub@redhat.com>
639 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
640 memset and value initialization afterwards.
642 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
644 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
646 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
649 * common/config/aarch64/aarch64-common.c
650 (aarch_option_optimization_table[]): Switch
651 off fomit-frame-pointer
653 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
654 Chung-Ju Wu <jasonwucj@gmail.com>
656 * config/nds32/nds32-multiple.md (load_multiple): Disallow
658 (store_multiple): Ditto.
660 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
662 * config.gcc: Add --with-cpu support for nds32 target.
663 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
664 * config/nds32/nds32.opt: Add -mcpu= option.
666 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
668 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
669 isel=yes): Warn for these deprecated options.
671 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
673 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
674 ISA_2_5_MASKS_EMBEDDED.
676 2018-02-23 Jakub Jelinek <jakub@redhat.com>
678 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
679 p->max as pointers rather than using iterative_hash_expr.
681 2018-02-23 Carl Love <cel@us.ibm.com>
683 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
684 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
686 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
687 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
688 P8V_BUILTIN_VEC_VUNSIGNED2.
690 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
693 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
694 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
696 * lra-constraints.c (curr_insn_transform): Set up
697 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
699 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
700 * lra-eliminations.c (spill_pseudos): Ditto.
701 (process_insn_for_elimination): Ditto.
702 * lra-lives.c (reg_early_clobber_p): Use the new macros.
703 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
704 LRA_NON_CLOBBERED_ALT.
706 2018-02-22 Martin Sebor <msebor@redhat.com>
708 PR tree-optimization/84480
709 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
710 to maybe_diag_stxncpy_trunc. Call it.
711 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
712 from gimple_fold_builtin_strcpy. Print inlining stack.
713 (handle_builtin_stxncpy): Print inlining stack.
714 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
716 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
719 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
720 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
721 and -fcheck-pointer-bounds are used together.
722 (indirect_thunk_prefix): New enum.
723 (indirect_thunk_need_prefix): New function.
724 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
725 "_nt" instead of "_bnd" for NOTRACK prefix.
726 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
727 (output_indirect_thunk_function): Likewise.
729 (ix86_code_end): Update output_indirect_thunk_function calls.
730 (ix86_output_indirect_branch_via_reg): Replace
731 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
732 (ix86_output_indirect_branch_via_push): Likewise.
733 (ix86_output_function_return): Likewise.
734 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
735 incompatible with -fcf-protection=branch and
736 -fcheck-pointer-bounds.
738 2018-02-22 Steve Ellcey <sellcey@cavium.com>
741 * config/aarch64/aarch64.c (aarch64_print_address_internal):
742 Change gcc_assert call to output_operand_lossage.
744 2018-02-22 Steve Ellcey <sellcey@cavium.com>
746 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
748 2018-02-22 DJ Delorie <dj@redhat.com>
749 Sebastian Perta <sebastian.perta@renesas.com>
750 Oleg Endo <olegendo@gcc.gnu.org>
752 * config/rx/rx.c (rx_rtx_costs): New function.
753 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
755 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
757 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
759 2018-02-22 Martin Liska <mliska@suse.cz>
762 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
763 Add "native" as a possible value.
765 2018-02-22 Martin Liska <mliska@suse.cz>
768 * config/i386/i386.c (ix86_option_override_internal):
769 Add "native" as a possible value for -march and -mtune.
771 2018-02-22 Jakub Jelinek <jakub@redhat.com>
774 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
775 to all type variants.
777 PR tree-optimization/84503
778 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
779 width as info->bitpos + info->bitsize - start.
780 (merged_store_group::merge_overlapping): Simplify width computation.
781 (check_no_overlap): New function.
782 (imm_store_chain_info::try_coalesce_bswap): Compute expected
783 start + width and last_order of the group, fail if check_no_overlap
785 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
786 to group if check_no_overlap fails.
788 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
790 * config/rs6000/altivec.md: Delete contraint arguments to
791 define_expand, define_split, and define_peephole2, and in
792 define_insn_and_split if always unused.
793 * config/rs6000/darwin.md: Ditto.
794 * config/rs6000/dfp.md: Ditto.
795 * config/rs6000/rs6000.md: Ditto.
796 * config/rs6000/sync.md: Ditto.
797 * config/rs6000/vector.md: Ditto.
798 * config/rs6000/vsx.md: Ditto.
800 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
802 * config/rs6000/altivec.md: Write output control strings as braced
803 blocks instead of double-quoted strings.
804 * config/rs6000/darwin.md: Ditto.
805 * config/rs6000/rs6000.md: Ditto.
806 * config/rs6000/vector.md: Ditto.
807 * config/rs6000/vsx.md: Ditto.
809 2018-02-21 Jason Merrill <jason@redhat.com>
811 PR c++/84314 - ICE with templates and fastcall attribute.
812 * attribs.c (build_type_attribute_qual_variant): Remove assert.
814 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
816 * ipa-cp.c (determine_versionability): Fix comment typos.
818 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
821 * ipa-cp.c (determine_versionability): Do not version functions caling
824 2018-02-21 Martin Liska <mliska@suse.cz>
827 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
828 Add "native" as a possible value.
829 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
830 the macro when native cpu detection is available.
832 2018-02-21 Martin Liska <mliska@suse.cz>
835 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
836 Add "native" as a possible value.
837 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
838 when native cpu detection is available.
840 2018-02-21 Jakub Jelinek <jakub@redhat.com>
841 Martin Sebor <msebor@redhat.com>
843 PR tree-optimization/84478
844 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
846 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
847 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
848 support which is conservatively correct, for 2 only stay conservative
849 for maxlen. Formatting and comment capitalization fixes. Add STRICT
850 argument to the 2 argument get_range_strlen, adjust 6 arg
851 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
853 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
854 (gimple_fold_builtin_strlen): Pass true as last argument to
857 2018-02-20 Martin Sebor <msebor@redhat.com>
860 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
861 (builtin_memref::set_base_and_offset): Same. Handle inner references.
862 (builtin_memref::builtin_memref): Factor out parts into
863 set_base_and_offset and call it.
865 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
868 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
869 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
870 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
871 search at the associated MODE_INT.
873 2018-02-20 Jeff Law <law@redhat.com>
876 PR tree-optimization/81592
878 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
879 for range data rather than using global data.
880 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
881 range data rather than using global data.
882 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
883 pass it to children as needed.
884 (struct directive::fmtresult): Similarly.
885 (struct directive::set_width): Similarly.
886 (struct directive::set_precision): Similarly.
887 (format_integer, format_directive, parse_directive): Similarly.
888 (format_none): Accept unnamed vr_values parameter.
889 (format_percent, format_floating, format_character): Similarly.
890 (format_string, format_plain): Similarly.
891 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
892 the EVRP range analyzer for range data rather than using global data.
893 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
894 gimple-ssa-evrp-analyze.h
895 (class sprintf_dom_walker): Add after_dom_children member function.
896 Add evrp_range_analyzer member.
897 (sprintf_dom_walker::before_dom_children): Call into the EVRP
898 range analyzer as needed.
899 (sprintf_dom_walker::after_dom_children): New member function.
900 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
902 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
903 (evrp_range_analyzer::pop_to_marker): Likewise.
905 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
907 PR tree-optimization/84419
908 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
909 with the required type if its current type is compatible but
912 2018-02-20 Jakub Jelinek <jakub@redhat.com>
915 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
918 2018-02-20 Martin Liska <mliska@suse.cz>
921 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
922 possible values if we don't have a hint.
924 2018-02-20 Martin Liska <mliska@suse.cz>
928 * final.c (shorten_branches): Build align_tab array with one
930 * opts.c (finish_options): Add alignment option limit check.
931 (MAX_CODE_ALIGN): Likewise.
932 (MAX_CODE_ALIGN_VALUE): Likewise.
933 * doc/invoke.texi: Document maximum allowed option value for
934 all -falign-* options.
936 2018-02-19 Jakub Jelinek <jakub@redhat.com>
939 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
940 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
941 * var-tracking.c (emit_note_insn_var_location): Remove all references
942 to NOTE_INSN_CALL_ARG_LOCATION.
943 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
944 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
945 Use copy_rtx_if_shared.
946 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
947 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
948 (dwarf2out_var_location): Remove handling of
949 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
951 * final.c (final_scan_insn): Remove all references to
952 NOTE_INSN_CALL_ARG_LOCATION.
953 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
954 before dumping final insns.
955 * except.c (emit_note_eh_region_end): Remove all references to
956 NOTE_INSN_CALL_ARG_LOCATION.
957 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
958 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
959 * config/arc/arc.c (hwloop_optimize): Likewise.
960 * config/arm/arm.c (create_fix_barrier): Likewise.
961 * config/s390/s390.c (s390_chunkify_start): Likewise.
962 * config/sh/sh.c (find_barrier): Likewise.
963 * config/i386/i386.c (rest_of_insert_endbranch,
964 ix86_seh_fixup_eh_fallthru): Likewise.
965 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
966 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
967 * config/frv/frv.c (frv_function_prologue): Likewise.
968 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
970 (note_outside_basic_block_p): Remove all references to
971 NOTE_INSN_CALL_ARG_LOCATION.
972 * gengtype.c (adjust_field_rtx_def): Likewise.
973 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
975 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
976 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
979 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
982 PR tree-optimization/84452
983 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
984 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
987 2018-02-19 Martin Liska <mliska@suse.cz>
990 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
992 2018-02-19 Martin Liska <mliska@suse.cz>
993 Richard Sandiford <richard.sandiford@linaro.org>
995 PR tree-optimization/82491
996 * gimple-fold.c (get_base_constructor): Make earlier bail out
999 2018-02-19 Carl Love <cel@us.ibm.com>
1001 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1002 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1004 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1005 P8V_BUILTIN_VEC_NEG.
1007 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1009 * config/rl78/rl78.md (movdf): New define expand.
1011 2018-02-19 Martin Liska <mliska@suse.cz>
1014 * doc/invoke.texi: Fix typo.
1015 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1017 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1019 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1020 handle rs6000_single_float and rs6000_double_float specially for
1023 2018-02-16 Jeff Law <law@redhat.com>
1025 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1028 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1031 * ipa-inline.c (inline_small_functions): Fix a typo.
1033 2018-02-16 Nathan Sidwell <nathan@acm.org>
1035 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1037 2018-02-16 Carl Love <cel@us.ibm.com>
1039 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1040 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1041 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1042 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1043 expansion to P8V_BUILTIN_VEC_FLOAT2.
1045 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1047 PR rtl-optimization/70023
1048 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1049 src_regno into account.
1051 2018-02-16 Carl Love <cel@us.ibm.com>
1053 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1054 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1055 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1056 * config/rs6000/rs6000.c: Remove case statements for
1057 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1058 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1059 and P9V_BUILTIN_VEC_VINSERT4B.
1060 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1061 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1062 * config/rs6000/vsx.md:
1063 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1066 2018-02-16 Carl Love <cel@us.ibm.com>
1068 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1070 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1072 * config/rs6000/rs6000-c.c: Add the definitions for
1073 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1074 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1075 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1076 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1077 definition for insert4b and define insn *insert3b_internal.
1078 * doc/extend.texi: Add documentation for vec_extract4b.
1080 2018-02-16 Nathan Sidwell <nathan@acm.org>
1082 * doc/extend.texi (Backwards Compatibility): Mention friend
1083 injection. Note for-scope is deprecated.
1084 * doc/invoke.texi (-ffriend-injection): Deprecate.
1086 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1088 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1089 that moved to I2, also allow destinations that are a paradoxical
1090 subreg (instead of a normal reg).
1092 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1095 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1098 2018-02-16 Richard Biener <rguenther@suse.de>
1100 PR tree-optimization/84037
1101 PR tree-optimization/84016
1103 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1104 Adjust vec_construct for the fact we need additional higher latency
1105 128bit inserts for AVX256 and AVX512 vector builds.
1106 (ix86_add_stmt_cost): Scale vector construction cost for
1109 2018-02-16 Richard Biener <rguenther@suse.de>
1111 PR tree-optimization/84417
1112 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1113 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1114 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1116 2018-02-16 Martin Liska <mliska@suse.cz>
1119 * internal-fn.def (ASAN_CHECK): Set proper flags.
1120 (ASAN_MARK): Likewise.
1122 2018-02-16 Julia Koval <julia.koval@intel.com>
1124 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1125 from PTA_CANNONLAKE.
1127 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1130 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1131 Use ++iter rather than iter++ for std::list iterators.
1132 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1133 defer deleting them until all nodes in the forest are processed. Do
1134 free even leaf nodes. Change to_process into auto_vec.
1137 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1138 * vec.h (vec_default_construct): Use memset instead of placement new
1139 if BROKEN_VALUE_INITIALIZATION is defined.
1140 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1141 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1144 PR rtl-optimization/83723
1145 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1146 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1147 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1149 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1151 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1153 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1155 PR rtl-optimization/81443
1156 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1157 from inner REGs to paradoxical SUBREGs.
1159 2018-02-16 Richard Biener <rguenther@suse.de>
1161 PR tree-optimization/84399
1162 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1163 For operands we can analyze at their definition make sure we can
1164 analyze them at each use as well.
1166 2018-02-16 Richard Biener <rguenther@suse.de>
1168 PR tree-optimization/84190
1169 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1170 volatile accesses if the decl isn't volatile.
1172 2018-02-15 Jason Merrill <jason@redhat.com>
1174 PR c++/84314 - ICE with templates and fastcall attribute.
1175 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1176 TYPE_CANONICAL on an existing type.
1178 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1180 PR tree-optimization/84383
1181 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1182 dstoff nor call operand_equal_p if dstbase is NULL.
1184 PR tree-optimization/84334
1185 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1186 also a CONSTANT_CLASS_P, punt.
1188 2018-02-14 Jim Wilson <jimw@sifive.com>
1190 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1191 first SMALL_OPERAND check. New local min_second_step. Move assert
1192 to where locals are set. Add TARGET_RVC support.
1193 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1195 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1197 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1199 2018-02-14 Martin Sebor <msebor@redhat.com>
1201 PR tree-optimization/83698
1202 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1203 arrays constrain the offset range to their bounds.
1204 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1205 (builtin_access::overlap): Avoid setting the size of overlap if it's
1207 (maybe_diag_overlap): Also consider arrays when deciding what values
1208 of offsets to include in diagnostics.
1210 2018-02-14 Martin Sebor <msebor@redhat.com>
1213 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1214 that correspond to the kind of a declaration.
1216 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1219 * config/pa/pa.md: Load address of PIC label using the linkage table
1220 if the label is nonlocal.
1222 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1224 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1225 warning message if user requests -maltivec=be.
1226 * doc/invoke.texi: Document deprecation of -maltivec=be.
1228 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1231 * config/rs6000/rs6000-c.c: Update definitions for
1232 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1233 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1235 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1238 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1239 add _get_ssp intrinsics. Remove argument from
1240 __builtin_ia32_rdssp[d|q].
1241 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1242 * config/i386/i386-builtin.def: Remove argument from
1243 __builtin_ia32_rdssp[d|q].
1244 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1245 ix86_expand_special_args_builtin for _rdssp[d|q].
1246 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1247 Clear register before usage.
1248 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1249 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1251 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1253 PR tree-optimization/84357
1254 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1255 operand 1 of an ARRAY_REF too.
1257 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1260 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1261 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1263 (set_of_reg): New struct.
1264 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1265 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1266 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1268 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1269 Split into bitclr, bitset, bitinvert patterns if appropriate.
1270 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1271 use rx_fuse_in_memory_bitop.
1272 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1273 to named insn, correct maximum insn length.
1275 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1278 * machmode.def: Define a complex mode for PARTIAL_INT.
1279 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1281 * doc/rtl.texi: Document CSPImode.
1282 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1284 (msp430_hard_regno_nregs_with_padding): Likewise.
1286 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1289 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1291 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1293 PR rtl-optimization/84169
1294 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1295 we generated a parallel as new i3 and we split that to new i2 and i3
1296 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1297 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1298 those to i2, not i1. Partially rewrite this scan code.
1300 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1303 * stor-layout.c (place_field): For variable length fields, adjust
1304 offset_align afterwards not just based on the field's alignment,
1305 but also on the size.
1308 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1309 of exps and logs in the use_exp2 case.
1311 2018-02-13 Jeff Law <law@redhat.com>
1313 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1316 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1319 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1323 * common.opt (gas-loc-support, gas-locview-support): New.
1324 (ginline-points, ginternal-reset-location-views): New.
1325 * doc/invoke.texi: Document them. Use @itemx where intended.
1326 (gvariable-location-views): Adjust.
1327 * target.def (reset_location_view): New.
1328 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1329 (TARGET_RESET_LOCATION_VIEW): New.
1330 * doc/tm.texi: Rebuilt.
1331 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1332 (dwarf2out_default_as_locview_support): New.
1333 (output_asm_line_debug_info): Use option variables.
1334 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1335 (output_loc_list): Likewise.
1336 (add_high_low_attributes): Check option variables.
1337 Don't output entry view attribute in strict mode.
1338 (gen_inlined_subroutine_die): Check option variables.
1339 (dwarf2out_inline_entry): Likewise.
1340 (init_sections_and_labels): Likewise.
1341 (dwarf2out_early_finish): Likewise.
1342 (maybe_reset_location_view): New, from...
1343 (dwarf2out_var_location): ... here. Call it.
1344 * debug.h (dwarf2out_default_as_loc_support): Declare.
1345 (dwarf2out_default_as_locview_support): Declare.
1346 * hooks.c (hook_int_rtx_insn_0): New.
1347 * hooks.h (hook_int_rtx_insn_0): Declare.
1348 * toplev.c (process_options): Take -gas-loc-support and
1349 -gas-locview-support from dwarf2out. Enable
1350 -gvariable-location-views by default only with locview
1351 assembler support. Enable -ginternal-reset-location-views by
1352 default only if the target defines the corresponding hook.
1353 Enable -ginline-points by default if location views are
1354 enabled; force it disabled if statement frontiers are
1356 * tree-inline.c (expand_call_inline): Check option variables.
1357 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1359 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1361 PR tree-optimization/84321
1362 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1363 handling. Also check whether the anti-range contains any values
1364 that satisfy the mask; switch to a VR_RANGE if not.
1366 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1369 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1371 2018-02-13 Martin Jambor <mjambor@suse.cz>
1374 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1375 of call statements, also set location of a load to a temporary.
1377 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1379 * config/rl78/rl78.c (add_vector_labels): New function.
1380 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1381 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1382 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1383 which checks that no arguments are passed.
1384 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1385 * doc/extend.texi: Documentation for the new attribute.
1387 2018-02-13 Andreas Schwab <schwab@suse.de>
1389 * config/riscv/linux.h (CPP_SPEC): Define.
1391 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1394 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1395 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1396 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1397 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1398 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1399 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1400 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1401 temporarily for AES and PCLMUL builtins.
1403 PR tree-optimization/84339
1404 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1405 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1409 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1410 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1411 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1413 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1415 * omp-simd-clone.h: New file.
1416 * omp-simd-clone.c: Include omp-simd-clone.h.
1417 (expand_simd_clones): No longer static.
1418 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1419 cgraph.h and omp-simd-clone.h.
1420 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1421 (vect_recog_widen_shift_pattern): Formatting fix.
1422 (vect_pattern_recog_1): Don't check optab for calls.
1425 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1426 operands[2] into a REG before using gen_lowpart on it.
1428 2018-02-12 Jeff Law <law@redhat.com>
1431 * config/sh/sh.c (find_barrier): Consider a sibling call
1434 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1435 successfully back substituting a reg.
1437 2018-02-12 Richard Biener <rguenther@suse.de>
1439 PR tree-optimization/84037
1440 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1441 parameter, move visited init to caller.
1442 (vect_slp_analyze_operations): Separate cost from validity
1443 check, initialize visited once for all instances.
1444 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1446 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1448 (vect_model_promotion_demotion_cost): Likewise.
1449 (vectorizable_bswap): Guard cost modeling with !slp_node
1450 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1452 (vectorizable_call): Likewise.
1453 (vectorizable_conversion): Likewise.
1454 (vectorizable_assignment): Likewise.
1455 (vectorizable_shift): Likewise.
1456 (vectorizable_operation): Likewise.
1457 (vectorizable_store): Likewise.
1458 (vectorizable_load): Likewise.
1459 (vectorizable_condition): Likewise.
1460 (vectorizable_comparison): Likewise.
1462 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1465 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1466 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1467 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1469 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1472 * params.def (inline-min-speedup): Increase from 8 to 15.
1473 (max-inline-insns-auto): Decrease from 40 to 30.
1474 * ipa-split.c (consider_split): Add some buffer for function to
1475 be considered inlining candidate.
1476 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1479 2018-02-12 Richard Biener <rguenther@suse.de>
1481 PR tree-optimization/84037
1482 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1483 matched stmts if we cannot swap the non-matched ones.
1485 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1487 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1488 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1489 _mm_maskz_scalef_round_ss): New intrinsics.
1490 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1491 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1492 __builtin_ia32_scalefss_round): Remove.
1493 (__builtin_ia32_scalefsd_mask_round,
1494 __builtin_ia32_scalefss_mask_round): New intrinsics.
1495 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1496 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1497 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1498 "<round_constraint>")): Changed to ...
1499 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1500 "<round_scalar_constraint>")): ... this.
1501 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1502 %0, %1, %2<round_op3>}"): Changed to ...
1503 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1504 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1505 %2<round_scalar_mask_op3>}"): ... this.
1506 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1508 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1510 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1511 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1512 (_mm_maskz_sqrt_round_ss): New intrinsics.
1513 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1514 (__builtin_ia32_sqrtsd_mask_round)
1515 (__builtin_ia32_sqrtss_mask_round): New builtins.
1516 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1517 (__builtin_ia32_sqrtss_round): Remove.
1518 (__builtin_ia32_sqrtsd_mask_round)
1519 (__builtin_ia32_sqrtss_mask_round): New builtins.
1520 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1521 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1522 ((match_operand:VF_128 1 "vector_operand"
1523 "xBm,<round_constraint>")): Changed to ...
1524 ((match_operand:VF_128 1 "vector_operand"
1525 "xBm,<round_scalar_constraint>")): ... this.
1526 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1527 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1528 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1529 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1530 %<iptr>1<round_scalar_mask_op3>}): ... this.
1531 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1532 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1534 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1537 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1538 Cast vec_cmpeq result to correct type.
1539 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1540 Cast vec_cmpgt result to correct type.
1542 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1544 * final.c (final_scan_insn_1): Renamed from...
1545 (final_scan_insn): ... this. New wrapper, to recover
1546 seen from the outermost call in recursive ones.
1547 * config/sparc/sparc.c (output_return): Drop seen from call.
1548 (output_sibcall): Likewise.
1549 * config/visium/visium.c (output_branch): Likewise.
1551 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1553 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1556 2018-02-10 Alan Modra <amodra@gmail.com>
1559 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1560 Specify LR as an input.
1562 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1565 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1566 remove_member_access_dummy_vars): New functions.
1567 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1568 lower_omp_1, execute_lower_omp): Use them.
1570 PR rtl-optimization/84308
1571 * shrink-wrap.c (spread_components): Release todo vector.
1573 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1575 PR rtl-optimization/57193
1576 * ira-color.c (struct allocno_color_data): Add member
1577 conflict_allocno_hard_prefs.
1578 (update_conflict_allocno_hard_prefs): New.
1579 (bucket_allocno_compare_func): Add a preference based on
1580 conflict_allocno_hard_prefs.
1581 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1582 (color_allocnos): Remove a dead code. Initiate
1583 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1585 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1588 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1589 constraint from =wa to wa. Avoid a subreg on the output operand,
1590 instead use a pseudo and subreg it in a move.
1591 (p9_xxbrd_<mode>): Changed to ...
1592 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1593 (p9_xxbrd_v2df): New expander.
1594 (p9_xxbrw_<mode>): Changed to ...
1595 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1596 (p9_xxbrw_v4sf): New expander.
1598 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1600 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1602 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1605 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1606 multiply in 32-bit mode.
1607 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1608 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1611 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1613 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1614 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1615 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1616 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1618 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1621 * dwarf2out.c (is_trivial_indirect_ref): New function.
1622 (dwarf2out_late_global_decl): Do not generate a location
1623 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1624 and that are not defined in the current unit.
1626 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1628 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1629 instead of a libcall for UNORDERED.
1631 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1634 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1635 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1637 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1640 * config/s390/s390.c (s390_set_current_function): Invoke
1641 s390_indirect_branch_settings also if fndecl didn't change.
1643 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1645 * config/rs6000/rs6000.md (blockage): Set length to zero.
1647 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1649 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1651 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1654 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1655 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1659 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1660 PARALLEL incoming that failed vt_get_decl_and_offset check.
1663 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1664 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1666 (get_variable_section): For decls in named .bss* sections pass true as
1667 second argument to bss_initializer_p.
1669 2018-02-09 Marek Polacek <polacek@redhat.com>
1670 Jakub Jelinek <jakub@redhat.com>
1673 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1674 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1675 Sync some changes from cxx_fold_indirect_ref.
1677 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1679 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1681 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1682 (BLOCK_INLINE_ENTRY_LABEL): New.
1683 (dwarf2out_var_location): Disregard inline entry markers.
1684 (inline_entry_data): New struct.
1685 (inline_entry_data_hasher): New hashtable type.
1686 (inline_entry_data_hasher::hash): New.
1687 (inline_entry_data_hasher::equal): New.
1688 (inline_entry_data_table): New variable.
1689 (add_high_low_attributes): Add DW_AT_entry_pc and
1690 DW_AT_GNU_entry_view attributes if a pending entry is found
1691 in inline_entry_data_table. Add old entry_pc attribute only
1692 if debug nonbinding markers are disabled.
1693 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1694 markers are enabled.
1695 (block_within_block_p, dwarf2out_inline_entry): New.
1696 (dwarf2out_finish): Check that no entries remained in
1697 inline_entry_data_table.
1698 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1699 (final_scan_insn, notice_source_line): Likewise.
1700 (rest_of_clean_state): Skip inline entry markers.
1701 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1703 * gimple.c (gimple_build_debug_inline_entry): New.
1704 * gimple.h (enum gimple_debug_subcode): Add
1705 GIMPLE_DEBUG_INLINE_ENTRY.
1706 (gimple_build_debug_inline_entry): Declare.
1707 (gimple_debug_inline_entry_p): New.
1708 (gimple_debug_nonbind_marker_p): Adjust.
1709 * insn-notes.def (INLINE_ENTRY): New.
1710 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1711 inline entry marker notes.
1712 (print_insn): Likewise.
1713 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1714 (INSN_DEBUG_MARKER_KIND): Likewise.
1715 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1716 * tree-inline.c (expand_call_inline): Build and insert
1717 debug_inline_entry stmt.
1718 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1719 inline entry blocks early, if nonbind markers are enabled.
1720 (dump_scope_block): Dump fragment info.
1721 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1722 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1723 (gimple_build_debug_inline_entry): New.
1724 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1725 Enable/disable inline entry points too.
1726 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1727 (DEBUG_INSN): Describe inline entry markers.
1729 * common.opt (gvariable-location-views): New.
1730 (gvariable-location-views=incompat5): New.
1731 * config.in: Rebuilt.
1732 * configure: Rebuilt.
1733 * configure.ac: Test assembler for view support.
1734 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1735 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1736 * dwarf2out.c (var_loc_view): New typedef.
1737 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1738 (dwarf2out_locviews_in_attribute): New.
1739 (dwarf2out_locviews_in_loclist): New.
1740 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1741 (enum dw_line_info_opcode): Add LI_adv_address.
1742 (struct dw_line_info_table): Add view.
1743 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1744 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1745 (zero_view_p): New variable.
1746 (ZERO_VIEW_P): New macro.
1747 (output_asm_line_debug_info): New.
1748 (struct var_loc_node): Add view.
1749 (add_AT_view_list, AT_loc_list): New.
1750 (add_var_loc_to_decl): Add view param. Test it against last.
1751 (new_loc_list): Add view params. Record them.
1752 (AT_loc_list_ptr): Handle loc and view lists.
1753 (view_list_to_loc_list_val_node): New.
1754 (print_dw_val): Handle dw_val_class_view_list.
1755 (size_of_die): Likewise.
1756 (value_format): Likewise.
1757 (loc_list_has_views): New.
1758 (gen_llsym): Set vl_symbol too.
1759 (maybe_gen_llsym, skip_loc_list_entry): New.
1760 (dwarf2out_maybe_output_loclist_view_pair): New.
1761 (output_loc_list): Output view list or entries too.
1762 (output_view_list_offset): New.
1763 (output_die): Handle dw_val_class_view_list.
1764 (output_dwarf_version): New.
1765 (output_compilation_unit_header): Use it.
1766 (output_skeleton_debug_sections): Likewise.
1767 (output_rnglists, output_line_info): Likewise.
1768 (output_pubnames, output_aranges): Update version comments.
1769 (output_one_line_info_table): Output view numbers in asm comments.
1770 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1771 Call maybe_gen_llsym.
1772 (loc_list_from_tree_1): Adjust.
1773 (add_AT_location_description): Create view list attribute if
1774 needed, check it's absent otherwise.
1775 (convert_cfa_to_fb_loc_list): Adjust.
1776 (maybe_emit_file): Call output_asm_line_debug_info for test.
1777 (dwarf2out_var_location): Reset views as needed. Precompute
1778 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1779 attribute. Set view.
1780 (new_line_info_table): Reset next view.
1781 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1782 (dwarf2out_source_line): Likewise. Output view resets and labels to
1783 the assembler, or select appropriate line info opcodes.
1784 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1785 (optimize_string_length): Catch it. Adjust.
1786 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1787 dw_val_class_view_list, and remove it if no longer needed.
1788 (hash_loc_list): Hash view numbers.
1789 (loc_list_hasher::equal): Compare them.
1790 (optimize_location_lists): Check whether a view list symbol is
1791 needed, and whether the locview attribute is present, and
1792 whether they match. Remove the locview attribute if no longer
1794 (index_location_lists): Call skip_loc_list_entry for test.
1795 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1796 Use output_dwarf_version.
1797 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1798 (struct dw_val_node): Add val_view_list.
1799 * final.c (SEEN_NEXT_VIEW): New.
1800 (set_next_view_needed): New.
1801 (clear_next_view_needed): New.
1802 (maybe_output_next_view): New.
1803 (final_start_function): Rename to...
1804 (final_start_function_1): ... this. Take pointer to FIRST,
1805 add SEEN parameter. Emit param bindings in the initial view.
1806 (final_start_function): Reintroduce SEEN-less interface.
1807 (final): Rename to...
1808 (final_1): ... this. Take SEEN parameter. Output final pending
1809 next view at the end.
1810 (final): Reintroduce seen-less interface.
1811 (final_scan_insn): Output pending next view before switching
1812 sections or ending a block. Mark the next view as needed when
1813 outputting variable locations. Notify debug backend of section
1814 changes, and of location view changes.
1815 (rest_of_handle_final): Adjust.
1816 * toplev.c (process_options): Autodetect value for debug variable
1817 location views option. Warn on incompat5 without -gdwarf-5.
1818 * doc/invoke.texi (gvariable-location-views): New.
1819 (gvariable-location-views=incompat5): New.
1820 (gno-variable-location-views): New.
1822 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1824 PR tree-optimization/84136
1825 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1826 that the result of find_edge is non-NULL.
1828 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1831 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1832 storing integer register in SImode. Fix cost of 256 and 512
1833 byte aligned SSE register store.
1835 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1837 * config/i386/i386.c (ix86_multiplication_cost): Fix
1838 multiplication cost for TARGET_AVX512DQ.
1840 2018-02-08 Marek Polacek <polacek@redhat.com>
1842 PR tree-optimization/84238
1843 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1846 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1848 PR tree-optimization/84265
1849 * tree-vect-stmts.c (vectorizable_store): Don't treat
1850 VMAT_CONTIGUOUS accesses as grouped.
1851 (vectorizable_load): Likewise.
1853 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1855 PR tree-optimization/81635
1856 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1857 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1858 (test_round_for_mask): New functions.
1859 (wide_int_cc_tests): Call test_round_for_mask.
1860 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1861 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1862 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1863 range returned by get_range_info.
1865 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1868 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1869 * symtab.c: Include builtins.h
1870 (symtab_node::output_to_lto_symbol_table_p): Move here
1871 from lto-streamer-out.c:output_symbol_p.
1872 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1873 (output_symbol_p): Move all logic to symtab.c
1874 (produce_symtab): Update.
1876 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1878 * config/s390/s390-opts.h (enum indirect_branch): Define.
1879 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1880 (s390_indirect_branch_via_thunk)
1881 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1882 (enum s390_indirect_branch_type): Define.
1883 * config/s390/s390.c (struct s390_frame_layout, struct
1884 machine_function): Remove.
1885 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1886 (indirect_branch_table_label_no, indirect_branch_table_name):
1888 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1889 (enum s390_indirect_branch_option): Define.
1890 (s390_return_addr_from_memory): New function.
1891 (s390_handle_string_attribute): New function.
1892 (s390_attribute_table): Add new attribute handler.
1893 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1894 (s390_indirect_branch_via_thunk): New function.
1895 (s390_indirect_branch_via_inline_thunk): New function.
1896 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1897 sibling call optimization for non z10 compiles.
1898 (s390_emit_call): Force indirect branch target to be a single
1899 register. Add r1 clobber for non-z10 compiles.
1900 (s390_emit_epilogue): Emit return jump via return_use expander.
1901 (s390_reorg): Handle JUMP_INSNs as execute targets.
1902 (s390_option_override_internal): Perform validity checks for the
1903 new command line options.
1904 (s390_indirect_branch_attrvalue): New function.
1905 (s390_indirect_branch_settings): New function.
1906 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1907 (s390_output_indirect_thunk_function): New function.
1908 (s390_code_end): Implement target hook.
1909 (s390_case_values_threshold): Implement target hook.
1910 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1912 * config/s390/s390.h (struct s390_frame_layout)
1913 (struct machine_function): Move here from s390.c.
1914 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1915 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1916 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1917 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1918 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1919 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1920 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1921 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1922 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1923 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1924 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1925 (mnemonic attribute): Add values which aren't recognized
1927 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1928 pattern for branch conversion. Fix mnemonic attribute.
1929 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1930 indirect branch via thunk if requested.
1931 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1932 ("*indirect_jump"): Disable for branch conversion using out of
1934 ("indirect_jump_via_thunk<mode>_z10")
1935 ("indirect_jump_via_thunk<mode>")
1936 ("indirect_jump_via_inlinethunk<mode>_z10")
1937 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1938 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1939 ("casesi_jump_via_inlinethunk<mode>_z10")
1940 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1941 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1942 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1943 ("*indirect2_jump"): Disable for branch conversion.
1944 ("casesi_jump"): Turn into expander and expand patterns for branch
1946 ("return_use"): New expander.
1947 ("*return"): Emit return via thunk and rename it to ...
1948 ("*return<mode>"): ... this one.
1949 * config/s390/s390.opt: Add new options and and enum for the
1952 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1954 * lra-constraints.c (match_reload): Unconditionally use
1955 gen_lowpart_SUBREG, rather than selecting between that
1956 and equivalent gen_rtx_SUBREG code.
1958 2018-02-08 Richard Biener <rguenther@suse.de>
1960 PR tree-optimization/84233
1961 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1962 changed flag instead of boguously re-using phi_inserted.
1964 2018-02-08 Martin Jambor <mjambor@suse.cz>
1966 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1967 static local variables.
1969 2018-02-08 Richard Biener <rguenther@suse.de>
1971 PR tree-optimization/84278
1972 * tree-vect-stmts.c (vectorizable_store): When looking for
1973 smaller vector types to perform grouped strided loads/stores
1974 make sure the mode is supported by the target.
1975 (vectorizable_load): Likewise.
1977 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1979 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1980 Increase LDP/STP opportunities by adding adjacent callee-saves.
1982 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1984 PR rtl-optimization/84068
1985 PR rtl-optimization/83459
1986 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1988 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1990 PR tree-optimization/84224
1991 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1992 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1995 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1998 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1999 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2000 count, remove one USE.
2002 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2004 * doc/install.texi (Configuration): Document the
2005 --with-long-double-format={ibm,ieee} PowerPC configuration
2009 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2010 Convert from define_expand to be define_insn_and_split. Rework
2011 float/double/_Float128 conversions to QI/HI/SImode to work with
2012 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2013 conversions to QI/HImode types did a store and then a load to
2014 truncate the value. For conversions to VSX registers, don't split
2015 the insn, instead emit the code directly. Use the code iterator
2016 any_fix to combine signed and unsigned conversions.
2017 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2018 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2019 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2020 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2021 (fix_<mode>di2_hw): Likewise.
2022 (fixuns_<mode>di2_hw): Likewise.
2023 (fix_<mode>si2_hw): Likewise.
2024 (fixuns_<mode>si2_hw): Likewise.
2025 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2026 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2027 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2028 fix<uns>_trunc<SFDF:mode>si2_p8.
2029 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2031 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2032 (fix<uns>_<mode>_mem): Likewise.
2033 (fctiw<u>z_<mode>_mem): Likewise.
2034 (fix<uns>_<mode>_mem): Likewise.
2035 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2036 the register allocator from doing a direct move to the GPRs to do
2037 a store, and instead use the ISA 3.0 store byte/half-word from
2038 vector register instruction. For IEEE 128-bit floating point,
2039 also optimize stores of 32-bit ints.
2040 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2042 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2044 * genextract.c (push_pathstr_operand): New function to support
2046 (walk_rtx): Call push_pathstr_operand.
2047 (print_path): Support [a-zA-Z].
2049 2018-02-07 Richard Biener <rguenther@suse.de>
2051 PR tree-optimization/84037
2052 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2053 (cse_and_gimplify_to_preheader): Declare.
2054 (vect_get_place_in_interleaving_chain): Likewise.
2055 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2057 (_loop_vec_info::~_loop_vec_info): Delete it.
2058 (cse_and_gimplify_to_preheader): New function.
2059 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2060 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2061 (vectorizable_load): Likewise. For grouped stores always base
2062 the IV on the first element.
2063 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2064 condition before gimplifying.
2066 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2068 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2069 *DIV_EXPR and *MOD_EXPR.
2071 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2074 * config/i386/i386.c (ix86_option_override_internal): Mask out
2075 the CF_SET bit when checking -fcf-protection.
2077 2018-02-07 Tom de Vries <tom@codesourcery.com>
2080 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2083 2018-02-07 Richard Biener <rguenther@suse.de>
2085 PR tree-optimization/84204
2086 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2089 PR tree-optimization/84205
2090 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2091 special-case isl_ast_op_zdiv_r.
2093 PR tree-optimization/84223
2094 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2095 Only add conditions from within the region.
2096 (gather_bbs::after_dom_children): Adjust.
2098 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2101 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2102 * config/avr/avr.md: Only post-reload split REG-REG moves if
2103 either register is GENERAL_REG_P.
2105 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2107 PR tree-optimization/84235
2108 * tree-ssa-scopedtables.c
2109 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2110 if the subtraction is performed in floating point type where NaNs are
2111 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2112 build 1. Formatting fix.
2114 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2117 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2118 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2119 and skip it regardless of bb boundaries. Use CALL_P macro,
2120 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2121 unnecessarily, formatting fix.
2123 2018-02-06 Michael Collison <michael.collison@arm.com>
2125 * config/arm/thumb2.md:
2126 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2127 (*thumb_mov_notscc): Ditto.
2129 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2132 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2133 unsigned_fix, not "s".
2135 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2137 * configure.ac (gcc_fn_eh_frame_ro): New function.
2138 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2139 correct .eh_frame permissions.
2140 * configure: Regenerate.
2142 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2144 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2147 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2149 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2150 Display warning message for -mno-speculate-indirect-jumps.
2152 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2154 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2156 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2158 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2160 PR tree-optimization/84225
2161 * tree-eh.c (find_trapping_overflow): Only call
2162 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2164 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2167 * config/i386/i386.c: Reimplement the check of possible options
2168 -mibt/-mshstk conbination. Change error messages.
2169 * doc/invoke.texi: Fix a typo: remove extra '='.
2171 2018-02-06 Marek Polacek <polacek@redhat.com>
2173 PR tree-optimization/84228
2174 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2176 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2179 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2180 emitted arch directives.
2181 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2182 __ARM_FEATURE_COPROC before changing architectures.
2184 2018-02-06 Richard Biener <rguenther@suse.de>
2186 * config/i386/i386.c (print_reg): Fix typo.
2187 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2189 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2191 * configure: Regenerate.
2193 2018-02-05 Martin Sebor <msebor@redhat.com>
2195 PR tree-optimization/83369
2196 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2199 2018-02-05 Martin Liska <mliska@suse.cz>
2201 * doc/invoke.texi: Cherry-pick upstream r323995.
2203 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2205 * ira.c (ira_init_register_move_cost): Adjust comment.
2207 2018-02-05 Martin Liska <mliska@suse.cz>
2209 PR gcov-profile/84137
2210 * doc/gcov.texi: Fix typo in documentation.
2212 2018-02-05 Martin Liska <mliska@suse.cz>
2214 PR gcov-profile/83879
2215 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2216 using dlopen functionality.
2218 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2220 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2221 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2222 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2223 _mm_maskz_range_round_ss): New intrinsics.
2224 (__builtin_ia32_rangesd128_round)
2225 (__builtin_ia32_rangess128_round): Remove.
2226 (__builtin_ia32_rangesd128_mask_round,
2227 __builtin_ia32_rangess128_mask_round): New builtins.
2228 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2229 __builtin_ia32_rangess128_round): Remove.
2230 (__builtin_ia32_rangesd128_mask_round,
2231 __builtin_ia32_rangess128_mask_round): New builtins.
2232 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2233 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2234 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2235 "<round_saeonly_constraint>")): Changed to ...
2236 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2237 "<round_saeonly_scalar_constraint>")): ... this.
2238 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2239 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2240 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2241 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2242 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2244 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2246 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2248 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2249 Remove all values except native, 8540 and 8548.
2251 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2253 * config/i386/i386.c (ix86_output_function_return): Pass
2254 INVALID_REGNUM, instead of -1, as invalid register number to
2255 indirect_thunk_name and output_indirect_thunk.
2257 2018-02-02 Julia Koval <julia.koval@intel.com>
2259 * config.gcc: Add -march=icelake.
2260 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2261 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2262 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2263 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2264 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2265 (processor_target_table): Add icelake.
2266 (ix86_option_override_internal): Handle new PTAs.
2267 (get_builtin_code_for_version): Handle icelake.
2268 (M_INTEL_COREI7_ICELAKE): New.
2269 (fold_builtin_cpu): Handle icelake.
2270 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2271 * doc/invoke.texi: Add -march=icelake.
2273 2018-02-02 Julia Koval <julia.koval@intel.com>
2275 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2276 to wide_int_bitmask.
2277 * wide-int-bitmask.h: New.
2279 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2282 * config/i386/i386.md: Replace Pmode with word_mode in
2283 builtin_setjmp_setup and builtin_longjmp to support x32.
2285 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2289 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2291 (rs6000_supported_cpu_names): New static variable.
2292 (linux_cpu_translation_table): Likewise.
2293 (elf_platform) <cpu>: Define new static variable and use it.
2294 Translate kernel AT_PLATFORM name to canonical name if needed.
2295 Error if platform name is unknown.
2297 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2300 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2302 2018-02-01 Jeff Law <law@redhat.com>
2305 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2306 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2307 the scratch if RELEASE_VIA_POP is false.
2308 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2309 If we have to save a temporary register, decrement SIZE appropriately.
2310 Pass new arguments to release_scratch_register_on_entry.
2311 (ix86_adjust_stack_and_probe): Likewise.
2312 (ix86_emit_probe_stack_range): Pass new arguments to
2313 release_scratch_register_on_entry.
2315 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2317 PR rtl-optimization/84157
2318 * combine.c (change_zero_ext): Use REG_P predicate in
2319 front of HARD_REGISTER_P predicate.
2321 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2323 * config/avr/avr.c (avr_option_override): Move disabling of
2324 -fdelete-null-pointer-checks to...
2325 * common/config/avr/avr-common.c (avr_option_optimization_table):
2328 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2330 PR tree-optimization/81635
2331 * tree-data-ref.c (split_constant_offset_1): For types that
2332 wrap on overflow, try to use range info to prove that wrapping
2335 2018-02-01 Renlin Li <renlin.li@arm.com>
2338 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2340 (aarch64_register_move_cost): Likewise.
2341 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2343 (REG_CLASS_NAMES): Likewise.
2344 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2345 TAILCALL_ADDR_REGS. Remove IP registers.
2346 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2348 2018-02-01 Richard Biener <rguenther@suse.de>
2350 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2351 for specifying RPO order and allow NULL for that.
2352 * domwalk.c (dom_walker::dom_walker): Likewise.
2353 (dom_walker::walk): Handle NULL RPO order.
2354 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2356 (rewrite_update_dom_walker): Likewise.
2357 (mark_def_dom_walker): Likewise.
2359 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2361 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2362 (aarch64_maybe_expand_sve_subreg_move): Declare.
2363 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2364 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2366 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2367 that are semantically a reverse operation.
2368 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2369 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2370 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2372 (aarch64_can_change_mode_class): For big-endian, forbid changes
2373 between two SVE modes if they have different element sizes.
2375 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2377 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2378 the TImode handling for big-endian targets.
2380 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2382 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2383 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2385 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2386 Remove BSWAP handing for big-endian targets and use the form of
2387 LD1RQ appropariate for the mode.
2389 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2391 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2392 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2395 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2398 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2399 check for operands that need to go through aarch64_sve_reload_be.
2401 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2403 PR tree-optimization/81661
2404 PR tree-optimization/84117
2405 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2406 * tree-eh.c: Include gimplify.h.
2407 (find_trapping_overflow, replace_trapping_overflow,
2408 rewrite_to_non_trapping_overflow): New functions.
2409 * tree-vect-loop.c: Include tree-eh.h.
2410 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2411 * tree-data-ref.c: Include tree-eh.h.
2412 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2414 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2416 PR rtl-optimization/84123
2417 * combine.c (change_zero_ext): Check if hard register satisfies
2418 can_change_dest_mode before calling gen_lowpart_SUBREG.
2420 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2423 * ira.c (ira_init_register_move_cost): Remove assert.
2425 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2427 PR rtl-optimization/84071
2428 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2429 * doc/tm.texi: Regenerate.
2431 2018-01-31 Richard Biener <rguenther@suse.de>
2433 PR tree-optimization/84132
2434 * tree-data-ref.c (analyze_miv_subscript): Properly
2435 check whether evolution_function_is_affine_multivariate_p
2436 before calling gcd_of_steps_may_divide_p.
2438 2018-01-31 Julia Koval <julia.koval@intel.com>
2441 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2442 * config/i386/i386.md (rdpid_rex64) New.
2443 (rdpid): Make 32bit only.
2445 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2448 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2449 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2451 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2454 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2456 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2458 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2460 PR rtl-optimization/84071
2461 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2462 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2464 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2466 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2467 (arc_attribute_table): Add 'aux' attribute.
2468 (arc_in_small_data_p): Consider aux like variables.
2469 (arc_is_aux_reg_p): New function.
2470 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2471 (arc_get_aux_arg): New function.
2472 (prepare_move_operands): Handle aux-register access.
2473 (arc_handle_aux_attribute): New function.
2474 * doc/extend.texi (ARC Variable attributes): Add subsection.
2476 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2478 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2479 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2480 (arc_attribute_table): Add 'uncached' attribute.
2481 (arc_print_operand): Print '.di' flag for uncached memory
2483 (arc_in_small_data_p): Do not consider for small data the uncached
2485 (arc_is_uncached_mem_p): New function.
2486 * config/arc/predicates.md (compact_store_memory_operand): Check
2487 for uncached memory accesses.
2488 (nonvol_nonimm_operand): Likewise.
2489 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2491 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2494 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2495 falign-loops=): Add Optimization flag.
2497 2018-01-30 Jeff Law <law@redhat.com>
2500 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2501 INT_REGISTERS_SAVED. Check it prior to calling
2502 get_scratch_register_on_entry.
2503 (ix86_adjust_stack_and_probe): Similarly.
2504 (ix86_emit_probe_stack_range): Similarly.
2505 (ix86_expand_prologue): Corresponding changes.
2507 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2510 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2511 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2513 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2516 * lra-constraints.c (curr_insn_transform): Process AND in the
2519 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2521 PR rtl-optimization/83986
2522 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2523 dependence against last_pending_memory_flush in addition to
2526 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2528 PR tree-optimization/81611
2529 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2532 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2535 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2538 2018-01-30 Richard Biener <rguenther@suse.de>
2539 Jakub Jelinek <jakub@redhat.com>
2541 PR tree-optimization/84111
2542 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2543 inner loops added during recursion, as they don't have up-to-date
2546 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2549 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2550 (can_inline_edge_by_limits_p): ... here.
2551 (can_early_inline_edge_p, check_callers,
2552 update_caller_keys, update_callee_keys, recursive_inlining,
2553 add_new_edges_to_heap, speculation_useful_p,
2554 inline_small_functions,
2555 inline_small_functions, flatten_function,
2556 inline_to_all_callers_1): Update.
2558 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2560 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2563 2018-01-30 Richard Biener <rguenther@suse.de>
2565 PR tree-optimization/83008
2566 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2567 invariant and constant vector uses in stmts when they need
2570 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2573 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2574 * configure: Regenerate.
2576 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2578 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2580 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2581 Use gen_rtx_REG rather than gen_lowpart.
2583 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2585 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2586 rather than 0 when creating partial subregs.
2588 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2590 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2593 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2596 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2597 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2598 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2599 flags. This restores the settings used before the 2017-07-24.
2600 Turning off pre increment/decrement/modify allows IVOPTS to
2601 optimize DF/SF loops where the index is an int.
2603 2018-01-29 Richard Biener <rguenther@suse.de>
2604 Kelvin Nilsen <kelvin@gcc.gnu.org>
2607 * tree-vect-stmts.c (vectorizable_call): Don't call
2608 targetm.vectorize_builtin_md_vectorized_function if callee is
2611 2018-01-22 Carl Love <cel@us.ibm.com>
2613 * doc/extend.tex: Fix typo in second arg in
2614 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2616 2018-01-29 Richard Biener <rguenther@suse.de>
2618 PR tree-optimization/84086
2619 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2620 (flush_ssaname_freelist): When SSA names were released reset
2621 the SCEV hash table.
2623 2018-01-29 Richard Biener <rguenther@suse.de>
2625 PR tree-optimization/84057
2626 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2627 removed paths when removing edges.
2629 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2631 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2632 -mfunction-return=@var{choice}.
2634 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2637 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2639 (layout::print_source_line): Likewise.
2640 (test_get_line_width_without_trailing_whitespace): Add test cases.
2642 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2645 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2648 2018-01-26 Jim Wilson <jimw@sifive.com>
2650 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2652 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2655 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2657 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2658 and CMP + SUB-immediate -> SUBS.
2660 2018-01-26 Martin Sebor <msebor@redhat.com>
2662 PR tree-optimization/83896
2663 * tree-ssa-strlen.c (get_string_len): Rename...
2664 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2665 Avoid assuming length is constant.
2666 (handle_char_store): Use HOST_WIDE_INT for string length.
2668 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2671 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2672 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2674 2018-01-26 Richard Biener <rguenther@suse.de>
2676 PR rtl-optimization/84003
2677 * dse.c (record_store): Only record redundant stores when
2678 the earlier store aliases at least all accesses the later one does.
2680 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2682 PR rtl-optimization/83985
2683 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2684 REG_CFA_RESTORE insns.
2685 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2686 REG_CFA_RESTORE insns here.
2689 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2690 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2692 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2694 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2695 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2696 (arc_init): Likewise.
2697 (arc_override_options): Likewise.
2698 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2700 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2702 * config/arc/arc.h (TARGET_DBNZ): Define.
2703 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2704 properly set the tune attribute.
2705 (dbnz): Use TARGET_DBNZ guard.
2706 * config/arc/arc.opt (mtune): Add core3 option.
2708 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2710 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2711 recognize new pic like addresses.
2712 (arc_delegitimize_address): Clean up.
2714 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2716 * config/arc/arc-arches.def: Option mrf16 valid for all
2718 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2719 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2720 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2721 * config/arc/arc-tables.opt: Regenerate.
2722 * config/arc/arc.c (arc_conditional_register_usage): Handle
2723 reduced register file case.
2724 (arc_file_start): Set must have build attributes.
2725 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2727 * config/arc/arc.opt (mrf16): Add new option.
2728 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2729 * config/arc/genmultilib.awk: Handle new mrf16 option.
2730 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2731 * config/arc/t-multilib: Regenerate.
2732 * doc/invoke.texi (ARC Options): Document mrf16 option.
2734 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2736 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2737 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2738 (arc_attribute_table): Add 'secure_call' attribute.
2739 (arc_print_operand): Print secure call operand.
2740 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2742 (arc_is_secure_call_p): New function. * config/arc/arc.md
2743 (call_i): Add support for sjli instruction.
2744 (call_value_i): Likewise.
2745 * config/arc/constraints.md (Csc): New constraint.
2747 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2748 John Eric Martin <John.Martin@emmicro-us.com>
2750 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2751 * config/arc/arc.c (_arc_jli_section): New struct.
2752 (arc_jli_section): New type.
2753 (rc_jli_sections): New static variable.
2754 (arc_handle_jli_attribute): New function.
2755 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2756 (arc_file_end): New function.
2757 (TARGET_ASM_FILE_END): Define.
2758 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2759 (arc_add_jli_section): New function.
2760 (jli_call_scan): Likewise.
2761 (arc_reorg): Call jli_call_scan.
2762 (arc_output_addsi): Remove 'S' from printing asm operand.
2763 (arc_is_jli_call_p): New function.
2764 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2766 (movhi_insn): Likewise.
2767 (movsi_insn): Likewise.
2768 (movsi_set_cc_insn): Likewise.
2769 (loadqi_update): Likewise.
2770 (load_zeroextendqisi_update): Likewise.
2771 (load_signextendqisi_update): Likewise.
2772 (loadhi_update): Likewise.
2773 (load_zeroextendhisi_update): Likewise.
2774 (load_signextendhisi_update): Likewise.
2775 (loadsi_update): Likewise.
2776 (loadsf_update): Likewise.
2777 (movsicc_insn): Likewise.
2778 (bset_insn): Likewise.
2779 (bxor_insn): Likewise.
2780 (bclr_insn): Likewise.
2781 (bmsk_insn): Likewise.
2782 (bicsi3_insn): Likewise.
2783 (cmpsi_cc_c_insn): Likewise.
2784 (movsi_ne): Likewise.
2785 (movsi_cond_exec): Likewise.
2786 (clrsbsi2): Likewise.
2797 (call_i): Remove 'S' asm letter, add jli instruction.
2798 (call_value_i): Likewise.
2799 * config/arc/arc.op (mjli-always): New option.
2800 * config/arc/constraints.md (Cji): New constraint.
2801 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2803 (subsf3_fpx): Likewise.
2804 (mulsf3_fpx): Likewise.
2805 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2807 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2809 * doc/invoke.texi (ARC): Document mjli-always option.
2811 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2813 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2814 avoid addition with 0 and use incw and decw where possible.
2816 2018-01-26 Richard Biener <rguenther@suse.de>
2818 PR tree-optimization/81082
2819 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2820 association if it requires casting to unsigned.
2821 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2822 from fold_plusminus_mult_expr to catch important cases late when
2823 range info is available.
2825 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2827 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2828 * configure.ac (hidden_linkonce): New test.
2829 * configure: Regenerate.
2830 * config.in: Regenerate.
2832 2018-01-26 Julia Koval <julia.koval@intel.com>
2834 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2835 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2836 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2837 _mm_mask_bitshuffle_epi64_mask): Fix type.
2838 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2839 USI_FTYPE_V4DI_V4DI_USI): Remove.
2840 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2841 __builtin_ia32_vpshufbitqmb256_mask,
2842 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2843 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2844 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2846 2018-01-26 Alan Modra <amodra@gmail.com>
2849 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2850 UNSPEC_VBPERMQ. Sort other unspecs.
2852 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2854 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2856 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2859 * predict.c (drop_profile): Do not push/pop cfun; update also
2861 (handle_missing_profiles): Fix logic looking for zero profiles.
2863 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2866 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2867 on functions with #pragma omp declare simd or functions with simd
2869 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2870 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2871 Remove trailing \n from warning_at calls.
2873 2018-01-25 Tom de Vries <tom@codesourcery.com>
2876 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2877 for neutered workers.
2879 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2882 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2883 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2885 2018-01-24 Jeff Law <law@redhat.com>
2888 * i386.c (get_probe_interval): Move to earlier point.
2889 (ix86_compute_frame_layout): If -fstack-clash-protection and
2890 the frame is larger than the probe interval, then use pushes
2891 to save registers rather than reg->mem moves.
2892 (ix86_expand_prologue): Remove conditional for int_registers_saved
2895 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2898 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2899 min/max for never referenced object.
2901 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2904 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2906 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2907 attributes from DECL_ATTRIBUTES (decl) without affecting
2908 DECL_ATTRIBUTES (current_function_decl).
2909 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2910 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2912 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2914 PR tree-optimization/83979
2915 * fold-const.c (fold_comparison): Use constant_boolean_node
2916 instead of boolean_{true,false}_node.
2918 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2920 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2923 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2925 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2926 Simplify the clause that sets the length attribute.
2927 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2928 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2929 clause that sets the length attribute.
2930 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2932 2018-01-24 Tom de Vries <tom@codesourcery.com>
2935 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2936 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2937 Add strict parameter.
2938 (prevent_branch_around_nothing): Insert dummy insn between branch to
2939 label and label with no ptx insn inbetween.
2940 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2942 2018-01-24 Tom de Vries <tom@codesourcery.com>
2945 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2946 for neutered threads in warp.
2947 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2949 2018-01-24 Richard Biener <rguenther@suse.de>
2951 PR tree-optimization/83176
2952 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2955 2018-01-24 Richard Biener <rguenther@suse.de>
2957 PR tree-optimization/82819
2958 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2959 code generating pluses that are no-ops in the target precision.
2961 2018-01-24 Richard Biener <rguenther@suse.de>
2964 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2966 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2968 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2969 to merge probabilities.
2970 * predict.c (probably_never_executed): Also mark as cold functions
2971 with global 0 profile and guessed local profile.
2972 * profile-count.c (profile_probability::combine_with_count): New
2974 * profile-count.h (profile_probability::operator*,
2975 profile_probability::operator*=, profile_probability::operator/,
2976 profile_probability::operator/=): Reduce precision to adjusted
2977 and set value to guessed on contradictory divisions.
2978 (profile_probability::combine_with_freq): Remove.
2979 (profile_probability::combine_wiht_count): Declare.
2980 (profile_count::force_nonzero):: Set to adjusted.
2981 (profile_count::probability_in):: Set quality to adjusted.
2982 * tree-ssa-tail-merge.c (replace_block_by): Use
2985 2018-01-23 Andrew Waterman <andrew@sifive.com>
2986 Jim Wilson <jimw@sifive.com>
2988 * config/riscv/riscv.c (riscv_stack_boundary): New.
2989 (riscv_option_override): Set riscv_stack_boundary. Handle
2990 riscv_preferred_stack_boundary_arg.
2991 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2992 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2993 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2994 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2995 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2996 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2998 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3001 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3002 of struct ix86_frame.
3003 (ix86_expand_epilogue): Likewise. Add a local variable for
3004 the reg_save_offset field in struct ix86_frame.
3006 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3008 PR tree-optimization/82604
3009 * tree-loop-distribution.c (enum partition_kind): New enum item
3010 PKIND_PARTIAL_MEMSET.
3011 (partition_builtin_p): Support above new enum item.
3012 (generate_code_for_partition): Ditto.
3013 (compute_access_range): Differentiate cases that equality can be
3014 proven at all loops, the innermost loops or no loops.
3015 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3016 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3017 (finalize_partitions, distribute_loop): Don't fuse partition of
3018 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3019 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3022 2018-01-23 Martin Liska <mliska@suse.cz>
3024 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3025 order to ignore the predictor.
3026 (PRED_POLYMORPHIC_CALL): Likewise.
3027 (PRED_RECURSIVE_CALL): Likewise.
3029 2018-01-23 Martin Liska <mliska@suse.cz>
3031 * tree-profile.c (tree_profiling): Print function header to
3032 aware reader which function we are working on.
3033 * value-prof.c (gimple_find_values_to_profile): Do not print
3034 not interesting value histograms.
3036 2018-01-23 Martin Liska <mliska@suse.cz>
3038 * profile-count.h (enum profile_quality): Add
3039 profile_uninitialized as the first value. Do not number values
3040 as they are zero based.
3041 (profile_count::verify): Update sanity check.
3042 (profile_probability::verify): Likewise.
3044 2018-01-23 Nathan Sidwell <nathan@acm.org>
3046 * doc/invoke.texi (ffor-scope): Deprecate.
3048 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3050 PR tree-optimization/83510
3051 * domwalk.c (set_all_edges_as_executable): New function.
3052 (dom_walker::dom_walker): Convert bool param
3053 "skip_unreachable_blocks" to enum reachability. Move setup of
3054 edge flags to set_all_edges_as_executable and only do it when
3055 reachability is REACHABLE_BLOCKS.
3056 * domwalk.h (enum dom_walker::reachability): New enum.
3057 (dom_walker::dom_walker): Convert bool param
3058 "skip_unreachable_blocks" to enum reachability.
3059 (set_all_edges_as_executable): New decl.
3060 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3061 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3063 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3064 but converting true to REACHABLE_BLOCKS.
3065 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3067 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3068 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3069 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3071 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3072 if check_all_array_refs will be called.
3074 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3076 * tree.c (selftest::test_location_wrappers): Add more test
3079 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3081 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3082 (selftest::test_bit_in_range): Likewise.
3084 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3087 * doc/sourcebuild.texi (vect_float): Say that the selector
3088 only describes the situation when -funsafe-math-optimizations is on.
3089 (vect_float_strict): Document.
3091 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3093 PR tree-optimization/83965
3094 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3095 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3096 instead of checking only for a reduction.
3097 (vect_recog_widen_sum_pattern): Likewise.
3099 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3101 * predict.c (probably_never_executed): Only use precise profile info.
3102 (compute_function_frequency): Skip after inlining hack since we now
3103 have quality checking.
3105 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3107 * profile-count.h (profile_probability::very_unlikely,
3108 profile_probability::unlikely, profile_probability::even): Set
3109 precision to guessed.
3111 2018-01-23 Richard Biener <rguenther@suse.de>
3113 PR tree-optimization/83963
3114 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3115 Properly terminate dominator walk when crossing the exit edge not
3116 when visiting its source block.
3118 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3121 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3122 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3124 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3126 PR tree-optimization/83957
3127 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3128 semicolon after for body surrounded by braces.
3130 PR tree-optimization/83081
3131 * profile-count.h (profile_probability::split): New method.
3132 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3133 Use profile_probability::split.
3134 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3135 when splitting a single conditional jump into 2.
3137 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3139 PR tree-optimization/69452
3140 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3143 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3145 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3146 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3147 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3149 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3151 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3152 * config/rl78/rl78.md (movdi): New define_expand.
3153 * config/rl78/rl78.c (rl78_split_movdi): New function.
3155 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3158 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3160 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3161 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3162 128-bit to produce an UNSPEC move to get the double word with the
3163 signbit and then a shift directly to do signbit.
3164 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3165 implementation with a new version that just does either a direct
3166 move or a regular move. Move memory interface to separate insns.
3167 Move insns so they are next to the expander.
3168 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3169 with signbit move. Split big and little endian case.
3170 (signbit<mode>2_dm_mem_le): Likewise.
3171 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3172 (signbit<mode>2_dm2): Likewise.
3174 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3176 * config/rl78/rl78.md (anddi3): New define_expand.
3178 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3180 * config/rl78/rl78.md (umindi3): New define_expand.
3182 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3184 * config/rl78/rl78.md (smindi3): New define_expand.
3186 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3188 * config/rl78/rl78.md (smaxdi3): New define_expand.
3190 2018-01-22 Carl Love <cel@us.ibm.com>
3192 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3193 LVX_V1TI): Add macro expansion.
3194 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3195 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3196 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3197 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3198 Change check to determine if the instruction is a byte reversing
3199 entry. Fix typo in comment.
3200 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3201 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3202 Add def_builtin calls for new builtins.
3203 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3204 Add define_insn expansion.
3206 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3208 * config/rl78/rl78.md (umaxdi3): New define_expand.
3210 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3212 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3213 for non-QImode registers.
3215 2018-01-22 Richard Biener <rguenther@suse.de>
3217 PR tree-optimization/83963
3218 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3219 including the loop exit block.
3220 (scop_detection::merge_sese): Likewise.
3221 (scop_detection::add_scop): Do it here instead.
3223 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3225 * doc/sourcebuild.texi (arm_softfloat): Document.
3227 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3230 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3231 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3232 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3234 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3235 David Edelsohn <dje.gcc@gmail.com>
3238 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3239 Change "crset eq" to "crset 2".
3240 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3241 (*call_indirect_aix<mode>_nospec): Likewise.
3242 (*call_value_indirect_aix<mode>_nospec): Likewise.
3243 (*call_indirect_elfv2<mode>_nospec): Likewise.
3244 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3245 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3246 change assembly output from . to $.
3247 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3248 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3249 (*tablejump<mode>_internal1_nospec): Likewise.
3251 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3254 * config/sh/sh_optimize_sett_clrt.cc:
3255 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3257 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3259 PR tree-optimization/83940
3260 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3261 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3262 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3263 use it to pass back the definition type.
3264 (vect_check_store_rhs): Likewise rhs_dt_out.
3265 (vect_build_gather_load_calls): Add a mask_dt argument and use
3266 it instead of a call to vect_is_simple_use.
3267 (vectorizable_store): Update calls to vect_check_load_store_mask
3268 and vect_check_store_rhs. Use the dt returned by the latter instead
3269 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3270 instead of calls to vect_is_simple_use. Pass the scalar rather
3271 than the vector operand to vect_is_simple_use when handling
3272 second and subsequent copies of an rhs value.
3273 (vectorizable_load): Update calls to vect_check_load_store_mask
3274 and vect_build_gather_load_calls. Use the cached mask_dt and
3275 gs_info.offset_dt instead of calls to vect_is_simple_use.
3277 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3280 * tree-emutls.c: Include gimplify.h.
3281 (lower_emutls_2): New function.
3282 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3283 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3284 it before further processing.
3287 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3288 UINTVAL (trueop1) instead of INTVAL (op1).
3290 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3294 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3295 INCOMING_FRAME_SP_OFFSET if not defined.
3296 (scan_trace): Add ENTRY argument. If true and
3297 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3298 emit a note to adjust the CFA offset.
3299 (create_cfi_notes): Adjust scan_trace callers.
3300 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3301 INCOMING_FRAME_SP_OFFSET in the CIE.
3302 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3303 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3305 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3306 * doc/tm.texi: Regenerated.
3308 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3310 PR rtl-optimization/83147
3311 * lra-constraints.c (remove_inheritance_pseudos): Use
3312 lra_substitute_pseudo_within_insn.
3314 2018-01-19 Tom de Vries <tom@codesourcery.com>
3315 Cesar Philippidis <cesar@codesourcery.com>
3318 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3320 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3323 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3324 spaces for function labels.
3326 2018-01-19 Martin Liska <mliska@suse.cz>
3328 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3329 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3330 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3331 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3332 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3333 (PRED_CONST_RETURN): Change from 69 to 65.
3334 (PRED_NULL_RETURN): Change from 91 to 71.
3335 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3336 (PRED_LOOP_GUARD): Change from 66 to 73.
3338 2018-01-19 Martin Liska <mliska@suse.cz>
3340 * predict.c (predict_insn_def): Add new assert.
3341 (struct branch_predictor): Change type to signed integer.
3342 (test_prediction_value_range): Amend test to cover
3344 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3345 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3346 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3347 (PRED_LOOP_IV_COMPARE): Likewise.
3348 * predict.h (PROB_UNINITIALIZED): Define new constant.
3350 2018-01-19 Martin Liska <mliska@suse.cz>
3352 * predict.c (dump_prediction): Add new format for
3353 analyze_brprob.py script which is enabled with -details
3355 * profile-count.h (precise_p): New function.
3357 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3359 PR tree-optimization/83922
3360 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3361 there are no statements that need masking.
3362 (vect_active_double_reduction_p): New function.
3363 (vect_analyze_loop_operations): Use it when handling phis that
3364 are not in the loop header.
3366 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3368 PR tree-optimization/83914
3369 * tree-vect-loop.c (vectorizable_induction): Don't convert
3370 init_expr or apply the peeling adjustment for inductions
3371 that are nested within the vectorized loop.
3373 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3375 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3378 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3382 * function.h (gimplify_parameters): Add gimple_seq * argument.
3383 * function.c: Include gimple.h and options.h.
3384 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3385 for the added local temporaries if needed.
3386 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3387 if there are any parameter cleanups, wrap whole body into a
3388 try/finally with the cleanups.
3390 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3393 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3394 Use GET_MODE_CLASS for scalar floating point.
3396 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3400 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3401 Fix call of call_cgraph_insertion_hooks.
3403 2018-01-18 Martin Sebor <msebor@redhat.com>
3405 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3407 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3410 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3413 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3416 * common.opt: (-ffile-prefix-map): New option.
3417 * opts.c (common_handle_option): Defer it.
3418 * opts-global.c (handle_common_deferred_options): Handle it.
3419 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3420 * file-prefix-map.h: New file.
3421 (remap_debug_filename, add_debug_prefix_map): ...here.
3422 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3423 * final.c (debug_prefix_map, add_debug_prefix_map
3424 remap_debug_filename): Move to...
3425 * file-prefix-map.c: New file.
3426 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3427 generalize, get rid of alloca(), use strrchr() instead of strchr().
3428 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3429 Implement in terms of add_prefix_map().
3430 (remap_macro_filename, remap_debug_filename): Implement in term of
3432 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3433 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3434 * dbxout.c: Include file-prefix-map.h.
3435 * varasm.c: Likewise.
3436 * vmsdbgout.c: Likewise.
3437 * xcoffout.c: Likewise.
3438 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3439 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3440 * doc/invoke.texi (-ffile-prefix-map): Document.
3441 (-fdebug-prefix-map): Update description.
3443 2018-01-18 Martin Liska <mliska@suse.cz>
3445 * config/i386/i386.c (indirect_thunk_name): Document that also
3447 (output_indirect_thunk): Document why both instructions
3448 (pause and lfence) are generated.
3450 2018-01-18 Richard Biener <rguenther@suse.de>
3452 PR tree-optimization/83887
3453 * graphite-scop-detection.c
3454 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3455 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3456 (scop_detection::merge_sese): Re-implement with a flood-fill
3457 algorithm that properly finds a SESE region if it exists.
3459 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3462 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3463 pointer_diff optimizations use view_convert instead of convert.
3465 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3467 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3468 Generate different code for -mno-speculate-indirect-jumps.
3469 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3470 (*call_indirect_aix<mode>): Disable for
3471 -mno-speculate-indirect-jumps.
3472 (*call_indirect_aix<mode>_nospec): New define_insn.
3473 (*call_value_indirect_aix<mode>): Disable for
3474 -mno-speculate-indirect-jumps.
3475 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3476 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3477 -mno-speculate-indirect-jumps.
3478 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3480 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3482 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3483 long double type, set the flags for noting the default long double
3484 type, even if we don't pass or return a long double type.
3486 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3489 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3492 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3494 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3495 support for merge[hl].
3496 (fold_mergehl_helper): New helper function.
3497 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3498 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3499 (altivec_vmrglw_direct): Add xxmrglw insn.
3501 2018-01-17 Andrew Waterman <andrew@sifive.com>
3503 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3504 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3506 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3509 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3510 call the lto_location_cache before reading the
3511 DECL_SOURCE_LOCATION of the types.
3513 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3514 Richard Sandiford <richard.sandiford@linaro.org>
3516 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3517 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3518 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3519 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3520 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3522 * config/aarch64/constraints.md (aarch64_movti_operand):
3524 * config/aarch64/predicates.md (Uti): Add new constraint.
3526 2018-01-17 Carl Love <cel@us.ibm.com>
3528 * config/rs6000/vsx.md (define_expand xl_len_r,
3529 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3530 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3532 (define_expand, define_insn): Move the shift left from the
3533 define_insn to the define_expand for lxvl and stxvl instructions.
3534 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3535 and XL_LEN_R definitions to PURE.
3537 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3539 * config/i386/i386.c (indirect_thunk_name): Declare regno
3540 as unsigned int. Compare regno with INVALID_REGNUM.
3541 (output_indirect_thunk): Ditto.
3542 (output_indirect_thunk_function): Ditto.
3543 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3544 in the call to output_indirect_thunk_function.
3546 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3549 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3550 rather than the size of inner_type to determine the stack slot size
3551 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3553 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3556 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3559 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3561 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3562 endian Linux systems to optionally enable multilibs for selecting
3563 the long double type if the user configured an explicit type.
3564 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3565 have no long double multilibs if not defined.
3566 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3567 warn if the user used -mabi={ieee,ibm}longdouble and we built
3568 multilibs for long double.
3569 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3570 appropriate multilib option.
3571 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3573 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3574 for building long double multilibs.
3575 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3577 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3579 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3582 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3584 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3587 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3590 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3593 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3595 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3596 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3598 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3600 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3601 different rtl trees depending on TARGET_64BIT.
3602 (rs6000_gen_lvx): Likewise.
3604 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3606 * config/visium/visium.md (nop): Tweak comment.
3607 (hazard_nop): Likewise.
3609 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3611 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3612 -mspeculate-indirect-jumps.
3613 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3614 for -mno-speculate-indirect-jumps.
3615 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3616 (*call_value_indirect_elfv2<mode>): Disable for
3617 -mno-speculate-indirect-jumps.
3618 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3619 (indirect_jump): Emit different RTL for
3620 -mno-speculate-indirect-jumps.
3621 (*indirect_jump<mode>): Disable for
3622 -mno-speculate-indirect-jumps.
3623 (*indirect_jump<mode>_nospec): New define_insn.
3624 (tablejump): Emit different RTL for
3625 -mno-speculate-indirect-jumps.
3626 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3627 (tablejumpsi_nospec): New define_expand.
3628 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3629 (tablejumpdi_nospec): New define_expand.
3630 (*tablejump<mode>_internal1): Disable for
3631 -mno-speculate-indirect-jumps.
3632 (*tablejump<mode>_internal1_nospec): New define_insn.
3633 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3636 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3638 * caller-save.c (insert_save): Drop unnecessary parameter. All
3641 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3642 Richard Biener <rguenth@suse.de>
3645 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3646 return early, inline manually is_gimple_sizepos. Make sure if we
3647 call gimplify_expr we don't end up with a gimple constant.
3648 * tree.c (variably_modified_type_p): Don't return true for
3649 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3650 * gimplify.h (is_gimple_sizepos): Remove.
3652 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3654 PR tree-optimization/83857
3655 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3656 vectorizable_live_operation for pure SLP statements.
3657 (vectorizable_live_operation): Handle PHIs.
3659 2018-01-16 Richard Biener <rguenther@suse.de>
3661 PR tree-optimization/83867
3662 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3663 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3665 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3668 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3669 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3670 If off is not INTEGER_CST, issue a may not be aligned warning
3671 rather than isn't aligned. Use isn%'t rather than isn't.
3672 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3674 <case MULT_EXPR>: Improve the case when bottom and one of the
3675 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3676 operand, in that case check if the other operand is multiple of
3677 bottom divided by the INTEGER_CST operand.
3679 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3682 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3683 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3684 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3685 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3686 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3687 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3688 (pa_function_arg_size): New function.
3690 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3692 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3693 in a separate statement.
3695 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3697 PR tree-optimization/83847
3698 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3699 group gathers and scatters.
3701 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3703 PR rtl-optimization/86620
3704 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3706 PR rtl-optimization/83213
3707 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3708 to last if both are JUMP_INSNs.
3710 PR tree-optimization/83843
3711 * gimple-ssa-store-merging.c
3712 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3713 store_immediate_info for bswap/nop orig_stores.
3715 2018-01-15 Andrew Waterman <andrew@sifive.com>
3717 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3719 <UDIV>: Increase cost if !TARGET_DIV.
3721 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3723 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3724 (define_attr "cr_logical_3op"): New.
3725 (cceq_ior_compare): Adjust.
3726 (cceq_ior_compare_complement): Adjust.
3727 (*cceq_rev_compare): Adjust.
3728 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3729 (is_cracked_insn): Adjust.
3730 (insn_must_be_first_in_group): Adjust.
3731 * config/rs6000/40x.md: Adjust.
3732 * config/rs6000/440.md: Adjust.
3733 * config/rs6000/476.md: Adjust.
3734 * config/rs6000/601.md: Adjust.
3735 * config/rs6000/603.md: Adjust.
3736 * config/rs6000/6xx.md: Adjust.
3737 * config/rs6000/7450.md: Adjust.
3738 * config/rs6000/7xx.md: Adjust.
3739 * config/rs6000/8540.md: Adjust.
3740 * config/rs6000/cell.md: Adjust.
3741 * config/rs6000/e300c2c3.md: Adjust.
3742 * config/rs6000/e500mc.md: Adjust.
3743 * config/rs6000/e500mc64.md: Adjust.
3744 * config/rs6000/e5500.md: Adjust.
3745 * config/rs6000/e6500.md: Adjust.
3746 * config/rs6000/mpc.md: Adjust.
3747 * config/rs6000/power4.md: Adjust.
3748 * config/rs6000/power5.md: Adjust.
3749 * config/rs6000/power6.md: Adjust.
3750 * config/rs6000/power7.md: Adjust.
3751 * config/rs6000/power8.md: Adjust.
3752 * config/rs6000/power9.md: Adjust.
3753 * config/rs6000/rs64.md: Adjust.
3754 * config/rs6000/titan.md: Adjust.
3756 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3758 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3759 ix86_indirect_branch_register logic.
3761 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3763 * config/i386/constraints.md (Bs): Update
3764 ix86_indirect_branch_register check. Don't check
3765 ix86_indirect_branch_register with GOT_memory_operand.
3767 * config/i386/predicates.md (GOT_memory_operand): Don't check
3768 ix86_indirect_branch_register here.
3769 (GOT32_symbol_operand): Likewise.
3771 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3773 * config/i386/predicates.md (constant_call_address_operand):
3774 Rewrite ix86_indirect_branch_register logic.
3775 (sibcall_insn_operand): Likewise.
3777 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3779 * config/i386/constraints.md (Bs): Replace
3780 ix86_indirect_branch_thunk_register with
3781 ix86_indirect_branch_register.
3783 * config/i386/i386.md (indirect_jump): Likewise.
3784 (tablejump): Likewise.
3785 (*sibcall_memory): Likewise.
3786 (*sibcall_value_memory): Likewise.
3787 Peepholes of indirect call and jump via memory: Likewise.
3788 * config/i386/i386.opt: Likewise.
3789 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3790 (GOT_memory_operand): Likewise.
3791 (call_insn_operand): Likewise.
3792 (sibcall_insn_operand): Likewise.
3793 (GOT32_symbol_operand): Likewise.
3795 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3798 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3799 type rather than type addr's type points to.
3800 (expand_omp_atomic_mutex): Likewise.
3801 (expand_omp_atomic): Likewise.
3803 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3806 * config/i386/i386.c (output_indirect_thunk_function): Use
3807 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3808 for __x86_return_thunk.
3810 2018-01-15 Richard Biener <rguenther@suse.de>
3813 * expmed.c (extract_bit_field_1): Fix typo.
3815 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3818 * config/arm/iterators.md (VF): New mode iterator.
3819 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3820 Remove integer-related logic from pattern.
3821 (neon_vabd<mode>_3): Likewise.
3823 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3826 * common.opt (fstrict-overflow): No longer an alias.
3827 (fwrapv-pointer): New option.
3828 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3829 also for pointer types based on flag_wrapv_pointer.
3830 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3831 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3832 opts->x_flag_wrapv got set.
3833 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3834 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3835 POINTER_TYPE_OVERFLOW_UNDEFINED.
3836 * match.pd: Likewise in address comparison pattern.
3837 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3839 2018-01-15 Richard Biener <rguenther@suse.de>
3842 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3843 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3844 Reset type names to their identifier if their TYPE_DECL doesn't
3845 have linkage (and thus is used for ODR and devirt).
3846 (save_debug_info_for_decl): Remove.
3847 (save_debug_info_for_type): Likewise.
3848 (add_tree_to_fld_list): Adjust.
3849 * tree-pretty-print.c (dump_generic_node): Make dumping of
3850 type names more robust.
3852 2018-01-15 Richard Biener <rguenther@suse.de>
3854 * BASE-VER: Bump to 8.0.1.
3856 2018-01-14 Martin Sebor <msebor@redhat.com>
3859 * builtins.c (check_access): Avoid warning when the no-warning bit
3862 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3864 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3865 * ira-color (allocno_hard_regs_compare): Likewise.
3867 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3870 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3871 Use .pushsection/.popsection.
3873 2018-01-14 Martin Sebor <msebor@redhat.com>
3876 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3878 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3880 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3881 entry from extra_headers.
3882 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3883 extra_headers, make the list bitwise identical to the i?86-*-* one.
3885 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3887 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3888 -mcmodel=large with -mindirect-branch=thunk,
3889 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3890 -mfunction-return=thunk-extern.
3891 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3892 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3893 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3895 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3897 * config/i386/i386.c (print_reg): Print the name of the full
3898 integer register without '%'.
3899 (ix86_print_operand): Handle 'V'.
3900 * doc/extend.texi: Document 'V' modifier.
3902 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3904 * config/i386/constraints.md (Bs): Disallow memory operand for
3905 -mindirect-branch-register.
3907 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3908 (GOT_memory_operand): Likewise.
3909 (call_insn_operand): Likewise.
3910 (sibcall_insn_operand): Likewise.
3911 (GOT32_symbol_operand): Likewise.
3912 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3913 for -mindirect-branch-register.
3914 (tablejump): Likewise.
3915 (*sibcall_memory): Likewise.
3916 (*sibcall_value_memory): Likewise.
3917 Disallow peepholes of indirect call and jump via memory for
3918 -mindirect-branch-register.
3919 (*call_pop): Replace m with Bw.
3920 (*call_value_pop): Likewise.
3921 (*sibcall_pop_memory): Replace m with Bs.
3922 * config/i386/i386.opt (mindirect-branch-register): New option.
3923 * doc/invoke.texi: Document -mindirect-branch-register option.
3925 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3927 * config/i386/i386-protos.h (ix86_output_function_return): New.
3928 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3929 set function_return_type.
3930 (indirect_thunk_name): Add ret_p to indicate thunk for function
3932 (output_indirect_thunk_function): Pass false to
3933 indirect_thunk_name.
3934 (ix86_output_indirect_branch_via_reg): Likewise.
3935 (ix86_output_indirect_branch_via_push): Likewise.
3936 (output_indirect_thunk_function): Create alias for function
3937 return thunk if regno < 0.
3938 (ix86_output_function_return): New function.
3939 (ix86_handle_fndecl_attribute): Handle function_return.
3940 (ix86_attribute_table): Add function_return.
3941 * config/i386/i386.h (machine_function): Add
3942 function_return_type.
3943 * config/i386/i386.md (simple_return_internal): Use
3944 ix86_output_function_return.
3945 (simple_return_internal_long): Likewise.
3946 * config/i386/i386.opt (mfunction-return=): New option.
3947 (indirect_branch): Mention -mfunction-return=.
3948 * doc/extend.texi: Document function_return function attribute.
3949 * doc/invoke.texi: Document -mfunction-return= option.
3951 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3953 * config/i386/i386-opts.h (indirect_branch): New.
3954 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3955 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3956 with local indirect jump when converting indirect call and jump.
3957 (ix86_set_indirect_branch_type): New.
3958 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3959 (indirectlabelno): New.
3960 (indirect_thunk_needed): Likewise.
3961 (indirect_thunk_bnd_needed): Likewise.
3962 (indirect_thunks_used): Likewise.
3963 (indirect_thunks_bnd_used): Likewise.
3964 (INDIRECT_LABEL): Likewise.
3965 (indirect_thunk_name): Likewise.
3966 (output_indirect_thunk): Likewise.
3967 (output_indirect_thunk_function): Likewise.
3968 (ix86_output_indirect_branch_via_reg): Likewise.
3969 (ix86_output_indirect_branch_via_push): Likewise.
3970 (ix86_output_indirect_branch): Likewise.
3971 (ix86_output_indirect_jmp): Likewise.
3972 (ix86_code_end): Call output_indirect_thunk_function if needed.
3973 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3975 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3976 (ix86_attribute_table): Add indirect_branch.
3977 * config/i386/i386.h (machine_function): Add indirect_branch_type
3978 and has_local_indirect_jump.
3979 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3981 (tablejump): Likewise.
3982 (*indirect_jump): Use ix86_output_indirect_jmp.
3983 (*tablejump_1): Likewise.
3984 (simple_return_indirect_internal): Likewise.
3985 * config/i386/i386.opt (mindirect-branch=): New option.
3986 (indirect_branch): New.
3989 (thunk-inline): Likewise.
3990 (thunk-extern): Likewise.
3991 * doc/extend.texi: Document indirect_branch function attribute.
3992 * doc/invoke.texi: Document -mindirect-branch= option.
3994 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3997 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3999 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4001 * ipa-inline.c (want_inline_small_function_p): Return false if
4002 inlining has already failed with CIF_FINAL_ERROR.
4003 (update_caller_keys): Call want_inline_small_function_p before
4005 (update_callee_keys): Likewise.
4007 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4009 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4011 (rs6000_quadword_masked_address_p): Likewise.
4012 (quad_aligned_load_p): Likewise.
4013 (quad_aligned_store_p): Likewise.
4014 (const_load_sequence_p): Add comment to describe the outer-most loop.
4015 (mimic_memory_attributes_and_flags): New function.
4016 (rs6000_gen_stvx): Likewise.
4017 (replace_swapped_aligned_store): Likewise.
4018 (rs6000_gen_lvx): Likewise.
4019 (replace_swapped_aligned_load): Likewise.
4020 (replace_swapped_load_constant): Capitalize argument name in
4021 comment describing this function.
4022 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4023 and stores that access quad-word aligned addresses and replace
4024 with stvx or lvx instructions when appropriate.
4025 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4026 New function prototype.
4027 (rs6000_quadword_masked_address_p): Likewise.
4028 (rs6000_gen_lvx): Likewise.
4029 (rs6000_gen_stvx): Likewise.
4030 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4031 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4032 when memory address is aligned.
4033 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4034 this split to select lvx instruction when memory address is aligned.
4035 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4036 instruction when memory address is aligned.
4037 (*vsx_le_perm_load_v16qi): Likewise.
4038 (four unnamed splitters): Modify to select the stvx instruction
4039 when memory is aligned.
4041 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4043 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4044 which appears in the queue multiple times.
4046 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4047 Alan Hayward <alan.hayward@arm.com>
4048 David Sherwood <david.sherwood@arm.com>
4050 * tree-vectorizer.h (vec_lower_bound): New structure.
4051 (_loop_vec_info): Add check_nonzero and lower_bounds.
4052 (LOOP_VINFO_CHECK_NONZERO): New macro.
4053 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4054 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4055 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4056 fields. Make seg_len the distance travelled, not including the
4058 (dr_direction_indicator): Declare.
4059 (dr_zero_step_indicator): Likewise.
4060 (dr_known_forward_stride_p): Likewise.
4061 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4063 (runtime_alias_check_p): Allow runtime alias checks with
4065 (operator ==): Compare access_size and align.
4066 (prune_runtime_alias_test_list): Rework for new distinction between
4067 the access_size and seg_len.
4068 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4070 (get_segment_min_max): New function.
4071 (create_intersect_range_checks): Use it.
4072 (dr_step_indicator): New function.
4073 (dr_direction_indicator): Likewise.
4074 (dr_zero_step_indicator): Likewise.
4075 (dr_known_forward_stride_p): Likewise.
4076 * tree-loop-distribution.c (data_ref_segment_size): Return
4077 DR_STEP * (niters - 1).
4078 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4080 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4081 (vect_preserves_scalar_order_p): New function, split out from...
4082 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4083 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4084 (vect_vfa_access_size): New function.
4085 (vect_vfa_align): Likewise.
4086 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4087 (dump_lower_bound): New function.
4088 (vect_check_lower_bound): Likewise.
4089 (vect_small_gap_p): Likewise.
4090 (vectorizable_with_step_bound_p): Likewise.
4091 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4092 depencies if the vectorization factor is 1. Convert the checks
4093 for nonzero steps into checks on the bounds of DR_STEP. Try using
4094 a bunds check for variable steps if the minimum required step is
4095 relatively small. Update calls to the dr_with_seg_len
4096 constructor and to vect_compile_time_alias.
4097 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4099 (vect_loop_versioning): Call it.
4100 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4102 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4104 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4105 Alan Hayward <alan.hayward@arm.com>
4106 David Sherwood <david.sherwood@arm.com>
4108 * doc/sourcebuild.texi (vect_scatter_store): Document.
4109 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4111 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4113 * genopinit.c (main): Add supports_vec_scatter_store and
4114 supports_vec_scatter_store_cached to target_optabs.
4115 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4116 IFN_MASK_SCATTER_STORE.
4117 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4119 * internal-fn.h (internal_store_fn_p): Declare.
4120 (internal_fn_stored_value_index): Likewise.
4121 * internal-fn.c (scatter_store_direct): New macro.
4122 (expand_scatter_store_optab_fn): New function.
4123 (direct_scatter_store_optab_supported_p): New macro.
4124 (internal_store_fn_p): New function.
4125 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4126 IFN_MASK_SCATTER_STORE.
4127 (internal_fn_mask_index): Likewise.
4128 (internal_fn_stored_value_index): New function.
4129 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4131 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4132 * optabs-query.c (supports_vec_scatter_store_p): New function.
4133 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4134 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4135 true for scatter stores.
4136 (vect_gather_scatter_fn_p): Handle scatter stores too.
4137 (vect_check_gather_scatter): Consider using scatter stores if
4138 supports_vec_scatter_store_p.
4139 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4141 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4142 internal_fn_stored_value_index.
4143 (check_load_store_masking): Handle scatter stores too.
4144 (vect_get_store_rhs): Make public.
4145 (vectorizable_call): Use internal_store_fn_p.
4146 (vectorizable_store): Handle scatter store internal functions.
4147 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4148 when deciding whether the end of the group has been reached.
4149 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4150 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4151 (mask_scatter_store<mode>): New insns.
4153 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4154 Alan Hayward <alan.hayward@arm.com>
4155 David Sherwood <david.sherwood@arm.com>
4157 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4158 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4159 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4161 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4162 Use vect_truncate_gather_scatter_offset if we can't treat the
4163 operation as a normal gather load or scatter store.
4164 (get_group_load_store_type): Take the gather_scatter_info
4165 as argument. Try using a gather load or scatter store for
4166 single-element groups.
4167 (get_load_store_type): Update calls to get_group_load_store_type
4168 and vect_use_strided_gather_scatters_p.
4170 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4171 Alan Hayward <alan.hayward@arm.com>
4172 David Sherwood <david.sherwood@arm.com>
4174 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4175 optional tree argument.
4176 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4178 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4179 but continue to use the current value as a fallback.
4180 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4181 to compare the updates.
4182 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4183 (get_load_store_type): Use it when handling a strided access.
4184 (vect_get_strided_load_store_ops): New function.
4185 (vect_get_data_ptr_increment): Likewise.
4186 (vectorizable_load): Handle strided gather loads. Always pass
4187 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4189 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4190 Alan Hayward <alan.hayward@arm.com>
4191 David Sherwood <david.sherwood@arm.com>
4193 * doc/md.texi (gather_load@var{m}): Document.
4194 (mask_gather_load@var{m}): Likewise.
4195 * genopinit.c (main): Add supports_vec_gather_load and
4196 supports_vec_gather_load_cached to target_optabs.
4197 * optabs-tree.c (init_tree_optimization_optabs): Use
4198 ggc_cleared_alloc to allocate target_optabs.
4199 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4200 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4202 * internal-fn.h (internal_load_fn_p): Declare.
4203 (internal_gather_scatter_fn_p): Likewise.
4204 (internal_fn_mask_index): Likewise.
4205 (internal_gather_scatter_fn_supported_p): Likewise.
4206 * internal-fn.c (gather_load_direct): New macro.
4207 (expand_gather_load_optab_fn): New function.
4208 (direct_gather_load_optab_supported_p): New macro.
4209 (direct_internal_fn_optab): New function.
4210 (internal_load_fn_p): Likewise.
4211 (internal_gather_scatter_fn_p): Likewise.
4212 (internal_fn_mask_index): Likewise.
4213 (internal_gather_scatter_fn_supported_p): Likewise.
4214 * optabs-query.c (supports_at_least_one_mode_p): New function.
4215 (supports_vec_gather_load_p): Likewise.
4216 * optabs-query.h (supports_vec_gather_load_p): Declare.
4217 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4218 and memory_type field.
4219 (NUM_PATTERNS): Bump to 15.
4220 * tree-vect-data-refs.c: Include internal-fn.h.
4221 (vect_gather_scatter_fn_p): New function.
4222 (vect_describe_gather_scatter_call): Likewise.
4223 (vect_check_gather_scatter): Try using internal functions for
4224 gather loads. Recognize existing calls to a gather load function.
4225 (vect_analyze_data_refs): Consider using gather loads if
4226 supports_vec_gather_load_p.
4227 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4228 (vect_get_gather_scatter_offset_type): Likewise.
4229 (vect_convert_mask_for_vectype): Likewise.
4230 (vect_add_conversion_to_patterm): Likewise.
4231 (vect_try_gather_scatter_pattern): Likewise.
4232 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4233 (vect_vect_recog_func_ptrs): Add it.
4234 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4235 internal_fn_mask_index and internal_gather_scatter_fn_p.
4236 (check_load_store_masking): Take the gather_scatter_info as an
4237 argument and handle gather loads.
4238 (vect_get_gather_scatter_ops): New function.
4239 (vectorizable_call): Check internal_load_fn_p.
4240 (vectorizable_load): Likewise. Handle gather load internal
4242 (vectorizable_store): Update call to check_load_store_masking.
4243 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4244 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4245 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4246 (aarch64_gather_scale_operand_d): New predicates.
4247 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4248 (mask_gather_load<mode>): New insns.
4250 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4251 Alan Hayward <alan.hayward@arm.com>
4252 David Sherwood <david.sherwood@arm.com>
4254 * optabs.def (fold_left_plus_optab): New optab.
4255 * doc/md.texi (fold_left_plus_@var{m}): Document.
4256 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4257 * internal-fn.c (fold_left_direct): Define.
4258 (expand_fold_left_optab_fn): Likewise.
4259 (direct_fold_left_optab_supported_p): Likewise.
4260 * fold-const-call.c (fold_const_fold_left): New function.
4261 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4262 * tree-parloops.c (valid_reduction_p): New function.
4263 (gather_scalar_reductions): Use it.
4264 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4265 (vect_finish_replace_stmt): Declare.
4266 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4267 (needs_fold_left_reduction_p): New function, split out from...
4268 (vect_is_simple_reduction): ...here. Accept reductions that
4269 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4270 (vect_force_simple_reduction): Also store the reduction type in
4271 the assignment's STMT_VINFO_REDUC_TYPE.
4272 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4273 (merge_with_identity): New function.
4274 (vect_expand_fold_left): Likewise.
4275 (vectorize_fold_left_reduction): Likewise.
4276 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4277 scalar phi in place for it. Check for target support and reject
4278 cases that would reassociate the operation. Defer the transform
4279 phase to vectorize_fold_left_reduction.
4280 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4281 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4282 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4284 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4286 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4287 call to ifc_temp_var.
4289 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4290 Alan Hayward <alan.hayward@arm.com>
4291 David Sherwood <david.sherwood@arm.com>
4293 * target.def (legitimize_address_displacement): Take the original
4294 offset as a poly_int.
4295 * targhooks.h (default_legitimize_address_displacement): Update
4297 * targhooks.c (default_legitimize_address_displacement): Likewise.
4298 * doc/tm.texi: Regenerate.
4299 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4300 as an argument, moving assert of ad->disp == ad->disp_term to...
4301 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4302 Try calling targetm.legitimize_address_displacement before expanding
4303 the address rather than afterwards, and adjust for the new interface.
4304 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4305 Match the new hook interface. Handle SVE addresses.
4306 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4309 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4311 * Makefile.in (OBJS): Add early-remat.o.
4312 * target.def (select_early_remat_modes): New hook.
4313 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4314 * doc/tm.texi: Regenerate.
4315 * targhooks.h (default_select_early_remat_modes): Declare.
4316 * targhooks.c (default_select_early_remat_modes): New function.
4317 * timevar.def (TV_EARLY_REMAT): New timevar.
4318 * passes.def (pass_early_remat): New pass.
4319 * tree-pass.h (make_pass_early_remat): Declare.
4320 * early-remat.c: New file.
4321 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4323 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4325 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4326 Alan Hayward <alan.hayward@arm.com>
4327 David Sherwood <david.sherwood@arm.com>
4329 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4330 vfm1 with a bound_epilog parameter.
4331 (vect_do_peeling): Update calls accordingly, and move the prologue
4332 call earlier in the function. Treat the base bound_epilog as 0 for
4333 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4334 this base when peeling for gaps.
4335 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4336 with fully-masked loops.
4337 (vect_estimate_min_profitable_iters): Handle the single peeled
4338 iteration in that case.
4340 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4341 Alan Hayward <alan.hayward@arm.com>
4342 David Sherwood <david.sherwood@arm.com>
4344 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4345 single-element interleaving even if the size is not a power of 2.
4346 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4347 accesses for single-element interleaving if the group size is
4350 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4351 Alan Hayward <alan.hayward@arm.com>
4352 David Sherwood <david.sherwood@arm.com>
4354 * doc/md.texi (fold_extract_last_@var{m}): Document.
4355 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4356 * optabs.def (fold_extract_last_optab): New optab.
4357 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4358 * internal-fn.c (fold_extract_direct): New macro.
4359 (expand_fold_extract_optab_fn): Likewise.
4360 (direct_fold_extract_optab_supported_p): Likewise.
4361 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4362 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4363 EXTRACT_LAST_REDUCTION.
4364 (get_initial_def_for_reduction): Do not create an initial vector
4365 for EXTRACT_LAST_REDUCTION reductions.
4366 (vectorizable_reduction): Leave the scalar phi in place for
4367 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4368 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4369 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4370 transform phase to vectorizable_condition.
4371 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4373 (vect_finish_stmt_generation): ...here.
4374 (vect_finish_replace_stmt): New function.
4375 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4376 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4378 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4380 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4381 Alan Hayward <alan.hayward@arm.com>
4382 David Sherwood <david.sherwood@arm.com>
4384 * doc/md.texi (extract_last_@var{m}): Document.
4385 * optabs.def (extract_last_optab): New optab.
4386 * internal-fn.def (EXTRACT_LAST): New internal function.
4387 * internal-fn.c (cond_unary_direct): New macro.
4388 (expand_cond_unary_optab_fn): Likewise.
4389 (direct_cond_unary_optab_supported_p): Likewise.
4390 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4391 loops using EXTRACT_LAST.
4392 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4393 (extract_last_<mode>): ...this optab.
4394 (vec_extract<mode><Vel>): Update accordingly.
4396 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4397 Alan Hayward <alan.hayward@arm.com>
4398 David Sherwood <david.sherwood@arm.com>
4400 * target.def (empty_mask_is_expensive): New hook.
4401 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4402 * doc/tm.texi: Regenerate.
4403 * targhooks.h (default_empty_mask_is_expensive): Declare.
4404 * targhooks.c (default_empty_mask_is_expensive): New function.
4405 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4406 if the target says that empty masks are expensive.
4407 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4409 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4411 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4412 Alan Hayward <alan.hayward@arm.com>
4413 David Sherwood <david.sherwood@arm.com>
4415 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4416 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4417 (vect_use_loop_mask_for_alignment_p): New function.
4418 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4419 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4420 niters_skip argument. Make sure that the first niters_skip elements
4421 of the first iteration are inactive.
4422 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4423 Update call to vect_set_loop_masks_directly.
4424 (get_misalign_in_elems): New function, split out from...
4425 (vect_gen_prolog_loop_niters): ...here.
4426 (vect_update_init_of_dr): Take a code argument that specifies whether
4427 the adjustment should be added or subtracted.
4428 (vect_update_init_of_drs): Likewise.
4429 (vect_prepare_for_masked_peels): New function.
4430 (vect_do_peeling): Skip prologue peeling if we're using a mask
4431 instead. Update call to vect_update_inits_of_drs.
4432 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4434 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4435 alignment. Do not include the number of peeled iterations in
4436 the minimum threshold in that case.
4437 (vectorizable_induction): Adjust the start value down by
4438 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4439 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4440 Take the number of skipped iterations into account when calculating
4442 * tree-vect-stmts.c (vect_gen_while_not): New function.
4444 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4445 Alan Hayward <alan.hayward@arm.com>
4446 David Sherwood <david.sherwood@arm.com>
4448 * doc/sourcebuild.texi (vect_fully_masked): Document.
4449 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4451 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4453 (vect_analyze_loop_2): ...here. Don't check the vectorization
4454 factor against the number of loop iterations if the loop is
4457 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4458 Alan Hayward <alan.hayward@arm.com>
4459 David Sherwood <david.sherwood@arm.com>
4461 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4462 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4463 (dump_groups): Update accordingly.
4464 (iv_use::mem_type): New member variable.
4465 (address_p): New function.
4466 (record_use): Add a mem_type argument and initialize the new
4468 (record_group_use): Add a mem_type argument. Use address_p.
4469 Remove obsolete null checks of base_object. Update call to record_use.
4470 (find_interesting_uses_op): Update call to record_group_use.
4471 (find_interesting_uses_cond): Likewise.
4472 (find_interesting_uses_address): Likewise.
4473 (get_mem_type_for_internal_fn): New function.
4474 (find_address_like_use): Likewise.
4475 (find_interesting_uses_stmt): Try find_address_like_use before
4476 calling find_interesting_uses_op.
4477 (addr_offset_valid_p): Use the iv mem_type field as the type
4478 of the addressed memory.
4479 (add_autoinc_candidates): Likewise.
4480 (get_address_cost): Likewise.
4481 (split_small_address_groups_p): Use address_p.
4482 (split_address_groups): Likewise.
4483 (add_iv_candidate_for_use): Likewise.
4484 (autoinc_possible_for_pair): Likewise.
4485 (rewrite_groups): Likewise.
4486 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4487 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4488 (get_alias_ptr_type_for_ptr_address): New function.
4489 (rewrite_use_address): Rewrite address uses in calls that were
4490 identified by find_address_like_use.
4492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4493 Alan Hayward <alan.hayward@arm.com>
4494 David Sherwood <david.sherwood@arm.com>
4496 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4498 * gimple-expr.h (is_gimple_addressable: Likewise.
4499 * gimple-expr.c (is_gimple_address): Likewise.
4500 * internal-fn.c (expand_call_mem_ref): New function.
4501 (expand_mask_load_optab_fn): Use it.
4502 (expand_mask_store_optab_fn): Likewise.
4504 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4505 Alan Hayward <alan.hayward@arm.com>
4506 David Sherwood <david.sherwood@arm.com>
4508 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4509 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4510 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4511 (cond_umax@var{mode}): Document.
4512 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4513 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4514 (cond_umin_optab, cond_umax_optab): New optabs.
4515 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4516 (COND_IOR, COND_XOR): New internal functions.
4517 * internal-fn.h (get_conditional_internal_fn): Declare.
4518 * internal-fn.c (cond_binary_direct): New macro.
4519 (expand_cond_binary_optab_fn): Likewise.
4520 (direct_cond_binary_optab_supported_p): Likewise.
4521 (get_conditional_internal_fn): New function.
4522 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4523 Cope with reduction statements that are vectorized as calls rather
4525 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4526 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4527 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4528 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4529 (UNSPEC_COND_EOR): New unspecs.
4530 (optab): Add mappings for them.
4531 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4532 (sve_int_op, sve_fp_op): New int attributes.
4534 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4535 Alan Hayward <alan.hayward@arm.com>
4536 David Sherwood <david.sherwood@arm.com>
4538 * optabs.def (while_ult_optab): New optab.
4539 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4540 * internal-fn.def (WHILE_ULT): New internal function.
4541 * internal-fn.h (direct_internal_fn_supported_p): New override
4542 that takes two types as argument.
4543 * internal-fn.c (while_direct): New macro.
4544 (expand_while_optab_fn): New function.
4545 (convert_optab_supported_p): Likewise.
4546 (direct_while_optab_supported_p): New macro.
4547 * wide-int.h (wi::udiv_ceil): New function.
4548 * tree-vectorizer.h (rgroup_masks): New structure.
4549 (vec_loop_masks): New typedef.
4550 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4552 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4553 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4554 (vect_max_vf): New function.
4555 (slpeel_make_loop_iterate_ntimes): Delete.
4556 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4557 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4558 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4559 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4560 internal-fn.h, stor-layout.h and optabs-query.h.
4561 (vect_set_loop_mask): New function.
4562 (add_preheader_seq): Likewise.
4563 (add_header_seq): Likewise.
4564 (interleave_supported_p): Likewise.
4565 (vect_maybe_permute_loop_masks): Likewise.
4566 (vect_set_loop_masks_directly): Likewise.
4567 (vect_set_loop_condition_masked): Likewise.
4568 (vect_set_loop_condition_unmasked): New function, split out from
4569 slpeel_make_loop_iterate_ntimes.
4570 (slpeel_make_loop_iterate_ntimes): Rename to..
4571 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4572 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4573 (vect_do_peeling): Update call accordingly.
4574 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4576 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4577 mask_compare_type, can_fully_mask_p and fully_masked_p.
4578 (release_vec_loop_masks): New function.
4579 (_loop_vec_info): Use it to free the loop masks.
4580 (can_produce_all_loop_masks_p): New function.
4581 (vect_get_max_nscalars_per_iter): Likewise.
4582 (vect_verify_full_masking): Likewise.
4583 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4584 retries, and free the mask rgroups before retrying. Check loop-wide
4585 reasons for disallowing fully-masked loops. Make the final decision
4586 about whether use a fully-masked loop or not.
4587 (vect_estimate_min_profitable_iters): Do not assume that peeling
4588 for the number of iterations will be needed for fully-masked loops.
4589 (vectorizable_reduction): Disable fully-masked loops.
4590 (vectorizable_live_operation): Likewise.
4591 (vect_halve_mask_nunits): New function.
4592 (vect_double_mask_nunits): Likewise.
4593 (vect_record_loop_mask): Likewise.
4594 (vect_get_loop_mask): Likewise.
4595 (vect_transform_loop): Handle the case in which the final loop
4596 iteration might handle a partial vector. Call vect_set_loop_condition
4597 instead of slpeel_make_loop_iterate_ntimes.
4598 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4599 (check_load_store_masking): New function.
4600 (prepare_load_store_mask): Likewise.
4601 (vectorizable_store): Handle fully-masked loops.
4602 (vectorizable_load): Likewise.
4603 (supportable_widening_operation): Use vect_halve_mask_nunits for
4605 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4606 (vect_gen_while): New function.
4607 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4608 (aarch64_uqdec<mode>): New insn.
4610 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4611 Alan Hayward <alan.hayward@arm.com>
4612 David Sherwood <david.sherwood@arm.com>
4614 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4615 (reduc_xor_scal_optab): New optabs.
4616 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4617 (reduc_xor_scal_@var{m}): Document.
4618 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4619 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4621 * fold-const-call.c (fold_const_call): Handle them.
4622 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4623 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4624 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4625 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4626 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4627 (UNSPEC_XORV): New unspecs.
4628 (optab): Add entries for them.
4629 (BITWISEV): New int iterator.
4630 (bit_reduc_op): New int attributes.
4632 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4633 Alan Hayward <alan.hayward@arm.com>
4634 David Sherwood <david.sherwood@arm.com>
4636 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4637 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4638 * optabs.def (vec_shl_insert_optab): New optab.
4639 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4640 (duplicate_and_interleave): Likewise.
4641 * tree-vect-loop.c: Include internal-fn.h.
4642 (neutral_op_for_slp_reduction): New function, split out from
4643 get_initial_defs_for_reduction.
4644 (get_initial_def_for_reduction): Handle option 2 for variable-length
4645 vectors by loading the neutral value into a vector and then shifting
4646 the initial value into element 0.
4647 (get_initial_defs_for_reduction): Replace the code argument with
4648 the neutral value calculated by neutral_op_for_slp_reduction.
4649 Use gimple_build_vector for constant-length vectors.
4650 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4651 but the first group_size elements have a neutral value.
4652 Use duplicate_and_interleave otherwise.
4653 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4654 Update call to get_initial_defs_for_reduction. Handle SLP
4655 reductions for variable-length vectors by creating one vector
4656 result for each scalar result, with the elements associated
4657 with other scalar results stubbed out with the neutral value.
4658 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4659 Require IFN_VEC_SHL_INSERT for double reductions on
4660 variable-length vectors, or SLP reductions that have
4661 a neutral value. Require can_duplicate_and_interleave_p
4662 support for variable-length unchained SLP reductions if there
4663 is no neutral value, such as for MIN/MAX reductions. Also require
4664 the number of vector elements to be a multiple of the number of
4665 SLP statements when doing variable-length unchained SLP reductions.
4666 Update call to vect_create_epilog_for_reduction.
4667 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4668 and remove initial values.
4669 (duplicate_and_interleave): Make public.
4670 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4671 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4673 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4674 Alan Hayward <alan.hayward@arm.com>
4675 David Sherwood <david.sherwood@arm.com>
4677 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4678 (can_duplicate_and_interleave_p): New function.
4679 (vect_get_and_check_slp_defs): Take the vector of statements
4680 rather than just the current one. Remove excess parentheses.
4681 Restriction rejectinon of vect_constant_def and vect_external_def
4682 for variable-length vectors to boolean types, or types for which
4683 can_duplicate_and_interleave_p is false.
4684 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4685 (duplicate_and_interleave): New function.
4686 (vect_get_constant_vectors): Use gimple_build_vector for
4687 constant-length vectors and suitable variable-length constant
4688 vectors. Use duplicate_and_interleave for other variable-length
4689 vectors. Don't defer the update when inserting new statements.
4691 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4692 Alan Hayward <alan.hayward@arm.com>
4693 David Sherwood <david.sherwood@arm.com>
4695 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4696 min_profitable_iters doesn't go negative.
4698 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4699 Alan Hayward <alan.hayward@arm.com>
4700 David Sherwood <david.sherwood@arm.com>
4702 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4703 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4704 * optabs.def (vec_mask_load_lanes_optab): New optab.
4705 (vec_mask_store_lanes_optab): Likewise.
4706 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4707 (MASK_STORE_LANES): Likewise.
4708 * internal-fn.c (mask_load_lanes_direct): New macro.
4709 (mask_store_lanes_direct): Likewise.
4710 (expand_mask_load_optab_fn): Handle masked operations.
4711 (expand_mask_load_lanes_optab_fn): New macro.
4712 (expand_mask_store_optab_fn): Handle masked operations.
4713 (expand_mask_store_lanes_optab_fn): New macro.
4714 (direct_mask_load_lanes_optab_supported_p): Likewise.
4715 (direct_mask_store_lanes_optab_supported_p): Likewise.
4716 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4718 (vect_load_lanes_supported): Likewise.
4719 * tree-vect-data-refs.c (strip_conversion): New function.
4720 (can_group_stmts_p): Likewise.
4721 (vect_analyze_data_ref_accesses): Use it instead of checking
4722 for a pair of assignments.
4723 (vect_store_lanes_supported): Take a masked_p parameter.
4724 (vect_load_lanes_supported): Likewise.
4725 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4726 vect_store_lanes_supported and vect_load_lanes_supported.
4727 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4728 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4729 parameter. Don't allow gaps for masked accesses.
4730 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4731 and vect_load_lanes_supported.
4732 (get_load_store_type): Take a masked_p parameter and update
4733 call to get_group_load_store_type.
4734 (vectorizable_store): Update call to get_load_store_type.
4735 Handle IFN_MASK_STORE_LANES.
4736 (vectorizable_load): Update call to get_load_store_type.
4737 Handle IFN_MASK_LOAD_LANES.
4739 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4740 Alan Hayward <alan.hayward@arm.com>
4741 David Sherwood <david.sherwood@arm.com>
4743 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4745 * config/aarch64/aarch64-protos.h
4746 (aarch64_sve_struct_memory_operand_p): Declare.
4747 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4748 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4749 (VPRED, vpred): Handle SVE structure modes.
4750 * config/aarch64/constraints.md (Utx): New constraint.
4751 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4752 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4753 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4754 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4755 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4756 structure modes. Split into pieces after RA.
4757 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4758 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4760 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4761 SVE structure modes.
4762 (aarch64_classify_address): Likewise.
4763 (sizetochar): Move earlier in file.
4764 (aarch64_print_operand): Handle SVE register lists.
4765 (aarch64_array_mode): New function.
4766 (aarch64_sve_struct_memory_operand_p): Likewise.
4767 (TARGET_ARRAY_MODE): Redefine.
4769 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4770 Alan Hayward <alan.hayward@arm.com>
4771 David Sherwood <david.sherwood@arm.com>
4773 * target.def (array_mode): New target hook.
4774 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4775 * doc/tm.texi: Regenerate.
4776 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4777 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4778 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4780 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4783 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4784 Alan Hayward <alan.hayward@arm.com>
4785 David Sherwood <david.sherwood@arm.com>
4787 * fold-const.c (fold_binary_loc): Check the argument types
4788 rather than the result type when testing for a vector operation.
4790 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4792 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4793 * doc/tm.texi: Regenerate.
4795 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4796 Alan Hayward <alan.hayward@arm.com>
4797 David Sherwood <david.sherwood@arm.com>
4799 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4800 (sve): Document new AArch64 extension.
4801 * doc/md.texi (w): Extend the description of the AArch64
4802 constraint to include SVE vectors.
4803 (Upl, Upa): Document new AArch64 predicate constraints.
4804 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4806 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4807 (msve-vector-bits=): New option.
4808 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4809 SVE when these are disabled.
4810 (sve): New extension.
4811 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4812 modes. Adjust their number of units based on aarch64_sve_vg.
4813 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4814 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4815 aarch64_addr_query_type.
4816 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4817 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4818 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4819 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4820 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4821 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4822 (aarch64_simd_imm_zero_p): Delete.
4823 (aarch64_check_zero_based_sve_index_immediate): Declare.
4824 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4825 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4826 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4827 (aarch64_sve_float_mul_immediate_p): Likewise.
4828 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4830 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4831 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4832 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4833 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4834 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4835 (aarch64_regmode_natural_size): Likewise.
4836 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4837 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4839 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4840 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4841 for VG and the SVE predicate registers.
4842 (V_ALIASES): Add a "z"-prefixed alias.
4843 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4844 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4845 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4846 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4847 (REG_CLASS_NAMES): Add entries for them.
4848 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4849 and the predicate registers.
4850 (aarch64_sve_vg): Declare.
4851 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4852 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4853 (REGMODE_NATURAL_SIZE): Define.
4854 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4856 * config/aarch64/aarch64.c: Include cfgrtl.h.
4857 (simd_immediate_info): Add a constructor for series vectors,
4858 and an associated step field.
4859 (aarch64_sve_vg): New variable.
4860 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4861 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4862 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4863 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4864 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4865 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4866 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4867 (aarch64_get_mask_mode): New functions.
4868 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4869 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4870 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4871 predicate modes and predicate registers. Explicitly restrict
4872 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4873 to store a vector mode if it is recognized by
4874 aarch64_classify_vector_mode.
4875 (aarch64_regmode_natural_size): New function.
4876 (aarch64_hard_regno_caller_save_mode): Return the original mode
4878 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4879 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4880 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4881 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4883 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4884 does not overlap dest if the function is frame-related. Handle
4886 (aarch64_split_add_offset): New function.
4887 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4888 them aarch64_add_offset.
4889 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4890 and update call to aarch64_sub_sp.
4891 (aarch64_add_cfa_expression): New function.
4892 (aarch64_expand_prologue): Pass extra temporary registers to the
4893 functions above. Handle the case in which we need to emit new
4894 DW_CFA_expressions for registers that were originally saved
4895 relative to the stack pointer, but now have to be expressed
4896 relative to the frame pointer.
4897 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4899 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4900 IP0 and IP1 values for SVE frames.
4901 (aarch64_expand_vec_series): New function.
4902 (aarch64_expand_sve_widened_duplicate): Likewise.
4903 (aarch64_expand_sve_const_vector): Likewise.
4904 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4905 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4906 into the register, rather than emitting a SET directly.
4907 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4908 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4909 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4910 (offset_9bit_signed_scaled_p): New functions.
4911 (aarch64_replicate_bitmask_imm): New function.
4912 (aarch64_bitmask_imm): Use it.
4913 (aarch64_cannot_force_const_mem): Reject expressions involving
4914 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4915 (aarch64_classify_index): Handle SVE indices, by requiring
4916 a plain register index with a scale that matches the element size.
4917 (aarch64_classify_address): Handle SVE addresses. Assert that
4918 the mode of the address is VOIDmode or an integer mode.
4919 Update call to aarch64_classify_symbol.
4920 (aarch64_classify_symbolic_expression): Update call to
4921 aarch64_classify_symbol.
4922 (aarch64_const_vec_all_in_range_p): New function.
4923 (aarch64_print_vector_float_operand): Likewise.
4924 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4925 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4926 and the FP immediates 1.0 and 0.5.
4927 (aarch64_print_address_internal): Handle SVE addresses.
4928 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4929 (aarch64_regno_regclass): Handle predicate registers.
4930 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4932 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4933 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4934 (aarch64_convert_sve_vector_bits): New function.
4935 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4936 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4938 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4939 Handle SVE vector and predicate modes. Accept VL-based constants
4940 that need only one temporary register, and VL offsets that require
4941 no temporary registers.
4942 (aarch64_conditional_register_usage): Mark the predicate registers
4943 as fixed if SVE isn't available.
4944 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4945 Return true for SVE vector and predicate modes.
4946 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4947 rather than an unsigned int. Handle SVE modes.
4948 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4950 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4952 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4953 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4954 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4955 (aarch64_sve_float_mul_immediate_p): New functions.
4956 (aarch64_sve_valid_immediate): New function.
4957 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4958 Explicitly reject structure modes. Check for INDEX constants.
4959 Handle PTRUE and PFALSE constants.
4960 (aarch64_check_zero_based_sve_index_immediate): New function.
4961 (aarch64_simd_imm_zero_p): Delete.
4962 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4963 vector modes. Accept constants in the range of CNT[BHWD].
4964 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4965 ask for an Advanced SIMD mode.
4966 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4967 (aarch64_simd_vector_alignment): Handle SVE predicates.
4968 (aarch64_vectorize_preferred_vector_alignment): New function.
4969 (aarch64_simd_vector_alignment_reachable): Use it instead of
4971 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4972 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4974 (MAX_VECT_LEN): Delete.
4975 (expand_vec_perm_d): Add a vec_flags field.
4976 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4977 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4978 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4980 (aarch64_evpc_rev): Rename to...
4981 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4982 (aarch64_evpc_rev_global): New function.
4983 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4984 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4986 (aarch64_evpc_sve_tbl): New function.
4987 (aarch64_expand_vec_perm_const_1): Update after rename of
4988 aarch64_evpc_rev. Handle SVE permutes too, trying
4989 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4990 than aarch64_evpc_tbl.
4991 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4992 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4993 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4994 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4995 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4996 (aarch64_expand_sve_vcond): New functions.
4997 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4998 of aarch64_vector_mode_p.
4999 (aarch64_dwarf_poly_indeterminate_value): New function.
5000 (aarch64_compute_pressure_classes): Likewise.
5001 (aarch64_can_change_mode_class): Likewise.
5002 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5003 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5004 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5005 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5006 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5007 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5008 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5009 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5011 (Dn, Dl, Dr): Accept const as well as const_vector.
5012 (Dz): Likewise. Compare against CONST0_RTX.
5013 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5014 of "vector" where appropriate.
5015 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5016 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5017 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5018 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5019 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5020 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5021 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5022 (v_int_equiv): Extend to SVE modes.
5023 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5025 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5026 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5027 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5028 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5029 (SVE_COND_FP_CMP): New int iterators.
5030 (perm_hilo): Handle the new unpack unspecs.
5031 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5033 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5034 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5035 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5036 (aarch64_equality_operator, aarch64_constant_vector_operand)
5037 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5038 (aarch64_sve_nonimmediate_operand): Likewise.
5039 (aarch64_sve_general_operand): Likewise.
5040 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5041 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5042 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5043 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5044 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5045 (aarch64_sve_float_arith_immediate): Likewise.
5046 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5047 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5048 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5049 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5050 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5051 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5052 (aarch64_sve_float_arith_operand): Likewise.
5053 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5054 (aarch64_sve_float_mul_operand): Likewise.
5055 (aarch64_sve_vec_perm_operand): Likewise.
5056 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5057 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5058 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5059 as well as const_vector.
5060 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5061 in file. Use CONST0_RTX and CONSTM1_RTX.
5062 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5063 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5064 Use aarch64_simd_imm_zero.
5065 * config/aarch64/aarch64-sve.md: New file.
5066 * config/aarch64/aarch64.md: Include it.
5067 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5068 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5069 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5070 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5071 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5072 (sve): New attribute.
5073 (enabled): Disable instructions with the sve attribute unless
5075 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5076 aarch64_expand_mov_immediate.
5077 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5078 CNT[BHSD] immediates.
5079 (movti): Split CONST_POLY_INT moves into two halves.
5080 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5081 Split additions that need a temporary here if the destination
5082 is the stack pointer.
5083 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5084 (*add<mode>3_poly_1): New instruction.
5085 (set_clobber_cc): New expander.
5087 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5089 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5090 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5091 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5092 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5093 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5094 Change innermode from fixed_mode_size to machine_mode.
5095 (simplify_subreg): Update call accordingly. Handle a constant-sized
5096 subreg of a variable-length CONST_VECTOR.
5098 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5099 Alan Hayward <alan.hayward@arm.com>
5100 David Sherwood <david.sherwood@arm.com>
5102 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5103 (add_offset_to_base): New function, split out from...
5104 (create_mem_ref): ...here. When handling a scale other than 1,
5105 check first whether the address is valid without the offset.
5106 Add it into the base if so, leaving the index and scale as-is.
5108 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5111 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5112 fold_for_warn before checking if arg2 is INTEGER_CST.
5114 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5116 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5117 (store_multiple_operation): Delete.
5118 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5119 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5120 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5121 guarded by TARGET_STRING.
5122 (rs6000_output_load_multiple): Delete.
5123 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5124 OPTION_MASK_STRING / TARGET_STRING handling.
5125 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5126 (const rs6000_opt_masks) <"string">: Change mask to 0.
5127 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5128 (MASK_STRING): Delete.
5129 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5131 (load_multiple): Delete.
5138 (store_multiple): Delete.
5145 (movmemsi_8reg): Delete.
5146 (corresponding unnamed define_insn): Delete.
5147 (movmemsi_6reg): Delete.
5148 (corresponding unnamed define_insn): Delete.
5149 (movmemsi_4reg): Delete.
5150 (corresponding unnamed define_insn): Delete.
5151 (movmemsi_2reg): Delete.
5152 (corresponding unnamed define_insn): Delete.
5153 (movmemsi_1reg): Delete.
5154 (corresponding unnamed define_insn): Delete.
5155 * config/rs6000/rs6000.opt (mno-string): New.
5156 (mstring): Replace by deprecation warning stub.
5157 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5159 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5161 * regrename.c (regrename_do_replace): If replacing the same
5162 reg multiple times, try to reuse last created gen_raw_REG.
5165 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5166 main to workaround a bug in GDB.
5168 2018-01-12 Tom de Vries <tom@codesourcery.com>
5171 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5173 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5175 PR rtl-optimization/80481
5176 * ira-color.c (get_cap_member): New function.
5177 (allocnos_conflict_by_live_ranges_p): Use it.
5178 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5179 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5181 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5184 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5185 (*saddl_se_1): Ditto.
5187 (*ssubl_se_1): Ditto.
5189 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5191 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5192 rather than wi::to_widest for DR_INITs.
5193 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5194 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5195 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5197 (vect_analyze_group_access_1): Note that here.
5199 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5201 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5202 polynomial type sizes.
5204 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5206 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5207 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5208 (gimple_add_tmp_var): Likewise.
5210 2018-01-12 Martin Liska <mliska@suse.cz>
5212 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5213 (gimple_alloc_sizes): Likewise.
5214 (dump_gimple_statistics): Use PRIu64 in printf format.
5215 * gimple.h: Change uint64_t to int.
5217 2018-01-12 Martin Liska <mliska@suse.cz>
5219 * tree-core.h: Use uint64_t instead of int.
5220 * tree.c (tree_node_counts): Likewise.
5221 (tree_node_sizes): Likewise.
5222 (dump_tree_statistics): Use PRIu64 in printf format.
5224 2018-01-12 Martin Liska <mliska@suse.cz>
5226 * Makefile.in: As qsort_chk is implemented in vec.c, add
5227 vec.o to linkage of gencfn-macros.
5228 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5229 passing the info to record_node_allocation_statistics.
5230 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5232 * ggc-common.c (struct ggc_usage): Add operator== and use
5233 it in operator< and compare function.
5234 * mem-stats.h (struct mem_usage): Likewise.
5235 * vec.c (struct vec_usage): Remove operator< and compare
5236 function. Can be simply inherited.
5238 2018-01-12 Martin Jambor <mjambor@suse.cz>
5241 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5242 * tree-ssa-math-opts.c: Include domwalk.h.
5243 (convert_mult_to_fma_1): New function.
5244 (fma_transformation_info): New type.
5245 (fma_deferring_state): Likewise.
5246 (cancel_fma_deferring): New function.
5247 (result_of_phi): Likewise.
5248 (last_fma_candidate_feeds_initial_phi): Likewise.
5249 (convert_mult_to_fma): Added deferring logic, split actual
5250 transformation to convert_mult_to_fma_1.
5251 (math_opts_dom_walker): New type.
5252 (math_opts_dom_walker::after_dom_children): New method, body moved
5253 here from pass_optimize_widening_mul::execute, added deferring logic
5255 (pass_optimize_widening_mul::execute): Moved most of code to
5256 math_opts_dom_walker::after_dom_children.
5257 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5258 * config/i386/i386.c (ix86_option_override_internal): Added
5259 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5261 2018-01-12 Richard Biener <rguenther@suse.de>
5264 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5265 inline instance vars.
5267 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5270 * config/rx/rx.c (rx_is_restricted_memory_address):
5273 2018-01-12 Richard Biener <rguenther@suse.de>
5275 PR tree-optimization/80846
5276 * target.def (split_reduction): New target hook.
5277 * targhooks.c (default_split_reduction): New function.
5278 * targhooks.h (default_split_reduction): Declare.
5279 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5280 target requests first reduce vectors by combining low and high
5282 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5283 (get_vectype_for_scalar_type_and_size): Export.
5284 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5285 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5286 * doc/tm.texi: Regenerate.
5287 * config/i386/i386.c (ix86_split_reduction): Implement
5288 TARGET_VECTORIZE_SPLIT_REDUCTION.
5290 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5293 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5294 in PIC mode except for TARGET_VXWORKS_RTP.
5295 * config/sparc/sparc.c: Include cfgrtl.h.
5296 (TARGET_INIT_PIC_REG): Define.
5297 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5298 (sparc_pic_register_p): New predicate.
5299 (sparc_legitimate_address_p): Use it.
5300 (sparc_legitimize_pic_address): Likewise.
5301 (sparc_delegitimize_address): Likewise.
5302 (sparc_mode_dependent_address_p): Likewise.
5303 (gen_load_pcrel_sym): Remove 4th parameter.
5304 (load_got_register): Adjust call to above. Remove obsolete stuff.
5305 (sparc_expand_prologue): Do not call load_got_register here.
5306 (sparc_flat_expand_prologue): Likewise.
5307 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5308 (sparc_use_pseudo_pic_reg): New function.
5309 (sparc_init_pic_reg): Likewise.
5310 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5311 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5313 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5315 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5316 Add item for branch_cost.
5318 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5320 PR rtl-optimization/83565
5321 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5322 not extend the result to a larger mode for rotate operations.
5323 (num_sign_bit_copies1): Likewise.
5325 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5328 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5330 Use values-Xc.o for -pedantic.
5331 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5333 2018-01-12 Martin Liska <mliska@suse.cz>
5336 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5338 (possible_polymorphic_call_targets): Use it.
5339 (ipa_devirt): Likewise.
5341 2018-01-12 Martin Liska <mliska@suse.cz>
5343 * profile-count.h (enum profile_quality): Use 0 as invalid
5344 enum value of profile_quality.
5346 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5348 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5349 -mext-string options.
5351 2018-01-12 Richard Biener <rguenther@suse.de>
5353 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5354 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5355 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5357 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5359 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5361 * configure.ac (--with-long-double-format): Add support for the
5362 configuration option to change the default long double format on
5364 * config.gcc (powerpc*-linux*-*): Likewise.
5365 * configure: Regenerate.
5366 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5367 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5368 used without modification.
5370 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5372 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5373 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5374 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5375 MISC_BUILTIN_SPEC_BARRIER.
5376 (rs6000_init_builtins): Likewise.
5377 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5379 (speculation_barrier): New define_insn.
5380 * doc/extend.texi: Document __builtin_speculation_barrier.
5382 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5385 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5386 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5387 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5389 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5390 integral modes instead of "ss" and "sd".
5391 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5392 vectors with 32-bit and 64-bit elements.
5393 (vecdupssescalarmodesuffix): New mode attribute.
5394 (vec_dup<mode>): Use it.
5396 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5399 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5400 frame if argument is passed on stack.
5402 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5405 * ree.c (combine_reaching_defs): Optimize also
5406 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5407 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5409 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5412 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5414 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5417 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5418 after they are computed.
5420 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5422 PR tree-optimization/83695
5423 * gimple-loop-linterchange.cc
5424 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5425 reset cached scev information after interchange.
5426 (pass_linterchange::execute): Remove call to scev_reset_htab.
5428 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5430 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5431 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5432 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5433 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5434 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5435 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5436 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5437 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5438 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5439 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5440 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5441 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5442 (V_lane_reg): Likewise.
5443 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5445 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5446 (vfmal_lane_low<mode>_intrinsic,
5447 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5448 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5449 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5450 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5451 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5452 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5454 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5456 * config/arm/arm-cpus.in (fp16fml): New feature.
5457 (ALL_SIMD): Add fp16fml.
5458 (armv8.2-a): Add fp16fml as an option.
5459 (armv8.3-a): Likewise.
5460 (armv8.4-a): Add fp16fml as part of fp16.
5461 * config/arm/arm.h (TARGET_FP16FML): Define.
5462 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5464 * config/arm/arm-modes.def (V2HF): Define.
5465 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5466 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5467 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5468 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5469 vfmsl_low, vfmsl_high): New set of builtins.
5470 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5471 (vfml_op): New code attribute.
5472 (VFMLHALVES): New int iterator.
5473 (VFML, VFMLSEL): New mode attributes.
5474 (V_reg): Define mapping for V2HF.
5475 (V_hi, V_lo): New mode attributes.
5476 (VF_constraint): Likewise.
5477 (vfml_half, vfml_half_selector): New int attributes.
5478 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5480 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5481 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5483 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5484 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5485 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5486 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5488 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5489 Document new effective target and option set.
5491 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5493 * config/arm/arm-cpus.in (armv8_4): New feature.
5494 (ARMv8_4a): New fgroup.
5495 (armv8.4-a): New arch.
5496 * config/arm/arm-tables.opt: Regenerate.
5497 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5498 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5499 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5500 Add matching rules for -march=armv8.4-a and extensions.
5501 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5503 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5506 * config/rx/rx.md (BW): New mode attribute.
5507 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5509 2018-01-11 Richard Biener <rguenther@suse.de>
5511 PR tree-optimization/83435
5512 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5513 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5514 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5516 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5517 Alan Hayward <alan.hayward@arm.com>
5518 David Sherwood <david.sherwood@arm.com>
5520 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5522 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5523 (aarch64_print_address_internal): Use it to check for a zero offset.
5525 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5526 Alan Hayward <alan.hayward@arm.com>
5527 David Sherwood <david.sherwood@arm.com>
5529 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5530 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5531 Return a poly_int64 rather than a HOST_WIDE_INT.
5532 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5533 rather than a HOST_WIDE_INT.
5534 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5535 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5536 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5537 final_offset from HOST_WIDE_INT to poly_int64.
5538 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5539 to_constant when getting the number of units in an Advanced SIMD
5541 (aarch64_builtin_vectorized_function): Check for a constant number
5543 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5545 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5546 attribute instead of GET_MODE_NUNITS.
5547 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5548 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5549 GET_MODE_SIZE for fixed-size registers.
5550 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5551 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5552 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5553 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5554 (aarch64_print_operand, aarch64_print_address_internal)
5555 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5556 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5557 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5558 Handle polynomial GET_MODE_SIZE.
5559 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5560 wider than SImode without modification.
5561 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5562 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5563 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5564 passing and returning SVE modes.
5565 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5566 rather than GEN_INT.
5567 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5568 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5569 (aarch64_allocate_and_probe_stack_space): Likewise.
5570 (aarch64_layout_frame): Cope with polynomial offsets.
5571 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5572 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5574 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5575 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5576 poly_int64 rather than a HOST_WIDE_INT.
5577 (aarch64_get_separate_components, aarch64_process_components)
5578 (aarch64_expand_prologue, aarch64_expand_epilogue)
5579 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5580 (aarch64_anchor_offset): New function, split out from...
5581 (aarch64_legitimize_address): ...here.
5582 (aarch64_builtin_vectorization_cost): Handle polynomial
5583 TYPE_VECTOR_SUBPARTS.
5584 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5586 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5587 number of elements from the PARALLEL rather than the mode.
5588 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5589 rather than GET_MODE_BITSIZE.
5590 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5591 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5592 (aarch64_expand_vec_perm_const_1): Handle polynomial
5593 d->perm.length () and d->perm elements.
5594 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5595 Apply to_constant to d->perm elements.
5596 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5597 polynomial CONST_VECTOR_NUNITS.
5598 (aarch64_move_pointer): Take amount as a poly_int64 rather
5600 (aarch64_progress_pointer): Avoid temporary variable.
5601 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5602 the mode attribute instead of GET_MODE.
5604 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5605 Alan Hayward <alan.hayward@arm.com>
5606 David Sherwood <david.sherwood@arm.com>
5608 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5609 x exists before using it.
5610 (aarch64_add_constant_internal): Rename to...
5611 (aarch64_add_offset_1): ...this. Replace regnum with separate
5612 src and dest rtxes. Handle the case in which they're different,
5613 including when the offset is zero. Replace scratchreg with an rtx.
5614 Use 2 additions if there is no spare register into which we can
5615 move a 16-bit constant.
5616 (aarch64_add_constant): Delete.
5617 (aarch64_add_offset): Replace reg with separate src and dest
5618 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5619 Use aarch64_add_offset_1.
5620 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5621 an rtx rather than an int. Take the delta as a poly_int64
5622 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5623 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5624 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5625 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5626 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5628 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5629 aarch64_add_constant.
5631 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5633 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5634 Use scalar_float_mode.
5636 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5638 * config/aarch64/aarch64-simd.md
5639 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5640 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5641 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5642 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5643 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5644 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5645 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5646 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5647 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5648 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5650 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5653 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5654 targ_options->x_arm_arch_string is non NULL.
5656 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5658 * config/aarch64/aarch64.h
5659 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5661 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5664 * expmed.c (emit_store_flag_force): Swap if const op0
5665 and change VOIDmode to mode of op0.
5667 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5669 PR rtl-optimization/83761
5670 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5671 than bytes to mode_for_size.
5673 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5676 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5677 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5680 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5683 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5684 when in layout mode.
5685 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5686 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5689 2018-01-10 Michael Collison <michael.collison@arm.com>
5691 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5692 * config/aarch64/aarch64-option-extension.def: Add
5693 AARCH64_OPT_EXTENSION of 'fp16fml'.
5694 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5695 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5696 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5697 * config/aarch64/constraints.md (Ui7): New constraint.
5698 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5699 (VFMLA_SEL_W): Ditto.
5702 (VFMLA16_LOW): New int iterator.
5703 (VFMLA16_HIGH): Ditto.
5704 (UNSPEC_FMLAL): New unspec.
5705 (UNSPEC_FMLSL): Ditto.
5706 (UNSPEC_FMLAL2): Ditto.
5707 (UNSPEC_FMLSL2): Ditto.
5708 (f16mac): New code attribute.
5709 * config/aarch64/aarch64-simd-builtins.def
5710 (aarch64_fmlal_lowv2sf): Ditto.
5711 (aarch64_fmlsl_lowv2sf): Ditto.
5712 (aarch64_fmlalq_lowv4sf): Ditto.
5713 (aarch64_fmlslq_lowv4sf): Ditto.
5714 (aarch64_fmlal_highv2sf): Ditto.
5715 (aarch64_fmlsl_highv2sf): Ditto.
5716 (aarch64_fmlalq_highv4sf): Ditto.
5717 (aarch64_fmlslq_highv4sf): Ditto.
5718 (aarch64_fmlal_lane_lowv2sf): Ditto.
5719 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5720 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5721 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5722 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5723 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5724 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5725 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5726 (aarch64_fmlal_lane_highv2sf): Ditto.
5727 (aarch64_fmlsl_lane_highv2sf): Ditto.
5728 (aarch64_fmlal_laneq_highv2sf): Ditto.
5729 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5730 (aarch64_fmlalq_lane_highv4sf): Ditto.
5731 (aarch64_fmlsl_lane_highv4sf): Ditto.
5732 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5733 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5734 * config/aarch64/aarch64-simd.md:
5735 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5736 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5737 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5738 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5739 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5740 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5741 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5742 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5743 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5744 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5745 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5746 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5747 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5748 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5749 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5750 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5751 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5752 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5753 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5754 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5755 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5756 (vfmlsl_low_u32): Ditto.
5757 (vfmlalq_low_u32): Ditto.
5758 (vfmlslq_low_u32): Ditto.
5759 (vfmlal_high_u32): Ditto.
5760 (vfmlsl_high_u32): Ditto.
5761 (vfmlalq_high_u32): Ditto.
5762 (vfmlslq_high_u32): Ditto.
5763 (vfmlal_lane_low_u32): Ditto.
5764 (vfmlsl_lane_low_u32): Ditto.
5765 (vfmlal_laneq_low_u32): Ditto.
5766 (vfmlsl_laneq_low_u32): Ditto.
5767 (vfmlalq_lane_low_u32): Ditto.
5768 (vfmlslq_lane_low_u32): Ditto.
5769 (vfmlalq_laneq_low_u32): Ditto.
5770 (vfmlslq_laneq_low_u32): Ditto.
5771 (vfmlal_lane_high_u32): Ditto.
5772 (vfmlsl_lane_high_u32): Ditto.
5773 (vfmlal_laneq_high_u32): Ditto.
5774 (vfmlsl_laneq_high_u32): Ditto.
5775 (vfmlalq_lane_high_u32): Ditto.
5776 (vfmlslq_lane_high_u32): Ditto.
5777 (vfmlalq_laneq_high_u32): Ditto.
5778 (vfmlslq_laneq_high_u32): Ditto.
5779 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5780 (AARCH64_FL_FOR_ARCH8_4): New.
5781 (AARCH64_ISA_F16FML): New ISA flag.
5782 (TARGET_F16FML): New feature flag for fp16fml.
5783 (doc/invoke.texi): Document new fp16fml option.
5785 2018-01-10 Michael Collison <michael.collison@arm.com>
5787 * config/aarch64/aarch64-builtins.c:
5788 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5789 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5790 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5791 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5792 (AARCH64_ISA_SHA3): New ISA flag.
5793 (TARGET_SHA3): New feature flag for sha3.
5794 * config/aarch64/iterators.md (sha512_op): New int attribute.
5795 (CRYPTO_SHA512): New int iterator.
5796 (UNSPEC_SHA512H): New unspec.
5797 (UNSPEC_SHA512H2): Ditto.
5798 (UNSPEC_SHA512SU0): Ditto.
5799 (UNSPEC_SHA512SU1): Ditto.
5800 * config/aarch64/aarch64-simd-builtins.def
5801 (aarch64_crypto_sha512hqv2di): New builtin.
5802 (aarch64_crypto_sha512h2qv2di): Ditto.
5803 (aarch64_crypto_sha512su0qv2di): Ditto.
5804 (aarch64_crypto_sha512su1qv2di): Ditto.
5805 (aarch64_eor3qv8hi): Ditto.
5806 (aarch64_rax1qv2di): Ditto.
5807 (aarch64_xarqv2di): Ditto.
5808 (aarch64_bcaxqv8hi): Ditto.
5809 * config/aarch64/aarch64-simd.md:
5810 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5811 (aarch64_crypto_sha512su0qv2di): Ditto.
5812 (aarch64_crypto_sha512su1qv2di): Ditto.
5813 (aarch64_eor3qv8hi): Ditto.
5814 (aarch64_rax1qv2di): Ditto.
5815 (aarch64_xarqv2di): Ditto.
5816 (aarch64_bcaxqv8hi): Ditto.
5817 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5818 (vsha512h2q_u64): Ditto.
5819 (vsha512su0q_u64): Ditto.
5820 (vsha512su1q_u64): Ditto.
5821 (veor3q_u16): Ditto.
5822 (vrax1q_u64): Ditto.
5824 (vbcaxq_u16): Ditto.
5825 * config/arm/types.md (crypto_sha512): New type attribute.
5826 (crypto_sha3): Ditto.
5827 (doc/invoke.texi): Document new sha3 option.
5829 2018-01-10 Michael Collison <michael.collison@arm.com>
5831 * config/aarch64/aarch64-builtins.c:
5832 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5833 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5834 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5835 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5836 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5837 (AARCH64_ISA_SM4): New ISA flag.
5838 (TARGET_SM4): New feature flag for sm4.
5839 * config/aarch64/aarch64-simd-builtins.def
5840 (aarch64_sm3ss1qv4si): Ditto.
5841 (aarch64_sm3tt1aq4si): Ditto.
5842 (aarch64_sm3tt1bq4si): Ditto.
5843 (aarch64_sm3tt2aq4si): Ditto.
5844 (aarch64_sm3tt2bq4si): Ditto.
5845 (aarch64_sm3partw1qv4si): Ditto.
5846 (aarch64_sm3partw2qv4si): Ditto.
5847 (aarch64_sm4eqv4si): Ditto.
5848 (aarch64_sm4ekeyqv4si): Ditto.
5849 * config/aarch64/aarch64-simd.md:
5850 (aarch64_sm3ss1qv4si): Ditto.
5851 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5852 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5853 (aarch64_sm4eqv4si): Ditto.
5854 (aarch64_sm4ekeyqv4si): Ditto.
5855 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5856 (sm3part_op): Ditto.
5857 (CRYPTO_SM3TT): Ditto.
5858 (CRYPTO_SM3PART): Ditto.
5859 (UNSPEC_SM3SS1): New unspec.
5860 (UNSPEC_SM3TT1A): Ditto.
5861 (UNSPEC_SM3TT1B): Ditto.
5862 (UNSPEC_SM3TT2A): Ditto.
5863 (UNSPEC_SM3TT2B): Ditto.
5864 (UNSPEC_SM3PARTW1): Ditto.
5865 (UNSPEC_SM3PARTW2): Ditto.
5866 (UNSPEC_SM4E): Ditto.
5867 (UNSPEC_SM4EKEY): Ditto.
5868 * config/aarch64/constraints.md (Ui2): New constraint.
5869 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5870 * config/arm/types.md (crypto_sm3): New type attribute.
5871 (crypto_sm4): Ditto.
5872 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5873 (vsm3tt1aq_u32): Ditto.
5874 (vsm3tt1bq_u32): Ditto.
5875 (vsm3tt2aq_u32): Ditto.
5876 (vsm3tt2bq_u32): Ditto.
5877 (vsm3partw1q_u32): Ditto.
5878 (vsm3partw2q_u32): Ditto.
5879 (vsm4eq_u32): Ditto.
5880 (vsm4ekeyq_u32): Ditto.
5881 (doc/invoke.texi): Document new sm4 option.
5883 2018-01-10 Michael Collison <michael.collison@arm.com>
5885 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5886 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5887 (AARCH64_FL_FOR_ARCH8_4): New.
5888 (AARCH64_FL_V8_4): New flag.
5889 (doc/invoke.texi): Document new armv8.4-a option.
5891 2018-01-10 Michael Collison <michael.collison@arm.com>
5893 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5894 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5895 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5896 * config/aarch64/aarch64-option-extension.def: Add
5897 AARCH64_OPT_EXTENSION of 'sha2'.
5898 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5899 (crypto): Disable sha2 and aes if crypto disabled.
5900 (crypto): Enable aes and sha2 if enabled.
5901 (simd): Disable sha2 and aes if simd disabled.
5902 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5904 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5905 (TARGET_SHA2): New feature flag for sha2.
5906 (TARGET_AES): New feature flag for aes.
5907 * config/aarch64/aarch64-simd.md:
5908 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5909 conditional on TARGET_AES.
5910 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5911 (aarch64_crypto_sha1hsi): Make pattern conditional
5913 (aarch64_crypto_sha1hv4si): Ditto.
5914 (aarch64_be_crypto_sha1hv4si): Ditto.
5915 (aarch64_crypto_sha1su1v4si): Ditto.
5916 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5917 (aarch64_crypto_sha1su0v4si): Ditto.
5918 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5919 (aarch64_crypto_sha256su0v4si): Ditto.
5920 (aarch64_crypto_sha256su1v4si): Ditto.
5921 (doc/invoke.texi): Document new aes and sha2 options.
5923 2018-01-10 Martin Sebor <msebor@redhat.com>
5925 PR tree-optimization/83781
5926 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5929 2018-01-11 Martin Sebor <msebor@gmail.com>
5930 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5932 PR tree-optimization/83501
5933 PR tree-optimization/81703
5935 * tree-ssa-strlen.c (get_string_cst): Rename...
5936 (get_string_len): ...to this. Handle global constants.
5937 (handle_char_store): Adjust.
5939 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5940 Jim Wilson <jimw@sifive.com>
5942 * config/riscv/riscv-protos.h (riscv_output_return): New.
5943 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5944 (riscv_attribute_table, riscv_output_return),
5945 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5946 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5947 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5949 (riscv_expand_prologue): Add early return for naked function.
5950 (riscv_expand_epilogue): Likewise.
5951 (riscv_function_ok_for_sibcall): Return false for naked function.
5952 (riscv_set_current_function): New.
5953 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5954 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5955 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5956 * doc/extend.texi (RISC-V Function Attributes): New.
5958 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5960 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5961 check for 128-bit long double before checking TCmode.
5962 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5963 128-bit long doubles before checking TFmode or TCmode.
5964 (FLOAT128_IBM_P): Likewise.
5966 2018-01-10 Martin Sebor <msebor@redhat.com>
5968 PR tree-optimization/83671
5969 * builtins.c (c_strlen): Unconditionally return zero for the empty
5971 Use -Warray-bounds for warnings.
5972 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5973 for non-constant array indices with COMPONENT_REF, arrays of
5974 arrays, and pointers to arrays.
5975 (gimple_fold_builtin_strlen): Determine and set length range for
5976 non-constant character arrays.
5978 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5981 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5984 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5986 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5988 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5991 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5992 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5993 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5994 indexed_or_indirect_operand predicate.
5995 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5996 (*vsx_le_perm_load_v8hi): Likewise.
5997 (*vsx_le_perm_load_v16qi): Likewise.
5998 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5999 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6000 (*vsx_le_perm_store_v8hi): Likewise.
6001 (*vsx_le_perm_store_v16qi): Likewise.
6002 (eight unnamed splitters): Likewise.
6004 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6006 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6007 * config/rs6000/emmintrin.h: Likewise.
6008 * config/rs6000/mmintrin.h: Likewise.
6009 * config/rs6000/xmmintrin.h: Likewise.
6011 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6014 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6016 * tree.c (tree_nop_conversion): Return true for location wrapper
6018 (maybe_wrap_with_location): New function.
6019 (selftest::check_strip_nops): New function.
6020 (selftest::test_location_wrappers): New function.
6021 (selftest::tree_c_tests): Call it.
6022 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6023 (maybe_wrap_with_location): New decl.
6024 (EXPR_LOCATION_WRAPPER_P): New macro.
6025 (location_wrapper_p): New inline function.
6026 (tree_strip_any_location_wrapper): New inline function.
6028 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6031 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6032 stack_realign_offset for the largest alignment of stack slot
6034 (ix86_find_max_used_stack_alignment): New function.
6035 (ix86_finalize_stack_frame_flags): Use it. Set
6036 max_used_stack_alignment if we don't realign stack.
6037 * config/i386/i386.h (machine_function): Add
6038 max_used_stack_alignment.
6040 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6042 * config/arm/arm.opt (-mbranch-cost): New option.
6043 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6046 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6049 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6050 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6052 2018-01-10 Richard Biener <rguenther@suse.de>
6055 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6056 early out so it also covers the case where we have a non-NULL
6059 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6061 PR tree-optimization/83753
6062 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6063 for non-strided grouped accesses if the number of elements is 1.
6065 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6068 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6069 * i386.h (TARGET_USE_GATHER): Define.
6070 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6072 2018-01-10 Martin Liska <mliska@suse.cz>
6075 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6076 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6078 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6079 CLEANUP_NO_PARTITIONING is not set.
6081 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6083 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6084 for vectors, as a partial revert of r254296.
6085 * rtl.h (const_vec_p): Delete.
6086 (const_vec_duplicate_p): Don't test for vector CONSTs.
6087 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6088 * expmed.c (make_tree): Likewise.
6091 * common.md (E, F): Use CONSTANT_P instead of checking for
6093 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6094 checking for CONST_VECTOR.
6096 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6099 * predict.c (force_edge_cold): Handle in more sane way edges
6102 2018-01-09 Carl Love <cel@us.ibm.com>
6104 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6106 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6107 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6108 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6109 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6110 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6111 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6112 * config/rs6000/rs6000-protos.h: Add extern defition for
6113 rs6000_generate_float2_double_code.
6114 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6116 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6117 (float2_v2df): Add define_expand.
6119 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6122 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6123 op_mode in the force_to_mode call.
6125 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6127 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6128 instead of checking each element individually.
6129 (aarch64_evpc_uzp): Likewise.
6130 (aarch64_evpc_zip): Likewise.
6131 (aarch64_evpc_ext): Likewise.
6132 (aarch64_evpc_rev): Likewise.
6133 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6134 instead of checking each element individually. Return true without
6136 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6137 whether all selected elements come from the same input, instead of
6138 checking each element individually. Remove calls to gen_rtx_REG,
6139 start_sequence and end_sequence and instead assert that no rtl is
6142 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6144 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6145 order of HIGH and CONST checks.
6147 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6149 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6150 if the destination isn't an SSA_NAME.
6152 2018-01-09 Richard Biener <rguenther@suse.de>
6154 PR tree-optimization/83668
6155 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6157 (canonicalize_loop_form): ... here, renamed from ...
6158 (canonicalize_loop_closed_ssa_form): ... this and amended to
6159 swap successor edges for loop exit blocks to make us use
6160 the RPO order we need for initial schedule generation.
6162 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6164 PR tree-optimization/64811
6165 * match.pd: When optimizing comparisons with Inf, avoid
6166 introducing or losing exceptions from comparisons with NaN.
6168 2018-01-09 Martin Liska <mliska@suse.cz>
6171 * asan.c (shadow_mem_size): Add gcc_assert.
6173 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6175 Don't save registers in main().
6178 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6179 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6180 * config/avr/avr.c (avr_set_current_function): Don't error if
6181 naked, OS_task or OS_main are specified at the same time.
6182 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6184 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6186 * common/config/avr/avr-common.c (avr_option_optimization_table):
6187 Switch on -mmain-is-OS_task for optimizing compilations.
6189 2018-01-09 Richard Biener <rguenther@suse.de>
6191 PR tree-optimization/83572
6192 * graphite.c: Include cfganal.h.
6193 (graphite_transform_loops): Connect infinite loops to exit
6194 and remove fake edges at the end.
6196 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6198 * ipa-inline.c (edge_badness): Revert accidental checkin.
6200 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6203 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6204 symbols; not inline clones.
6206 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6209 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6210 hard registers. Formatting fixes.
6212 PR preprocessor/83722
6213 * gcc.c (try_generate_repro): Pass
6214 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6215 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6218 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6219 Kito Cheng <kito.cheng@gmail.com>
6221 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6222 (riscv_leaf_function_p): Delete.
6223 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6225 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6227 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6229 (do_ifelse): New function.
6230 (do_isel): New function.
6231 (do_sub3): New function.
6232 (do_add3): New function.
6233 (do_load_mask_compare): New function.
6234 (do_overlap_load_compare): New function.
6235 (expand_compare_loop): New function.
6236 (expand_block_compare): Call expand_compare_loop() when appropriate.
6237 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6239 (-mblock-compare-inline-loop-limit): New option.
6241 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6244 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6245 Reverse order of second and third operands in first alternative.
6246 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6247 of first and second elements in UNSPEC_VPERMR vector.
6248 (altivec_expand_vec_perm_le): Likewise.
6250 2018-01-08 Jeff Law <law@redhat.com>
6252 PR rtl-optimizatin/81308
6253 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6254 (process_switch): If group_case_labels makes a change, then set
6256 (pass_convert_switch::execute): If a switch is converted, then
6257 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6259 PR rtl-optimization/81308
6260 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6263 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6265 PR target/83663 - Revert r255946
6266 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6267 generation for cases where splatting a value is not useful.
6268 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6269 across a vec_duplicate and a paradoxical subreg forming a vector
6270 mode to a vec_concat.
6272 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6274 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6275 -march=armv8.3-a variants.
6276 * config/arm/t-multilib: Likewise.
6277 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6279 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6281 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6283 (cceq_ior_compare_complement): Give it a name so I can use it, and
6284 change boolean_or_operator predicate to boolean_operator so it can
6285 be used to generate a crand.
6286 (eqne): New code iterator.
6287 (bd/bd_neg): New code_attrs.
6288 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6289 a single define_insn.
6290 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6291 decrement (bdnzt/bdnzf/bdzt/bdzf).
6292 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6293 with the new names of the branch decrement patterns, and added the
6294 names of the branch decrement conditional patterns.
6296 2018-01-08 Richard Biener <rguenther@suse.de>
6298 PR tree-optimization/83563
6299 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6302 2018-01-08 Richard Biener <rguenther@suse.de>
6305 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6307 2018-01-08 Richard Biener <rguenther@suse.de>
6309 PR tree-optimization/83685
6310 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6311 references to abnormals.
6313 2018-01-08 Richard Biener <rguenther@suse.de>
6316 * dwarf2out.c (output_indirect_strings): Handle empty
6317 skeleton_debug_str_hash.
6318 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6320 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6322 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6323 (emit_store_direct): Likewise.
6324 (arc_trampoline_adjust_address): Likewise.
6325 (arc_asm_trampoline_template): New function.
6326 (arc_initialize_trampoline): Use asm_trampoline_template.
6327 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6328 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6329 * config/arc/arc.md (flush_icache): Delete pattern.
6331 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6333 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6334 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6337 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6340 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6341 by not USED_FOR_TARGET.
6342 (make_pass_resolve_sw_modes): Likewise.
6344 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6346 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6349 2018-01-08 Richard Biener <rguenther@suse.de>
6352 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6354 2018-01-08 Richard Biener <rguenther@suse.de>
6357 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6359 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6362 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6363 basic blocks with a small number of successors.
6364 (convert_control_dep_chain_into_preds): Improve handling of
6366 (dump_predicates): Split apart into...
6367 (dump_pred_chain): ...here...
6368 (dump_pred_info): ...and here.
6369 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6370 (can_chain_union_be_invalidated_p): Improve check for invalidation
6372 (uninit_uses_cannot_happen): Avoid unnecessary if
6373 convert_control_dep_chain_into_preds yielded nothing.
6375 2018-01-06 Martin Sebor <msebor@redhat.com>
6377 PR tree-optimization/83640
6378 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6379 subtracting negative offset from size.
6380 (builtin_access::overlap): Adjust offset bounds of the access to fall
6381 within the size of the object if possible.
6383 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6385 PR rtl-optimization/83699
6386 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6387 extract_bit_field_as_subreg to cases in which the extracted
6388 value is also a vector.
6390 * lra-constraints.c (process_alt_operands): Test for the equivalence
6391 substitutions when detecting a possible reload cycle.
6393 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6396 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6397 by default if flag_selective_schedling{,2}. Formatting fixes.
6399 PR rtl-optimization/83682
6400 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6401 if it has non-VECTOR_MODE element mode.
6402 (vec_duplicate_p): Likewise.
6405 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6406 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6408 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6411 * config/i386/i386-builtin.def
6412 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6413 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6414 Require also OPTION_MASK_ISA_AVX512F in addition to
6415 OPTION_MASK_ISA_GFNI.
6416 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6417 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6418 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6419 to OPTION_MASK_ISA_GFNI.
6420 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6421 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6422 OPTION_MASK_ISA_AVX512BW.
6423 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6424 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6425 addition to OPTION_MASK_ISA_GFNI.
6426 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6427 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6428 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6429 to OPTION_MASK_ISA_GFNI.
6430 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6431 a requirement for all ISAs rather than any of them with a few
6433 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6435 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6436 bitmasks to be enabled with 3 exceptions, instead of requiring any
6437 enabled ISA with lots of exceptions.
6438 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6439 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6440 Change avx512bw in isa attribute to avx512f.
6441 * config/i386/sgxintrin.h: Add license boilerplate.
6442 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6443 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6444 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6445 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6447 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6448 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6449 temporarily sse2 rather than sse if not enabled already.
6452 * config/i386/sse.md (VI248_VLBW): Rename to ...
6453 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6454 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6455 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6456 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6457 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6458 mode iterator instead of VI248_VLBW.
6460 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6462 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6463 (record_modified): Skip clobbers; add debug output.
6464 (param_change_prob): Use sreal frequencies.
6466 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6468 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6469 punt for user-aligned variables.
6471 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6473 * tree-chrec.c (chrec_contains_symbols): Return true for
6476 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6479 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6480 of (x|y) == x for BICS pattern.
6482 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6484 PR tree-optimization/83605
6485 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6486 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6489 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6491 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6492 * config/epiphany/rtems.h: New file.
6494 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6495 Uros Bizjak <ubizjak@gmail.com>
6498 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6499 QIreg_operand instead of register_operand predicate.
6500 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6501 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6502 comments instead of -fmitigate[-_]rop.
6504 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6507 * cgraphunit.c (symbol_table::compile): Switch to text_section
6508 before calling assembly_start debug hook.
6509 * run-rtl-passes.c (run_rtl_passes): Likewise.
6512 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6514 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6515 range_int_cst_p rather than !symbolic_range_p before calling
6516 extract_range_from_multiplicative_op_1.
6518 2018-01-04 Jeff Law <law@redhat.com>
6520 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6521 redundant test in assertion.
6523 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6525 * doc/rtl.texi: Document machine_mode wrapper classes.
6527 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6529 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6532 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6534 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6535 the VEC_PERM_EXPR fold to fail.
6537 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6540 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6541 to switched_sections.
6543 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6546 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6549 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6552 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6553 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6555 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6558 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6559 is BLKmode and bitpos not zero or mode change is needed.
6561 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6564 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6567 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6570 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6571 instead of MULT rtx. Update all corresponding splitters.
6573 (*ssub<modesuffix>): Ditto.
6575 (*cmp_sadd_di): Update split patterns.
6576 (*cmp_sadd_si): Ditto.
6577 (*cmp_sadd_sidi): Ditto.
6578 (*cmp_ssub_di): Ditto.
6579 (*cmp_ssub_si): Ditto.
6580 (*cmp_ssub_sidi): Ditto.
6581 * config/alpha/predicates.md (const23_operand): New predicate.
6582 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6583 Look for ASHIFT, not MULT inner operand.
6584 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6586 2018-01-04 Martin Liska <mliska@suse.cz>
6588 PR gcov-profile/83669
6589 * gcov.c (output_intermediate_file): Add version to intermediate
6591 * doc/gcov.texi: Document new field 'version' in intermediate
6592 file format. Fix location of '-k' option of gcov command.
6594 2018-01-04 Martin Liska <mliska@suse.cz>
6597 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6599 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6601 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6603 2018-01-03 Martin Sebor <msebor@redhat.com>
6605 PR tree-optimization/83655
6606 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6607 checking calls with invalid arguments.
6609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6611 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6612 (vectorizable_mask_load_store): Delete.
6613 (vectorizable_call): Return false for masked loads and stores.
6614 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6615 instead of gimple_assign_rhs1.
6616 (vectorizable_load): Handle IFN_MASK_LOAD.
6617 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6619 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6621 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6623 (vectorizable_mask_load_store): ...here.
6624 (vectorizable_load): ...and here.
6626 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6628 * tree-vect-stmts.c (vect_build_all_ones_mask)
6629 (vect_build_zero_merge_argument): New functions, split out from...
6630 (vectorizable_load): ...here.
6632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6634 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6636 (vectorizable_mask_load_store): ...here.
6637 (vectorizable_store): ...and here.
6639 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6641 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6643 (vectorizable_mask_load_store): ...here.
6645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6647 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6648 (vect_model_store_cost): Take a vec_load_store_type instead of a
6650 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6651 (vect_model_store_cost): Take a vec_load_store_type instead of a
6653 (vectorizable_mask_load_store): Update accordingly.
6654 (vectorizable_store): Likewise.
6655 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6659 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6660 IFN_MASK_LOAD calls here rather than...
6661 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6664 Alan Hayward <alan.hayward@arm.com>
6665 David Sherwood <david.sherwood@arm.com>
6667 * expmed.c (extract_bit_field_1): For vector extracts,
6668 fall back to extract_bit_field_as_subreg if vec_extract
6671 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6672 Alan Hayward <alan.hayward@arm.com>
6673 David Sherwood <david.sherwood@arm.com>
6675 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6676 they are variable or constant sized.
6677 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6678 slots for constant-sized data.
6680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6681 Alan Hayward <alan.hayward@arm.com>
6682 David Sherwood <david.sherwood@arm.com>
6684 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6685 handling COND_EXPRs with boolean comparisons, try to find a better
6686 basis for the mask type than the boolean itself.
6688 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6690 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6691 is calculated and how it can be overridden.
6692 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6693 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6695 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6698 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6699 Alan Hayward <alan.hayward@arm.com>
6700 David Sherwood <david.sherwood@arm.com>
6702 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6703 Remove the mode argument.
6704 (aarch64_simd_valid_immediate): Remove the mode and inverse
6706 * config/aarch64/iterators.md (bitsize): New iterator.
6707 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6708 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6709 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6710 aarch64_simd_valid_immediate.
6711 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6712 (aarch64_reg_or_bic_imm): Likewise.
6713 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6714 with an insn_type enum and msl with a modifier_type enum.
6715 Replace element_width with a scalar_mode. Change the shift
6716 to unsigned int. Add constructors for scalar_float_mode and
6717 scalar_int_mode elements.
6718 (aarch64_vect_float_const_representable_p): Delete.
6719 (aarch64_can_const_movi_rtx_p)
6720 (aarch64_simd_scalar_immediate_valid_for_move)
6721 (aarch64_simd_make_constant): Update call to
6722 aarch64_simd_valid_immediate.
6723 (aarch64_advsimd_valid_immediate_hs): New function.
6724 (aarch64_advsimd_valid_immediate): Likewise.
6725 (aarch64_simd_valid_immediate): Remove mode and inverse
6726 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6727 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6728 and aarch64_float_const_representable_p on the result.
6729 (aarch64_output_simd_mov_immediate): Remove mode argument.
6730 Update call to aarch64_simd_valid_immediate and use of
6731 simd_immediate_info.
6732 (aarch64_output_scalar_simd_mov_immediate): Update call
6735 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6736 Alan Hayward <alan.hayward@arm.com>
6737 David Sherwood <david.sherwood@arm.com>
6739 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6740 (mode_nunits): Likewise CONST_MODE_NUNITS.
6741 * machmode.def (ADJUST_NUNITS): Document.
6742 * genmodes.c (mode_data::need_nunits_adj): New field.
6743 (blank_mode): Update accordingly.
6744 (adj_nunits): New variable.
6745 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6747 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6748 listed in adj_nunits.
6749 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6750 listed in adj_nunits. Don't emit case statements for such modes.
6751 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6752 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6753 nothing if adj_nunits is nonnull.
6754 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6755 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6756 (emit_mode_fbit): Update use of print_maybe_const_decl.
6757 (emit_move_size): Likewise. Treat the array as non-const
6759 (emit_mode_adjustments): Handle adj_nunits.
6761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6763 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6764 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6765 (VECTOR_MODES): Use it.
6766 (make_vector_modes): Take the prefix as an argument.
6768 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6769 Alan Hayward <alan.hayward@arm.com>
6770 David Sherwood <david.sherwood@arm.com>
6772 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6773 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6774 for MODE_VECTOR_BOOL.
6775 * machmode.def (VECTOR_BOOL_MODE): Document.
6776 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6777 (make_vector_bool_mode): New function.
6778 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6780 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6781 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6783 * stor-layout.c (int_mode_for_mode): Likewise.
6784 * tree.c (build_vector_type_for_mode): Likewise.
6785 * varasm.c (output_constant_pool_2): Likewise.
6786 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6787 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6788 for MODE_VECTOR_BOOL.
6789 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6790 of mode class checks.
6791 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6792 instead of a list of mode class checks.
6793 (expand_vector_scalar_condition): Likewise.
6794 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6797 Alan Hayward <alan.hayward@arm.com>
6798 David Sherwood <david.sherwood@arm.com>
6800 * machmode.h (mode_size): Change from unsigned short to
6802 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6803 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6804 or if measurement_type is not polynomial.
6805 (fixed_size_mode::includes_p): Check for constant-sized modes.
6806 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6807 return a poly_uint16 rather than an unsigned short.
6808 (emit_mode_size): Change the type of mode_size from unsigned short
6809 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6810 (emit_mode_adjustments): Cope with polynomial vector sizes.
6811 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6813 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6815 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6816 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6817 * caller-save.c (setup_save_areas): Likewise.
6818 (replace_reg_with_saved_mem): Likewise.
6819 * calls.c (emit_library_call_value_1): Likewise.
6820 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6821 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6822 (gen_lowpart_for_combine): Likewise.
6823 * convert.c (convert_to_integer_1): Likewise.
6824 * cse.c (equiv_constant, cse_insn): Likewise.
6825 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6826 (cselib_subst_to_values): Likewise.
6827 * dce.c (word_dce_process_block): Likewise.
6828 * df-problems.c (df_word_lr_mark_ref): Likewise.
6829 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6830 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6831 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6832 (rtl_for_decl_location): Likewise.
6833 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6834 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6835 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6836 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6837 (expand_expr_real_1): Likewise.
6838 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6839 (pad_below): Likewise.
6840 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6841 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6842 * ira.c (get_subreg_tracking_sizes): Likewise.
6843 * ira-build.c (ira_create_allocno_objects): Likewise.
6844 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6845 (ira_sort_regnos_for_alter_reg): Likewise.
6846 * ira-costs.c (record_operand_costs): Likewise.
6847 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6848 (resolve_simple_move): Likewise.
6849 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6850 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6851 (lra_constraints): Likewise.
6852 (CONST_POOL_OK_P): Reject variable-sized modes.
6853 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6854 (add_pseudo_to_slot, lra_spill): Likewise.
6855 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6856 * optabs-query.c (get_best_extraction_insn): Likewise.
6857 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6858 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6859 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6860 * recog.c (offsettable_address_addr_space_p): Likewise.
6861 * regcprop.c (maybe_mode_change): Likewise.
6862 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6863 * regrename.c (build_def_use): Likewise.
6864 * regstat.c (dump_reg_info): Likewise.
6865 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6866 (find_reloads, find_reloads_subreg_address): Likewise.
6867 * reload1.c (eliminate_regs_1): Likewise.
6868 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6869 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6870 (simplify_binary_operation_1, simplify_subreg): Likewise.
6871 * targhooks.c (default_function_arg_padding): Likewise.
6872 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6873 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6874 (verify_gimple_assign_ternary): Likewise.
6875 * tree-inline.c (estimate_move_cost): Likewise.
6876 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6877 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6878 (get_address_cost_ainc): Likewise.
6879 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6880 (vect_supportable_dr_alignment): Likewise.
6881 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6882 (vectorizable_reduction): Likewise.
6883 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6884 (vectorizable_operation, vectorizable_load): Likewise.
6885 * tree.c (build_same_sized_truth_vector_type): Likewise.
6886 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6887 * var-tracking.c (emit_note_insn_var_location): Likewise.
6888 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6889 (ADDR_VEC_ALIGN): Likewise.
6891 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6892 Alan Hayward <alan.hayward@arm.com>
6893 David Sherwood <david.sherwood@arm.com>
6895 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6897 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6898 or if measurement_type is polynomial.
6899 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6900 * combine.c (make_extraction): Likewise.
6901 * dse.c (find_shift_sequence): Likewise.
6902 * dwarf2out.c (mem_loc_descriptor): Likewise.
6903 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6904 (extract_bit_field, extract_low_bits): Likewise.
6905 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6906 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6907 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6908 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6909 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6910 * reload.c (find_reloads): Likewise.
6911 * reload1.c (alter_reg): Likewise.
6912 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6913 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6914 * tree-if-conv.c (predicate_mem_writes): Likewise.
6915 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6916 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6917 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6918 * valtrack.c (dead_debug_insert_temp): Likewise.
6919 * varasm.c (mergeable_constant_section): Likewise.
6920 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6922 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6923 Alan Hayward <alan.hayward@arm.com>
6924 David Sherwood <david.sherwood@arm.com>
6926 * expr.c (expand_assignment): Cope with polynomial mode sizes
6927 when assigning to a CONCAT.
6929 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6930 Alan Hayward <alan.hayward@arm.com>
6931 David Sherwood <david.sherwood@arm.com>
6933 * machmode.h (mode_precision): Change from unsigned short to
6935 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6937 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6938 or if measurement_type is not polynomial.
6939 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6940 in which the mode is already known to be a scalar_int_mode.
6941 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6942 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6944 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6945 for GET_MODE_PRECISION.
6946 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6947 for GET_MODE_PRECISION.
6948 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6950 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6951 (expand_field_assignment, make_extraction): Likewise.
6952 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6953 (get_last_value): Likewise.
6954 * convert.c (convert_to_integer_1): Likewise.
6955 * cse.c (cse_insn): Likewise.
6956 * expr.c (expand_expr_real_1): Likewise.
6957 * lra-constraints.c (simplify_operand_subreg): Likewise.
6958 * optabs-query.c (can_atomic_load_p): Likewise.
6959 * optabs.c (expand_atomic_load): Likewise.
6960 (expand_atomic_store): Likewise.
6961 * ree.c (combine_reaching_defs): Likewise.
6962 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6963 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6964 * tree.h (type_has_mode_precision_p): Likewise.
6965 * ubsan.c (instrument_si_overflow): Likewise.
6967 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6968 Alan Hayward <alan.hayward@arm.com>
6969 David Sherwood <david.sherwood@arm.com>
6971 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6972 polynomial numbers of units.
6973 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6974 (valid_vector_subparts_p): New function.
6975 (build_vector_type): Remove temporary shim and take the number
6976 of units as a poly_uint64 rather than an int.
6977 (build_opaque_vector_type): Take the number of units as a
6978 poly_uint64 rather than an int.
6979 * tree.c (build_vector_from_ctor): Handle polynomial
6980 TYPE_VECTOR_SUBPARTS.
6981 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6982 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6983 (build_vector_from_val): If the number of units is variable,
6984 use build_vec_duplicate_cst for constant operands and
6985 VEC_DUPLICATE_EXPR otherwise.
6986 (make_vector_type): Remove temporary is_constant ().
6987 (build_vector_type, build_opaque_vector_type): Take the number of
6988 units as a poly_uint64 rather than an int.
6989 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6991 * cfgexpand.c (expand_debug_expr): Likewise.
6992 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6993 (store_constructor, expand_expr_real_1): Likewise.
6994 (const_scalar_mask_from_tree): Likewise.
6995 * fold-const-call.c (fold_const_reduction): Likewise.
6996 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6997 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6998 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6999 (fold_relational_const): Likewise.
7000 (native_interpret_vector): Likewise. Change the size from an
7001 int to an unsigned int.
7002 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7003 TYPE_VECTOR_SUBPARTS.
7004 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7005 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7006 duplicating a non-constant operand into a variable-length vector.
7007 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7008 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7009 * ipa-icf.c (sem_variable::equals): Likewise.
7010 * match.pd: Likewise.
7011 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7012 * print-tree.c (print_node): Likewise.
7013 * stor-layout.c (layout_type): Likewise.
7014 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7015 * tree-cfg.c (verify_gimple_comparison): Likewise.
7016 (verify_gimple_assign_binary): Likewise.
7017 (verify_gimple_assign_ternary): Likewise.
7018 (verify_gimple_assign_single): Likewise.
7019 * tree-pretty-print.c (dump_generic_node): Likewise.
7020 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7021 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7022 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7023 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7024 (vect_shift_permute_load_chain): Likewise.
7025 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7026 (expand_vector_condition, optimize_vector_constructor): Likewise.
7027 (lower_vec_perm, get_compute_type): Likewise.
7028 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7029 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7030 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7031 (vect_recog_mask_conversion_pattern): Likewise.
7032 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7033 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7034 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7035 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7036 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7037 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7038 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7039 (supportable_widening_operation): Likewise.
7040 (supportable_narrowing_operation): Likewise.
7041 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7043 * varasm.c (output_constant): Likewise.
7045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7046 Alan Hayward <alan.hayward@arm.com>
7047 David Sherwood <david.sherwood@arm.com>
7049 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7050 so that both the length == 3 and length != 3 cases set up their
7051 own permute vectors. Add comments explaining why we know the
7052 number of elements is constant.
7053 (vect_permute_load_chain): Likewise.
7055 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7056 Alan Hayward <alan.hayward@arm.com>
7057 David Sherwood <david.sherwood@arm.com>
7059 * machmode.h (mode_nunits): Change from unsigned char to
7061 (ONLY_FIXED_SIZE_MODES): New macro.
7062 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7063 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7064 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7066 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7067 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7068 or if measurement_type is not polynomial.
7069 * genmodes.c (ZERO_COEFFS): New macro.
7070 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7072 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7073 Use ZERO_COEFFS when emitting initializers.
7074 * data-streamer.h (bp_pack_poly_value): New function.
7075 (bp_unpack_poly_value): Likewise.
7076 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7077 for GET_MODE_NUNITS.
7078 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7079 for GET_MODE_NUNITS.
7080 * tree.c (make_vector_type): Remove temporary shim and make
7081 the real function take the number of units as a poly_uint64
7083 (build_vector_type_for_mode): Handle polynomial nunits.
7084 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7085 * emit-rtl.c (const_vec_series_p_1): Likewise.
7086 (gen_rtx_CONST_VECTOR): Likewise.
7087 * fold-const.c (test_vec_duplicate_folding): Likewise.
7088 * genrecog.c (validate_pattern): Likewise.
7089 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7090 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7091 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7092 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7093 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7094 * rtlanal.c (subreg_get_info): Likewise.
7095 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7096 (vect_grouped_load_supported): Likewise.
7097 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7098 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7099 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7100 (simplify_const_unary_operation, simplify_binary_operation_1)
7101 (simplify_const_binary_operation, simplify_ternary_operation)
7102 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7103 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7104 instead of CONST_VECTOR_NUNITS.
7105 * varasm.c (output_constant_pool_2): Likewise.
7106 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7107 explicit-encoded elements in the XVEC for variable-length vectors.
7109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7111 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7113 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7114 Alan Hayward <alan.hayward@arm.com>
7115 David Sherwood <david.sherwood@arm.com>
7117 * coretypes.h (fixed_size_mode): Declare.
7118 (fixed_size_mode_pod): New typedef.
7119 * builtins.h (target_builtins::x_apply_args_mode)
7120 (target_builtins::x_apply_result_mode): Change type to
7121 fixed_size_mode_pod.
7122 * builtins.c (apply_args_size, apply_result_size, result_vector)
7123 (expand_builtin_apply_args_1, expand_builtin_apply)
7124 (expand_builtin_return): Update accordingly.
7126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7128 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7129 * cselib.c (cselib_hash_rtx): Likewise.
7130 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7131 CONST_VECTOR encoding.
7133 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7134 Jeff Law <law@redhat.com>
7137 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7138 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7139 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7140 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7143 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7144 explicitly probe *sp in a noreturn function if there were any callee
7145 register saves or frame pointer is needed.
7147 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7150 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7151 BLKmode for ternary, binary or unary expressions.
7154 * var-tracking.c (delete_vta_debug_insn): New inline function.
7155 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7156 insns from get_insns () to NULL instead of each bb separately.
7157 Use delete_vta_debug_insn. No longer static.
7158 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7159 delete_vta_debug_insns callers.
7160 * rtl.h (delete_vta_debug_insns): Declare.
7161 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7162 instead of variable_tracking_main.
7164 2018-01-03 Martin Sebor <msebor@redhat.com>
7166 PR tree-optimization/83603
7167 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7168 arguments past the endof the argument list in functions declared
7169 without a prototype.
7170 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7171 Avoid checking when arguments are null.
7173 2018-01-03 Martin Sebor <msebor@redhat.com>
7176 * doc/extend.texi (attribute const): Fix a typo.
7177 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7178 issuing -Wsuggest-attribute for void functions.
7180 2018-01-03 Martin Sebor <msebor@redhat.com>
7182 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7183 offset_int::from instead of wide_int::to_shwi.
7184 (maybe_diag_overlap): Remove assertion.
7185 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7186 * gimple-ssa-sprintf.c (format_directive): Same.
7187 (parse_directive): Same.
7188 (sprintf_dom_walker::compute_format_length): Same.
7189 (try_substitute_return_value): Same.
7191 2018-01-03 Jeff Law <law@redhat.com>
7194 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7195 non-constant residual for zero at runtime and avoid probing in
7196 that case. Reorganize code for trailing problem to mirror handling
7199 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7201 PR tree-optimization/83501
7202 * tree-ssa-strlen.c (get_string_cst): New.
7203 (handle_char_store): Call get_string_cst.
7205 2018-01-03 Martin Liska <mliska@suse.cz>
7207 PR tree-optimization/83593
7208 * tree-ssa-strlen.c: Include tree-cfg.h.
7209 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7210 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7211 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7213 (strlen_dom_walker::before_dom_children): Call
7214 gimple_purge_dead_eh_edges. Dump tranformation with details
7216 (strlen_dom_walker::before_dom_children): Update call by adding
7217 new argument cleanup_eh.
7218 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7220 2018-01-03 Martin Liska <mliska@suse.cz>
7223 * cif-code.def (VARIADIC_THUNK): New enum value.
7224 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7227 2018-01-03 Jan Beulich <jbeulich@suse.com>
7229 * sse.md (mov<mode>_internal): Tighten condition for when to use
7230 vmovdqu<ssescalarsize> for TI and OI modes.
7232 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7234 Update copyright years.
7236 2018-01-03 Martin Liska <mliska@suse.cz>
7239 * ipa-visibility.c (function_and_variable_visibility): Skip
7240 functions with noipa attribure.
7242 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7244 * gcc.c (process_command): Update copyright notice dates.
7245 * gcov-dump.c (print_version): Ditto.
7246 * gcov.c (print_version): Ditto.
7247 * gcov-tool.c (print_version): Ditto.
7248 * gengtype.c (create_file): Ditto.
7249 * doc/cpp.texi: Bump @copying's copyright year.
7250 * doc/cppinternals.texi: Ditto.
7251 * doc/gcc.texi: Ditto.
7252 * doc/gccint.texi: Ditto.
7253 * doc/gcov.texi: Ditto.
7254 * doc/install.texi: Ditto.
7255 * doc/invoke.texi: Ditto.
7257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7259 * vector-builder.h (vector_builder::m_full_nelts): Change from
7260 unsigned int to poly_uint64.
7261 (vector_builder::full_nelts): Update prototype accordingly.
7262 (vector_builder::new_vector): Likewise.
7263 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7264 (vector_builder::operator ==): Likewise.
7265 (vector_builder::finalize): Likewise.
7266 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7267 Take the number of elements as a poly_uint64 rather than an
7269 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7270 from unsigned int to poly_uint64.
7271 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7272 (vec_perm_indices::new_vector): Likewise.
7273 (vec_perm_indices::length): Likewise.
7274 (vec_perm_indices::nelts_per_input): Likewise.
7275 (vec_perm_indices::input_nelts): Likewise.
7276 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7277 number of elements per input as a poly_uint64 rather than an
7278 unsigned int. Use the original encoding for variable-length
7279 vectors, rather than clamping each individual element.
7280 For the second and subsequent elements in each pattern,
7281 clamp the step and base before clamping their sum.
7282 (vec_perm_indices::series_p): Handle polynomial element counts.
7283 (vec_perm_indices::all_in_range_p): Likewise.
7284 (vec_perm_indices_to_tree): Likewise.
7285 (vec_perm_indices_to_rtx): Likewise.
7286 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7287 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7288 (tree_vector_builder::new_binary_operation): Handle polynomial
7289 element counts. Return false if we need to know the number
7290 of elements at compile time.
7291 * fold-const.c (fold_vec_perm): Punt if the number of elements
7292 isn't known at compile time.
7294 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7296 * vec-perm-indices.h (vec_perm_builder): Change element type
7297 from HOST_WIDE_INT to poly_int64.
7298 (vec_perm_indices::element_type): Update accordingly.
7299 (vec_perm_indices::clamp): Handle polynomial element_types.
7300 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7301 (vec_perm_indices::all_in_range_p): Likewise.
7302 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7304 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7305 polynomial vec_perm_indices element types.
7306 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7307 * fold-const.c (fold_vec_perm): Likewise.
7308 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7309 * tree-vect-generic.c (lower_vec_perm): Likewise.
7310 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7311 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7312 element type to HOST_WIDE_INT.
7314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7315 Alan Hayward <alan.hayward@arm.com>
7316 David Sherwood <david.sherwood@arm.com>
7318 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7319 rather than an int. Use plus_constant.
7320 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7321 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7324 Alan Hayward <alan.hayward@arm.com>
7325 David Sherwood <david.sherwood@arm.com>
7327 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7328 a HOST_WIDE_INT to a poly_int64.
7330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7331 Alan Hayward <alan.hayward@arm.com>
7332 David Sherwood <david.sherwood@arm.com>
7334 * calls.c (load_register_parameters): Cope with polynomial
7335 mode sizes. Require a constant size for BLKmode parameters
7336 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7337 forces a parameter to be padded at the lsb end in order to
7338 fill a complete number of words, require the parameter size
7339 to be ordered wrt UNITS_PER_WORD.
7341 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7342 Alan Hayward <alan.hayward@arm.com>
7343 David Sherwood <david.sherwood@arm.com>
7345 * reload1.c (spill_stack_slot_width): Change element type
7346 from unsigned int to poly_uint64_pod.
7347 (alter_reg): Treat mode sizes as polynomial.
7349 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7350 Alan Hayward <alan.hayward@arm.com>
7351 David Sherwood <david.sherwood@arm.com>
7353 * reload.c (complex_word_subreg_p): New function.
7354 (reload_inner_reg_of_subreg, push_reload): Use it.
7356 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7357 Alan Hayward <alan.hayward@arm.com>
7358 David Sherwood <david.sherwood@arm.com>
7360 * lra-constraints.c (process_alt_operands): Reject matched
7361 operands whose sizes aren't ordered.
7362 (match_reload): Refer to this check here.
7364 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7365 Alan Hayward <alan.hayward@arm.com>
7366 David Sherwood <david.sherwood@arm.com>
7368 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7369 that the mode size is in the set {1, 2, 4, 8, 16}.
7371 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7372 Alan Hayward <alan.hayward@arm.com>
7373 David Sherwood <david.sherwood@arm.com>
7375 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7376 Use plus_constant instead of gen_rtx_PLUS.
7378 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7379 Alan Hayward <alan.hayward@arm.com>
7380 David Sherwood <david.sherwood@arm.com>
7382 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7383 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7384 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7385 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7386 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7387 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7388 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7389 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7390 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7391 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7393 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7394 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7395 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7396 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7397 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7398 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7399 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7400 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7401 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7402 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7404 * expr.c (emit_move_resolve_push): Treat the input and result
7405 of PUSH_ROUNDING as a poly_int64.
7406 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7407 (emit_push_insn): Likewise.
7408 * lra-eliminations.c (mark_not_eliminable): Likewise.
7409 * recog.c (push_operand): Likewise.
7410 * reload1.c (elimination_effects): Likewise.
7411 * rtlanal.c (nonzero_bits1): Likewise.
7412 * calls.c (store_one_arg): Likewise. Require the padding to be
7413 known at compile time.
7415 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7416 Alan Hayward <alan.hayward@arm.com>
7417 David Sherwood <david.sherwood@arm.com>
7419 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7420 Use plus_constant instead of gen_rtx_PLUS.
7422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7423 Alan Hayward <alan.hayward@arm.com>
7424 David Sherwood <david.sherwood@arm.com>
7426 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7430 Alan Hayward <alan.hayward@arm.com>
7431 David Sherwood <david.sherwood@arm.com>
7433 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7434 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7435 via stack temporaries. Treat the mode size as polynomial too.
7437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7438 Alan Hayward <alan.hayward@arm.com>
7439 David Sherwood <david.sherwood@arm.com>
7441 * expr.c (expand_expr_real_2): When handling conversions involving
7442 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7443 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7444 as a poly_uint64 too.
7446 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7447 Alan Hayward <alan.hayward@arm.com>
7448 David Sherwood <david.sherwood@arm.com>
7450 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7453 Alan Hayward <alan.hayward@arm.com>
7454 David Sherwood <david.sherwood@arm.com>
7456 * combine.c (can_change_dest_mode): Handle polynomial
7457 REGMODE_NATURAL_SIZE.
7458 * expmed.c (store_bit_field_1): Likewise.
7459 * expr.c (store_constructor): Likewise.
7460 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7461 and polynomial REGMODE_NATURAL_SIZE.
7462 (gen_lowpart_common): Likewise.
7463 * reginfo.c (record_subregs_of_mode): Likewise.
7464 * rtlanal.c (read_modify_subreg_p): Likewise.
7466 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7467 Alan Hayward <alan.hayward@arm.com>
7468 David Sherwood <david.sherwood@arm.com>
7470 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7471 numbers of elements.
7473 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7474 Alan Hayward <alan.hayward@arm.com>
7475 David Sherwood <david.sherwood@arm.com>
7477 * match.pd: Cope with polynomial numbers of vector elements.
7479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7480 Alan Hayward <alan.hayward@arm.com>
7481 David Sherwood <david.sherwood@arm.com>
7483 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7484 in a POINTER_PLUS_EXPR.
7486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7487 Alan Hayward <alan.hayward@arm.com>
7488 David Sherwood <david.sherwood@arm.com>
7490 * omp-simd-clone.c (simd_clone_subparts): New function.
7491 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7492 (ipa_simd_modify_function_body): Likewise.
7494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7495 Alan Hayward <alan.hayward@arm.com>
7496 David Sherwood <david.sherwood@arm.com>
7498 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7499 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7500 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7501 (expand_vector_condition, vector_element): Likewise.
7502 (subparts_gt): New function.
7503 (get_compute_type): Use subparts_gt.
7504 (count_type_subparts): Delete.
7505 (expand_vector_operations_1): Use subparts_gt instead of
7506 count_type_subparts.
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 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7513 (vect_compile_time_alias): ...this new function. Do the calculation
7514 on poly_ints rather than trees.
7515 (vect_prune_runtime_alias_test_list): Update call accordingly.
7517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7518 Alan Hayward <alan.hayward@arm.com>
7519 David Sherwood <david.sherwood@arm.com>
7521 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7523 (vect_schedule_slp_instance): Likewise.
7525 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7526 Alan Hayward <alan.hayward@arm.com>
7527 David Sherwood <david.sherwood@arm.com>
7529 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7530 constant and extern definitions for variable-length vectors.
7531 (vect_get_constant_vectors): Note that the number of units
7532 is known to be constant.
7534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7535 Alan Hayward <alan.hayward@arm.com>
7536 David Sherwood <david.sherwood@arm.com>
7538 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7539 of units as polynomial. Choose between WIDE and NARROW based
7542 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7543 Alan Hayward <alan.hayward@arm.com>
7544 David Sherwood <david.sherwood@arm.com>
7546 * tree-vect-stmts.c (simd_clone_subparts): New function.
7547 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7550 Alan Hayward <alan.hayward@arm.com>
7551 David Sherwood <david.sherwood@arm.com>
7553 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7554 vectors as polynomial. Use build_index_vector for
7557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7558 Alan Hayward <alan.hayward@arm.com>
7559 David Sherwood <david.sherwood@arm.com>
7561 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7562 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7563 for variable-length vectors.
7564 (vectorizable_mask_load_store): Treat the number of units as
7565 polynomial, asserting that it is constant if the condition has
7566 already been enforced.
7567 (vectorizable_store, vectorizable_load): Likewise.
7569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7570 Alan Hayward <alan.hayward@arm.com>
7571 David Sherwood <david.sherwood@arm.com>
7573 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7574 of units as polynomial. Punt if we can't tell at compile time
7575 which vector contains the final result.
7577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7578 Alan Hayward <alan.hayward@arm.com>
7579 David Sherwood <david.sherwood@arm.com>
7581 * tree-vect-loop.c (vectorizable_induction): Treat the number
7582 of units as polynomial. Punt on SLP inductions. Use an integer
7583 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7584 cast of such a series for variable-length floating-point
7587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7588 Alan Hayward <alan.hayward@arm.com>
7589 David Sherwood <david.sherwood@arm.com>
7591 * tree.h (build_index_vector): Declare.
7592 * tree.c (build_index_vector): New function.
7593 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7594 of units as polynomial, forcibly converting it to a constant if
7595 vectorizable_reduction has already enforced the condition.
7596 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7597 to create a {1,2,3,...} vector.
7598 (vectorizable_reduction): Treat the number of units as polynomial.
7599 Choose vectype_in based on the largest scalar element size rather
7600 than the smallest number of units. Enforce the restrictions
7603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7604 Alan Hayward <alan.hayward@arm.com>
7605 David Sherwood <david.sherwood@arm.com>
7607 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7608 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7615 * target.def (autovectorize_vector_sizes): Return the vector sizes
7616 by pointer, using vector_sizes rather than a bitmask.
7617 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7618 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7619 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7621 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7622 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7623 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7624 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7625 * omp-general.c (omp_max_vf): Likewise.
7626 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7627 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7628 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7629 * tree-vect-slp.c (vect_slp_bb): Likewise.
7630 * doc/tm.texi: Regenerate.
7631 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7633 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7634 the vector size as a poly_uint64 rather than an unsigned int.
7635 (current_vector_size): Change from an unsigned int to a poly_uint64.
7636 (get_vectype_for_scalar_type): Update accordingly.
7637 * tree.h (build_truth_vector_type): Take the size and number of
7638 units as a poly_uint64 rather than an unsigned int.
7639 (build_vector_type): Add a temporary overload that takes
7640 the number of units as a poly_uint64 rather than an unsigned int.
7641 * tree.c (make_vector_type): Likewise.
7642 (build_truth_vector_type): Take the number of units as a poly_uint64
7643 rather than an unsigned int.
7645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7646 Alan Hayward <alan.hayward@arm.com>
7647 David Sherwood <david.sherwood@arm.com>
7649 * target.def (get_mask_mode): Take the number of units and length
7650 as poly_uint64s rather than unsigned ints.
7651 * targhooks.h (default_get_mask_mode): Update accordingly.
7652 * targhooks.c (default_get_mask_mode): Likewise.
7653 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7654 * doc/tm.texi: Regenerate.
7656 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7657 Alan Hayward <alan.hayward@arm.com>
7658 David Sherwood <david.sherwood@arm.com>
7660 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7661 * omp-general.c (omp_max_vf): Likewise.
7662 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7663 (expand_omp_simd): Handle polynomial safelen.
7664 * omp-low.c (omplow_simd_context): Add a default constructor.
7665 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7666 (lower_rec_simd_input_clauses): Update accordingly.
7667 (lower_rec_input_clauses): 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-vectorizer.h (vect_nunits_for_cost): New function.
7674 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7675 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7676 (vect_analyze_slp_cost): Likewise.
7677 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7678 (vect_model_load_cost): Likewise.
7680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7681 Alan Hayward <alan.hayward@arm.com>
7682 David Sherwood <david.sherwood@arm.com>
7684 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7685 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7686 from an unsigned int * to a poly_uint64_pod *.
7687 (calculate_unrolling_factor): New function.
7688 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7691 Alan Hayward <alan.hayward@arm.com>
7692 David Sherwood <david.sherwood@arm.com>
7694 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7695 from an unsigned int to a poly_uint64.
7696 (_loop_vec_info::slp_unrolling_factor): Likewise.
7697 (_loop_vec_info::vectorization_factor): Change from an int
7699 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7700 (vect_get_num_vectors): New function.
7701 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7702 (vect_get_num_copies): Use vect_get_num_vectors.
7703 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7704 to an unsigned int *.
7705 (vect_analyze_data_refs): Change min_vf from an int * to a
7707 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7708 than an unsigned HOST_WIDE_INT.
7709 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7710 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7711 to an unsigned int *.
7712 (vect_analyze_data_ref_dependences): Likewise.
7713 (vect_compute_data_ref_alignment): Handle polynomial vf.
7714 (vect_enhance_data_refs_alignment): Likewise.
7715 (vect_prune_runtime_alias_test_list): Likewise.
7716 (vect_shift_permute_load_chain): Likewise.
7717 (vect_supportable_dr_alignment): Likewise.
7718 (dependence_distance_ge_vf): Take the vectorization factor as a
7719 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7720 (vect_analyze_data_refs): Change min_vf from an int * to a
7722 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7723 vfm1 as a poly_uint64 rather than an int. Make the same change
7724 for the returned bound_scalar.
7725 (vect_gen_vector_loop_niters): Handle polynomial vf.
7726 (vect_do_peeling): Likewise. Update call to
7727 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7728 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7730 * tree-vect-loop.c (vect_determine_vectorization_factor)
7731 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7732 (vect_get_known_peeling_cost): Likewise.
7733 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7734 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7735 (vect_transform_loop): Likewise. Use the lowest possible VF when
7736 updating the upper bounds of the loop.
7737 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7739 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7740 polynomial unroll factors.
7741 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7742 (vect_make_slp_decision): Likewise.
7743 (vect_supported_load_permutation_p): Likewise, and polynomial
7745 (vect_analyze_slp_cost): Handle polynomial vf.
7746 (vect_slp_analyze_node_operations): Likewise.
7747 (vect_slp_analyze_bb_1): Likewise.
7748 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7749 than an unsigned HOST_WIDE_INT.
7750 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7751 (vectorizable_load): Handle polynomial vf.
7752 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7754 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7756 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7757 Alan Hayward <alan.hayward@arm.com>
7758 David Sherwood <david.sherwood@arm.com>
7760 * match.pd: Handle bit operations involving three constants
7761 and try to fold one pair.
7763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7765 * tree-vect-loop-manip.c: Include gimple-fold.h.
7766 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7767 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7768 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7769 Add a path that uses a step of VF instead of 1, but disable it
7771 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7772 and niters_no_overflow parameters. Update calls to
7773 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7774 Create a new SSA name if the latter choses to use a ste other
7775 than zero, and return it via niters_vector_mult_vf_var.
7776 * tree-vect-loop.c (vect_transform_loop): Update calls to
7777 vect_do_peeling, vect_gen_vector_loop_niters and
7778 slpeel_make_loop_iterate_ntimes.
7779 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7780 (vect_gen_vector_loop_niters): Update declarations after above changes.
7782 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7784 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7785 128-bit round to integer instructions.
7786 (ceil<mode>2): Likewise.
7787 (btrunc<mode>2): Likewise.
7788 (round<mode>2): Likewise.
7790 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7792 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7793 unaligned VSX load/store on P8/P9.
7794 (expand_block_clear): Allow the use of unaligned VSX
7795 load/store on P8/P9.
7797 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7799 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7801 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7802 swap associated with both a load and a store.
7804 2018-01-02 Andrew Waterman <andrew@sifive.com>
7806 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7807 * config/riscv/riscv.md (clear_cache): Use it.
7809 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7811 * web.c: Remove out-of-date comment.
7813 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7815 * expr.c (fixup_args_size_notes): Check that any existing
7816 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7817 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7818 (emit_single_push_insn): ...here.
7820 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7822 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7823 (const_vector_encoded_nelts): New function.
7824 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7825 (const_vector_int_elt, const_vector_elt): Declare.
7826 * emit-rtl.c (const_vector_int_elt_1): New function.
7827 (const_vector_elt): Likewise.
7828 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7829 of CONST_VECTOR_ELT.
7831 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7833 * expr.c: Include rtx-vector-builder.h.
7834 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7835 directly on the tree encoding.
7836 (const_vector_from_tree): Likewise.
7837 * optabs.c: Include rtx-vector-builder.h.
7838 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7839 sequence of "u" values.
7840 * vec-perm-indices.c: Include rtx-vector-builder.h.
7841 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7842 directly on the vec_perm_indices encoding.
7844 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7846 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7847 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7848 * rtx-vector-builder.h: New file.
7849 * rtx-vector-builder.c: Likewise.
7850 * rtl.h (rtx_def::u2): Add a const_vector field.
7851 (CONST_VECTOR_NPATTERNS): New macro.
7852 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7853 (CONST_VECTOR_DUPLICATE_P): Likewise.
7854 (CONST_VECTOR_STEPPED_P): Likewise.
7855 (CONST_VECTOR_ENCODED_ELT): Likewise.
7856 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7857 (unwrap_const_vec_duplicate): Likewise.
7858 (const_vec_series_p): Check for a non-duplicated vector encoding.
7859 Say that the function only returns true for integer vectors.
7860 * emit-rtl.c: Include rtx-vector-builder.h.
7861 (gen_const_vec_duplicate_1): Delete.
7862 (gen_const_vector): Call gen_const_vec_duplicate instead of
7863 gen_const_vec_duplicate_1.
7864 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7865 (gen_const_vec_duplicate): Use rtx_vector_builder.
7866 (gen_const_vec_series): Likewise.
7867 (gen_rtx_CONST_VECTOR): Likewise.
7868 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7869 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7870 Build a new vector rather than modifying a CONST_VECTOR in-place.
7871 (handle_special_swappables): Update call accordingly.
7872 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7873 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7874 Build a new vector rather than modifying a CONST_VECTOR in-place.
7875 (handle_special_swappables): Update call accordingly.
7877 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7879 * simplify-rtx.c (simplify_const_binary_operation): Use
7880 CONST_VECTOR_ELT instead of XVECEXP.
7882 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7884 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7885 the selector elements to be different from the data elements
7886 if the selector is a VECTOR_CST.
7887 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7888 ssizetype for the selector.
7890 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7892 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7893 before testing each element individually.
7894 * tree-vect-generic.c (lower_vec_perm): Likewise.
7896 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7898 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7899 * selftest-run-tests.c (selftest::run_tests): Call it.
7900 * vector-builder.h (vector_builder::operator ==): New function.
7901 (vector_builder::operator !=): Likewise.
7902 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7903 (vec_perm_indices::all_from_input_p): New function.
7904 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7905 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7906 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7907 instead of reading the VECTOR_CST directly. Detect whether both
7908 vector inputs are the same before constructing the vec_perm_indices,
7909 and update the number of inputs argument accordingly. Use the
7910 utility functions added above. Only construct sel2 if we need to.
7912 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7914 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7915 the broadcast of the low byte.
7916 (expand_mult_highpart): Use an explicit encoding for the permutes.
7917 * optabs-query.c (can_mult_highpart_p): Likewise.
7918 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7919 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7920 (vectorizable_bswap): Likewise.
7921 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7922 explicit encoding for the power-of-2 permutes.
7923 (vect_permute_store_chain): Likewise.
7924 (vect_grouped_load_supported): Likewise.
7925 (vect_permute_load_chain): Likewise.
7927 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7929 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7930 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7931 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7932 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7933 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7934 (vect_gen_perm_mask_any): Likewise.
7936 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7938 * int-vector-builder.h: New file.
7939 * vec-perm-indices.h: Include int-vector-builder.h.
7940 (vec_perm_indices): Redefine as an int_vector_builder.
7941 (auto_vec_perm_indices): Delete.
7942 (vec_perm_builder): Redefine as a stand-alone class.
7943 (vec_perm_indices::vec_perm_indices): New function.
7944 (vec_perm_indices::clamp): Likewise.
7945 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7946 (vec_perm_indices::new_vector): New function.
7947 (vec_perm_indices::new_expanded_vector): Update for new
7948 vec_perm_indices class.
7949 (vec_perm_indices::rotate_inputs): New function.
7950 (vec_perm_indices::all_in_range_p): Operate directly on the
7951 encoded form, without computing elided elements.
7952 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7953 encoding. Update for new vec_perm_indices class.
7954 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7955 the given vec_perm_builder.
7956 (expand_vec_perm_var): Update vec_perm_builder constructor.
7957 (expand_mult_highpart): Use vec_perm_builder instead of
7958 auto_vec_perm_indices.
7959 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7960 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7961 or double series encoding as appropriate.
7962 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7963 vec_perm_indices instead of auto_vec_perm_indices.
7964 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7965 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7966 (vect_permute_store_chain): Likewise.
7967 (vect_grouped_load_supported): Likewise.
7968 (vect_permute_load_chain): Likewise.
7969 (vect_shift_permute_load_chain): Likewise.
7970 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7971 (vect_transform_slp_perm_load): Likewise.
7972 (vect_schedule_slp_instance): Likewise.
7973 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7974 (vectorizable_mask_load_store): Likewise.
7975 (vectorizable_bswap): Likewise.
7976 (vectorizable_store): Likewise.
7977 (vectorizable_load): Likewise.
7978 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7979 vec_perm_indices instead of auto_vec_perm_indices. Use
7980 tree_to_vec_perm_builder to read the vector from a tree.
7981 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7982 vec_perm_builder instead of a vec_perm_indices.
7983 (have_whole_vector_shift): Use vec_perm_builder and
7984 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7985 truncation to calc_vec_perm_mask_for_shift.
7986 (vect_create_epilog_for_reduction): Likewise.
7987 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7988 from auto_vec_perm_indices to vec_perm_indices.
7989 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7990 instead of changing individual elements.
7991 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7992 the vector in d.perm.
7993 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7994 from auto_vec_perm_indices to vec_perm_indices.
7995 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7996 instead of changing individual elements.
7997 (arm_vectorize_vec_perm_const): Use new_vector to install
7998 the vector in d.perm.
7999 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8000 Update vec_perm_builder constructor.
8001 (rs6000_expand_interleave): Likewise.
8002 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8003 (rs6000_expand_interleave): Likewise.
8005 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8007 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8008 to qimode could truncate the indices.
8009 * optabs.c (expand_vec_perm_var): Likewise.
8011 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8013 * Makefile.in (OBJS): Add vec-perm-indices.o.
8014 * vec-perm-indices.h: New file.
8015 * vec-perm-indices.c: Likewise.
8016 * target.h (vec_perm_indices): Replace with a forward class
8018 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8019 * optabs.h: Include vec-perm-indices.h.
8020 (expand_vec_perm): Delete.
8021 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8022 (expand_vec_perm_const): Declare.
8023 * target.def (vec_perm_const_ok): Replace with...
8024 (vec_perm_const): ...this new hook.
8025 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8026 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8027 * doc/tm.texi: Regenerate.
8028 * optabs.def (vec_perm_const): Delete.
8029 * doc/md.texi (vec_perm_const): Likewise.
8030 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8031 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8032 expand_vec_perm for constant permutation vectors. Assert that
8033 the mode of variable permutation vectors is the integer equivalent
8034 of the mode that is being permuted.
8035 * optabs-query.h (selector_fits_mode_p): Declare.
8036 * optabs-query.c: Include vec-perm-indices.h.
8037 (selector_fits_mode_p): New function.
8038 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8039 is defined, instead of checking whether the vec_perm_const_optab
8040 exists. Use targetm.vectorize.vec_perm_const instead of
8041 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8042 fit in the vector mode before using a variable permute.
8043 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8044 vec_perm_indices instead of an rtx.
8045 (expand_vec_perm): Replace with...
8046 (expand_vec_perm_const): ...this new function. Take the selector
8047 as a vec_perm_indices rather than an rtx. Also take the mode of
8048 the selector. Update call to shift_amt_for_vec_perm_mask.
8049 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8050 Use vec_perm_indices::new_expanded_vector to expand the original
8051 selector into bytes. Check whether the indices fit in the vector
8052 mode before using a variable permute.
8053 (expand_vec_perm_var): Make global.
8054 (expand_mult_highpart): Use expand_vec_perm_const.
8055 * fold-const.c: Includes vec-perm-indices.h.
8056 * tree-ssa-forwprop.c: Likewise.
8057 * tree-vect-data-refs.c: Likewise.
8058 * tree-vect-generic.c: Likewise.
8059 * tree-vect-loop.c: Likewise.
8060 * tree-vect-slp.c: Likewise.
8061 * tree-vect-stmts.c: Likewise.
8062 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8064 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8065 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8066 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8067 (aarch64_vectorize_vec_perm_const): ...this new function.
8068 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8069 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8070 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8071 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8072 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8073 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8074 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8076 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8077 check for NEON modes.
8078 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8079 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8080 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8081 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8083 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8084 the old VEC_PERM_CONST conditions.
8085 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8086 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8087 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8088 (ia64_vectorize_vec_perm_const_ok): Merge into...
8089 (ia64_vectorize_vec_perm_const): ...this new function.
8090 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8091 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8092 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8093 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8094 * config/mips/mips.c (mips_expand_vec_perm_const)
8095 (mips_vectorize_vec_perm_const_ok): Merge into...
8096 (mips_vectorize_vec_perm_const): ...this new function.
8097 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8098 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8099 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8100 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8101 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8102 (rs6000_expand_vec_perm_const): Delete.
8103 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8105 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8106 (altivec_expand_vec_perm_const_le): Take each operand individually.
8107 Operate on constant selectors rather than rtxes.
8108 (altivec_expand_vec_perm_const): Likewise. Update call to
8109 altivec_expand_vec_perm_const_le.
8110 (rs6000_expand_vec_perm_const): Delete.
8111 (rs6000_vectorize_vec_perm_const_ok): Delete.
8112 (rs6000_vectorize_vec_perm_const): New function.
8113 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8114 an element count and rtx array.
8115 (rs6000_expand_extract_even): Update call accordingly.
8116 (rs6000_expand_interleave): Likewise.
8117 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8118 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8119 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8120 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8121 (rs6000_expand_vec_perm_const): Delete.
8122 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8123 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8124 (altivec_expand_vec_perm_const_le): Take each operand individually.
8125 Operate on constant selectors rather than rtxes.
8126 (altivec_expand_vec_perm_const): Likewise. Update call to
8127 altivec_expand_vec_perm_const_le.
8128 (rs6000_expand_vec_perm_const): Delete.
8129 (rs6000_vectorize_vec_perm_const_ok): Delete.
8130 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8131 reference to the SPE evmerge intructions.
8132 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8133 an element count and rtx array.
8134 (rs6000_expand_extract_even): Update call accordingly.
8135 (rs6000_expand_interleave): Likewise.
8136 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8137 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8139 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8141 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8143 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8144 vector mode and that that mode matches the mode of the data
8146 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8147 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8148 directly using expand_vec_perm_1 when forcing selectors into
8150 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8152 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8154 * optabs-query.h (can_vec_perm_p): Delete.
8155 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8156 * optabs-query.c (can_vec_perm_p): Split into...
8157 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8158 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8159 particular selector is valid.
8160 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8161 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8162 (vect_grouped_load_supported): Likewise.
8163 (vect_shift_permute_load_chain): Likewise.
8164 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8165 (vect_transform_slp_perm_load): Likewise.
8166 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8167 (vectorizable_bswap): Likewise.
8168 (vect_gen_perm_mask_checked): Likewise.
8169 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8170 implementations of variable permutation vectors into account
8171 when deciding which selector to use.
8172 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8173 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8174 with a false third argument.
8175 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8176 to test whether the constant selector is valid and can_vec_perm_var_p
8177 to test whether a variable selector is valid.
8179 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8181 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8182 * optabs-query.c (can_vec_perm_p): Likewise.
8183 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8184 instead of vec_perm_indices.
8185 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8186 (vect_gen_perm_mask_checked): Likewise,
8187 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8188 (vect_gen_perm_mask_checked): Likewise,
8190 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8192 * optabs-query.h (qimode_for_vec_perm): Declare.
8193 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8194 (qimode_for_vec_perm): ...this new function.
8195 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8197 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8199 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8200 does not have a conditional at the top.
8202 2018-01-02 Richard Biener <rguenther@suse.de>
8204 * ipa-inline.c (big_speedup_p): Fix expression.
8206 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8209 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8212 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8216 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8217 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8218 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8219 cond_taken_branch_cost 3->4.
8221 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8223 PR tree-optimization/83581
8224 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8225 TODO_cleanup_cfg if any changes have been made.
8228 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8229 convert_modes if target mode has the right side, but different mode
8233 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8234 last argument when extracting from CONCAT. If either from_real or
8235 from_imag is NULL, use expansion through memory. If result is not
8236 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8237 the parts directly to inner mode, if even that fails, use expansion
8241 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8242 check for bswap in mode rather than HImode and use that in expand_unop
8245 Copyright (C) 2018 Free Software Foundation, Inc.
8247 Copying and distribution of this file, with or without modification,
8248 are permitted in any medium without royalty provided the copyright
8249 notice and this notice are preserved.