1 2018-11-19 Eric Botcazou <ebotcazou@adacore.com>
3 * compare-elim.c (struct comparison): Add not_in_a field.
4 (is_not): New static function.
6 (conforming_compare): Handle a NOT in the first operand.
7 (can_eliminate_compare): Likewise.
8 (find_comparison_dom_walker::before_dom_children): Likewise.
9 (try_eliminate_compare): Likewise.
10 * config/visium/visium.md (negsi2_insn_set_carry): Turn into...
11 (neg<mode>2_insn_set_carry): ...this and add missing NEG operation.
13 2018-11-19 Jonathan Wakely <jwakely@redhat.com>
15 * doc/extend.texi (Common Type Attributes): Fix typos.
17 2018-11-19 Martin Liska <mliska@suse.cz>
20 * coverage.c (coverage_begin_function): Add assert.
22 2018-11-18 Sandra Loosemore <sandra@codesourcery.com>
25 * doc/extend.texi (Common Function Attributes): Document that
26 no_instrument_function applies to -p and -pg, too.
27 * doc/invoke.texi (Instrumentation Options): Add cross-references
28 to docs for -p, -pg, and -finstrument-functions.
30 2018-11-18 Alan Modra <amodra@gmail.com>
32 * config/gnu-user.h (GNU_USER_TARGET_CRTI): Define.
33 (GNU_USER_TARGET_STARTFILE_SPEC): Use it here.
34 (GNU_USER_TARGET_CRTN): Define.
35 (GNU_USER_TARGET_ENDFILE_SPEC): Use it here.
36 (GNU_USER_TARGET_LINK_GCC_C_SEQUENCE_SPEC): Define.
38 2018-11-18 Alan Modra <amodra@gmail.com>
40 * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Delete
42 (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
44 2018-11-17 Nick Clifton <nickc@redhat.com>
45 Sandra Loosemore <sandra@codesourcery.com>
48 * doc/invoke.texi (Overall Options): Document that --help and
49 --help=<value> options cannot be combined.
51 2018-11-17 Jim Wilson <jimw@sifive.com>
53 * config/riscv/riscv.c (epilogue_cfa_sp_offset): New.
54 (riscv_restore_reg): If restoring HARD_FRAME_POINTER_REGNUM, and
55 epilogue_cfa_sp_offset set, then add REG_CFA_DEF_CFA regnote.
56 (riscv_expand_epilogue): Initialize epilogue_cfa_sp_offset. Set it
57 to step2 if frame_pointer_needed and step1 is 0.
59 2018-11-17 Sandra Loosemore <sandra@codesourcery.com>
62 * doc/cpp.texi (System Headers): Add note about implicit
63 extern "C" block on targets that define SYSTEM_IMPLICIT_EXTERN_C.
65 2018-11-17 Sandra Loosemore <sandra@codesourcery.com>
68 * doc/invoke.texi (C++ Dialect Options): Clarify usage of
69 -fno-implicit-templates.
71 2018-11-17 Jakub Jelinek <jakub@redhat.com>
73 PR tree-optimization/87546
74 * tree-vect-patterns.c (vect_look_through_possible_promotion): Add
75 min_precision variable, initially set it to orig_precision, only does
76 something if op_type's precision is <= min_precision and update
77 min_precision whenever calling set_op.
79 2018-11-16 Jan Hubicka <hubicka@ucw.cz>
82 * ipa-devirt.c (warn_odr): Look for main variant to get TYPE_DECL.
84 2018-11-16 Sandra Loosemore <sandra@codesourcery.com>
86 * doc/invoke.texi (Option Summary): Fix whitespace and line
87 breaks in @gccoptlist environments.
88 (Warning Options): Likewise.
89 (Optimize Options): Likewise.
90 (PowerPC SPE Options): Likewise.
91 (RS/6000 and PowerPC Options): Likewise.
93 2018-11-16 Jeff Law <law@redhat.com>
95 * config/mn10300/mn10300.md (adddi3_degenerate): Remove bogus
98 2018-11-16 Sandra Loosemore <sandra@codesourcery.com>
101 * doc/invoke.texi (Optimize Options): Update options enabled by
102 fprofile-generate, -fprofile-use, and -fauto-profile.
104 2018-11-16 Jozef Lawrynowicz <jozef.l@mittosystems.com>
107 * target-def.h: Initialize TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP.
108 Add them to the TARGET_ASM_{,UN}ALIGNED_INT_OP structs.
109 * target.def: Enumerate TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP in
111 * target.h: Add psi, pdi, pti to struct asm_int_op definition.
112 * targhooks.c (default_print_patchable_function_entry): Assert
113 asm_int_op does not return a NULL string.
114 * varasm.c (integer_asm_op): Return the op for a partial int type
115 when the requested size does not correspond to an integer type.
116 * config/msp430/msp430.c: Initialize TARGET_ASM_{,UN}ALIGNED_PSI_OP.
117 * doc/tm.texi: Regenerate.
119 2018-11-12 Jason Merrill <jason@redhat.com>
121 * gimplify.c (gimplify_case_label_expr): Handle hot/cold attributes.
123 2018-11-16 Michael Meissner <meissner@linux.ibm.com>
125 * config/rs6000/constraints.md (wF constraint): Remove power9
126 fusion documentation. Just document wF's use for power8 fusion.
127 * config/rs6000/predicates.md (p9_fusion_reg_operand): Delete.
128 (fusion_gpr_addis): Delete power9 fusion support. Change power8
129 fusion support to require the upper 12 bits to be all 0's or all
131 (fusion_gpr_mem_load): Add comment.
132 (fusion_addis_mem_combo_load): Remove power9 fusion support.
133 (fusion_addis_mem_combo_store): Delete.
134 (fusion_offsettable_mem_operand): Delete.
135 * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Do not set
137 (ISA_3_0_MASKS_SERVER): Delete power9 fusion.
138 (POWERPC_MASKS): Delete power9 fusion.
139 * config/rs6000/rs6000-protos.h (emit_fusion_load_store): Delete.
140 (fusion_p9_p): Delete.
141 (expand_fusion_p9_load): Delete.
142 (expand_fusion_p9_store): Delete.
143 (emit_fusion_p9_load): Delete.
144 (emit_fusion_p9_store): Delete.
145 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Delete power9
147 (rs6000_option_override_internal): Set power8 fusion based on
148 whether we are tuning for power8. Delete power9 fusion support.
149 (rs6000_opt_masks): Delete -mpower9-fusion switch.
150 (emit_fusion_load): Rename emit_fusion_load_store to
151 emit_fusion_load, and drop fusion store support. Update callers.
152 (emit_fusion_load_store): Likewise.
153 (emit_fusion_gpr_load): Likewise.
154 (fusion_p9_p): Delete.
155 (expand_fusion_p9_load): Delete.
156 (expand_fusion_p9_store): Delete.
157 (emit_fusion_p9_load): Delete.
158 (emit_fusion_p9_store): Delete.
159 * config/rs6000/rs6000.md (UNSPEC_FUSION_P9): Delete.
160 (GPR_FUSION): Delete.
161 (FPR_FUSION): Delete.
162 (power9 fusion peephole2s): Delete.
163 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Delete.
164 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Delete.
165 (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_load): Delete.
166 (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_store): Delete.
167 (fusion_p9_<mode>_constant): Delete.
168 * config/rs6000/rs6000.opt (-mpower9-fusion): Delete undocumented
169 power9 fusion switch.
170 * doc/md.texi (PowerPC constraints): Update wF constraint
171 documentation for power8 fusion only.
173 2018-11-16 Jakub Jelinek <jakub@redhat.com>
175 PR rtl-optimization/87475
176 * cfgrtl.c (patch_jump_insn): Allow redirection failure for
177 CROSSING_JUMP_P insns.
178 (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
180 2018-11-16 Uros Bizjak <ubizjak@gmail.com>
183 * config/i386/i386.md (floatunsdidf2): Allow only 64bit AVX512F targets.
184 * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
185 (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
187 2018-11-16 Jakub Jelinek <jakub@redhat.com>
190 * optabs.c (expand_binop): For op0_mode use GET_MODE (op0), unless it
191 is VOIDmode, in which case use int_mode. Similarly for op1_mode.
193 2018-11-16 Eric Botcazou <ebotcazou@adacore.com>
195 * tree-switch-conversion.h (switch_decision_tree::emit_case_nodes): Add
196 location_t parameter.
197 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
198 (switch_decision_tree::do_jump_if_equal): Likewise.
199 * tree-switch-conversion.c (switch_decision_tree::emit): Pass location
200 of switch statement to emit_case_nodes.
201 (switch_decision_tree::emit_cmp_and_jump_insns): Add LOC parameter and
202 set it on the newly built GIMPLE comparison statement.
203 (switch_decision_tree::do_jump_if_equal): Likewise.
204 (switch_decision_tree::emit_case_nodes): Add LOC parameter and pass it
205 in calls to do_jump_if_equal as well as recursive calls.
207 2018-11-16 Andi Kleen <ak@linux.intel.com>
209 * config/i386/i386.md: Allow memory operands to ptwrite.
211 2018-11-16 Michael Matz <matz@suse.de>
213 * system.h (PRsa): New macro.
214 (SIZE_AMOUNT): Cast number to uint64_t.
215 * alloc-pool.h (pool_usage::dump): Don't use %zu but PRsa.
216 (pool_usage::dump_footer): Likewise and also use PRIu64.
217 * bitmap.h (bitmap_usage::dump): Likewise.
218 * ggc-common.c (ggc_usage::dump): Likewise.
219 * ggc-page.c (ggc_print_statistics): Likewise.
220 * input.c (dump_line_table_statistics): Likewise.
221 * mem-stats.h (mem_usage::dump): Likewise.
222 (mem_usage::dump_footer): Likewise.
223 * rtl.c (dump_rtx_statistics): Likewise.
224 * tree-cfg.c (dump_cfg_stats): Likewise.
225 * tree-dfa.c (dump_dfa_stats): Likewise.
226 * tree-phinodes.c (phinodes_print_statistics): Likewise.
227 * tree-ssanames (ssanames_print_statistics): Likewise.
228 * vec.c (vec_usage::dump): Likewise.
229 (vec_usage::dump_footer): Likewise.
231 2018-11-16 Richard Biener <rguenther@suse.de>
233 PR tree-optimization/88011
234 * tree-vrp.c (extract_range_from_binary_expr): Fix error in
235 replacing set_value_range_to_undefined and
236 set_value_range_to_varying with method calls.
238 2018-11-16 Ilya Leoshkevich <iii@linux.ibm.com>
240 * config/s390/s390.md
241 (*r<noxa>sbg_<mode>_srl_bitmask): Do not delegate arithmetic to
243 (*r<noxa>sbg_<mode>_sll): Likewise.
244 (*r<noxa>sbg_<mode>_srl): Likewise.
245 (*r<noxa>sbg_sidi_srl): New pattern.
247 2018-11-16 Jerome Lambourg <lambourg@adacore.com>
249 * config/vxworks.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): New macro.
250 Default to TARGET_VXWORKS_RTP.
251 (SUPPORTS_INIT_PRIORITY): Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
252 of TARGET_VXWORKS_RTP.
253 * config/vxworksae.h: Also define TARGET_VXWORKS_HAVE_CTORS_DTORS.
254 * config/vxworks.c: Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
255 of TARGET_VXWORKS_RTP to set targetm.have_ctors_dtors.
257 2018-11-15 Sandra Loosemore <sandra@codesourcery.com>
260 * doc/extend.texi (Common Type Attributes): Make it explicit
261 that attribute "packed" can apply to C++ classes.
263 2018-11-15 Martin Sebor <msebor@redhat.com>
267 * tree.c (type_argument_type): New function.
268 * tree.h (type_argument_type): Declare it.
269 * gcc/doc/extend.texi (alloc_align): Update and clarify.
270 (alloc_size, nonnull, sentinel): Same.
272 2018-11-15 Andrew Stubbs <ams@codesourcery.com>
273 Kwok Cheung Yeung <kcy@codesourcery.com>
275 * tree-vect-stmts.c (vectorizable_store): Don't ICE when
276 int_mode_for_size fails.
277 (vectorizable_load): Likewise.
279 2018-11-15 David Malcolm <dmalcolm@redhat.com>
281 * doc/ux.texi (Group logically-related diagnostics): Move
282 discussion of auto_diagnostic_group into this new subsection.
283 Give an example of where this grouping is used.
285 2018-11-15 David Malcolm <dmalcolm@redhat.com>
288 * Makefile.in (OBJS): Move json.o to...
289 (OBJS-libcommon): ...here and add diagnostic-format-json.o.
290 * common.opt (fdiagnostics-format=): New option.
291 (diagnostics_output_format): New enum.
292 * diagnostic-format-json.cc: New file.
293 * diagnostic.c (default_diagnostic_final_cb): New function, taken
294 from start of diagnostic_finish.
295 (diagnostic_initialize): Initialize final_cb to
296 default_diagnostic_final_cb.
297 (diagnostic_finish): Move "being treated as errors" messages to
298 default_diagnostic_final_cb. Call any final_cb.
299 (default_diagnostic_finalizer): Add diagnostic_t param.
300 (diagnostic_report_diagnostic): Pass "orig_diag_kind" to
301 diagnostic_finalizer callback.
302 * diagnostic.h (enum diagnostics_output_format): New enum.
303 (diagnostic_finalizer_fn): Reimplement, adding diagnostic_t param.
304 (struct diagnostic_context): Add "final_cb".
305 (default_diagnostic_finalizer): Add diagnostic_t param.
306 (diagnostic_output_format_init): New decl.
307 * doc/invoke.texi (-fdiagnostics-format): New option.
308 * dwarf2out.c (gen_producer_string): Ignore
309 OPT_fdiagnostics_format_.
310 * gcc.c (driver_handle_option): Handle OPT_fdiagnostics_format_.
311 * lto-wrapper.c (append_diag_options): Ignore it.
312 * opts.c (common_handle_option): Handle it.
314 2018-11-15 David Malcolm <dmalcolm@redhat.com>
316 PR tree-optimization/88015
317 * graphite-isl-ast-to-gimple.c
318 (translate_isl_ast_to_gimple::scop_to_isl_ast): Add missing check
320 * graphite-sese-to-poly.c (build_poly_scop): Likewise.
322 2018-11-15 Richard Biener <rguenther@suse.de>
325 * gimple.c (gimple_call_flags): Union flags from decl, type
327 * trans-mem.c (is_tm_pure_call): Simplify.
329 2018-11-15 Richard Biener <rguenther@suse.de>
331 PR tree-optimization/88031
332 * tree-vect-loop.c (vectorizable_reduction): Move check
333 for multiple types earlier so we get the expected dump.
334 Simplify calls to vectorizable_condition.
335 * tree-vect-stmts.h (vectorizable_condition): Update prototype.
336 * tree-vect-stmts.c (vectorizable_condition): Instead of
337 reduc_def and reduc_index take just a flag. Simplify
338 code-generation now that we can rely on the defs being set up.
339 (vectorizable_comparison): Remove unused argument.
341 2018-11-15 Richard Biener <rguenther@suse.de>
343 PR tree-optimization/88030
344 * tree-complex.c (need_eh_cleanup): New global.
345 (update_complex_assignment): Mark blocks that need EH update.
346 (expand_complex_comparison): Likewise.
347 (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
348 perform EH cleanup and schedule CFG cleanup if that did anything.
350 2018-11-15 Jakub Jelinek <jakub@redhat.com>
352 PR rtl-optimization/88018
353 * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
354 on the presence of fallthru edge, rather than if it is a USE or not.
356 2018-11-15 Richard Biener <rguenther@suse.de>
359 * tree-data-ref.c (analyze_miv_subscript): Guard calls to
360 analyze_subscript_affine_affine properly.
362 2018-11-15 Xianmiao Qu <xianmiao_qu@c-sky.com>
364 * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove.
365 (GLIBC_DYNAMIC_LINKER): Define.
366 (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name.
368 2018-11-15 Bin Cheng <bin.cheng@linux.alibaba.com>
370 PR tree-optimization/84648
371 * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New.
372 (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case
373 by calling adjust_cond_for_loop_until_wrap.
375 2018-11-15 Sandra Loosemore <sandra@codesourcery.com>
378 * doc/extend.texi (Common Function Attributes): Clarify linker
379 restrictions on "aligned" attribute.
380 (Common Variable Attributes): Likewise. Mention that linker
381 restrictions don't apply to stack-allocated variables.
383 2018-11-15 Joern Rennecke <joern.rennecke@riscy-ip.com>
385 * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT
388 2018-11-14 Sandra Loosemore <sandra@codesourcery.com>
392 * doc/invoke.texi (Optimize Options): Remove bad example about
393 interaction between -flto and -O. Replace it with a note that
394 you need to compile with -O and not just link. Copy-edit -flto
395 discussion to reduce verbiage and improve flow.
397 2018-11-14 Sandra Loosemore <sandra@codesourcery.com>
400 * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og
401 also suppress many optimizations. Alphabetize option lists for
402 -O1, -O2, and -Os. Add list of options disabled with -Og, and
403 correct documentation for those options to say that.
404 * opts.c (default_options_table): Sort table by level and option
405 name, to make it easier to correlate to the manual.
407 2018-11-14 Uros Bizjak <ubizjak@gmail.com>
409 * config/i386/i386.c (ix86_print_operand_address_as): Simplify
410 printing of the names of segment registers.
412 2018-11-14 Nathan Sidwell <nathan@acm.org>
416 * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
419 2018-11-14 David Malcolm <dmalcolm@redhat.com>
421 * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC).
423 2018-11-14 Jakub Jelinek <jakub@redhat.com>
425 P1236R1 - Signed integers are two's complement
426 * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a.
429 * hash-map.h (hash_map::iterator::reference_pair): New class.
430 (hash_map::iterator::operator*): Return it rather than std::pair.
432 2018-11-14 Jeff Law <law@redhat.com>
434 * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force
435 iff the operand is a constant.
437 2018-11-14 Aldy Hernandez <aldyh@redhat.com>
439 * gimple-ssa-evrp-analyze.c
440 (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename
441 ignore_equivs_equal_p to equal_p.
442 * ipa-cp.c (meet_with_1): Use equal_p instead of
443 ignore_equivs_equal_p.
444 * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same.
445 * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove.
446 (value_range::operator==): Remove.
447 (value_range::operator!=): Remove.
448 (vrp_prop::visit_stmt): Use equal_p.
449 * tree-vrp.h (value_range): Remove operator==, operator!=,
450 ignore_equivs_equal_p.
451 * vr-values.c (update_value_range): Use equal_p.
453 2018-11-14 Michael Matz <matz@suse.de>
456 * gimplify.c (collect_fallthrough_labels): Add new argument,
457 return location via that, don't modify statements.
458 (warn_implicit_fallthrough_r): Adjust call, don't use
459 statement location directly.
461 2018-11-14 Richard Biener <rguenther@suse.de>
464 * tree-data-ref.c (split_constant_offset): Add wrapper
465 allocating a cache hash-map.
466 (split_constant_offset_1): Cache results of expanding
467 expressions from SSA def stmts.
469 2018-11-14 Richard Biener <rguenther@suse.de>
472 * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument
474 (lambda_vector_mult_const): Likewise.
475 (lambda_matrix_right_hermite): Use lambda_int temporaries.
477 2018-11-14 Wilco Dijkstra <wdijkstr@arm.com>
478 Jackson Woodruff <jackson.woodruff@arm.com>
480 PR 71026/tree-optimization
481 * match.pd: Simplify floating point comparisons.
483 2018-11-14 Jakub Jelinek <jakub@redhat.com>
485 PR rtl-optimization/87817
486 * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
487 last argument has low 8 bits clear, fold to 0.
489 2018-11-14 Iain Sandoe <iain@sandoe.co.uk>
491 * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
492 MAX_OFILE_ALIGNMENT as needed.
494 2018-11-14 Richard Biener <rguenther@suse.de>
496 PR tree-optimization/88019
497 * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
498 COND_EXPR arguments to non-trapping overflow.
500 2018-11-14 Claudiu Zissulescu <claziss@synopsys.com>
502 * config/arc/arc.h (reg_class): Reorder registers classes, remove
503 unused register classes.
504 (REG_CLASS_NAMES): Likewise.
505 (REG_CLASS_CONTENTS): Likewise.
506 (FIXED_REGISTERS): Make lp_count fixed.
507 (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
508 (PROGRAM_COUNTER_REGNO): Remove.
509 * config/arc/arc.c (arc_conditional_register_usage): Remove unused
510 register classes, use constants for register numbers, remove
511 reg_class_contents references.
512 (arc_process_double_reg_moves): Add asserts.
513 (arc_secondary_reload): Remove LPCOUNT_REG reference, use
514 lra_in_progress predicate.
515 (arc_init_reg_tables): Remove unused register classes.
516 (arc_register_move_cost): Likewise.
517 (arc_preferred_reload_class): Likewise.
518 (hwloop_optimize): Update rtx patterns involving lp_count register.
519 (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros.
520 * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
521 CHEAP_CORE_REGS. Former one will be used for LRA.
522 ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS. Former
523 one will be used for LRA.
524 ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS. Former
525 one will be used for LRA.
526 ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
527 Former one will be used for LRA.
528 ("f"): Delete constraint.
531 ("l"): Change it from register constraint to constraint.
532 * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
533 (movhi_insn): Likewise.
534 (movsi_insn): Update pattern.
537 (stack_tie): Remove 'b' constraint letter.
539 (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
540 (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
541 (R57_REG, R59_REG, PCL_REG): Likewise.
542 (ILINK1_REGNUM): Renamed to ILINK1_REG.
543 (ILINK2_REGNUM): Renamed to ILINK2_REG.
546 (Rcw): Remove unused reg classes.
547 * config/arc/predicates.md (dest_reg_operand): Just default on
548 register_operand predicate.
549 (mpy_dest_reg_operand): Likewise.
550 (move_dest_operand): Use macros instead of constants.
551 * config/arc/arc.opt (mlra): Switch to lra as default.
553 2018-11-14 Richard Biener <rguenther@suse.de>
555 PR tree-optimization/87974
556 * tree-vect-loop.c (vectorizable_reduction): When computing
557 the vectorized reduction PHI vector type ignore constant
560 2018-11-14 Jakub Jelinek <jakub@redhat.com>
562 PR tree-optimization/87977
563 * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
564 stmt, build a new one and replace the old one with it. Formatting fix.
565 Call release_ssa_name (x) if !has_other_use and !delete_div.
566 (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
567 verify lhs of stmt is still def.
569 2018-11-13 Peter Bergner <bergner@linux.ibm.com>
571 PR rtl-optimization/87507
572 * lower-subreg.c (operand_for_swap_move_operator): New function.
573 (simple_move): Strip simple operators.
574 (find_pseudo_copy): Likewise.
575 (resolve_operand_for_swap_move_operator): New function.
576 (resolve_simple_move): Strip simple operators and swap operands.
578 2018-11-13 Jakub Jelinek <jakub@redhat.com>
580 PR tree-optimization/87898
581 * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
582 (ipa_simd_modify_function_body): Remove debug stmts where the first
583 argument was changed into a non-decl.
585 2018-11-13 Peter Bergner <bergner@linux.ibm.com>
587 PR rtl-optimization/87899
588 * lra-lives.c (start_living): Update white space in comment.
589 (enum point_type): New.
590 (sparseset_contains_pseudos_p): New function.
591 (update_pseudo_point): Likewise.
592 (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
593 (make_hard_regno_dead): Likewise. Remove ignore_reg_for_conflicts
594 handling. Move early exit after adding conflicts.
595 (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro. Add early exit
596 if regno is already live. Remove all handling of program points.
597 (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro. Add early exit
598 after adding conflicts. Remove all handling of program points and
599 ignore_reg_for_conflicts.
600 (mark_regno_live): Use HARD_REGISTER_NUM_P macro. Remove return value
601 and do not guard call to mark_pseudo_live.
602 (mark_regno_dead): Use HARD_REGISTER_NUM_P macro. Remove return value
603 and do not guard call to mark_pseudo_dead.
604 (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
605 (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
606 Use new function update_pseudo_point. Handle register copies by
607 removing the source register from the live set. Handle INOUT operands.
608 Update to the next program point using the unused_set, dead_set and
610 (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
612 2018-11-13 David Malcolm <dmalcolm@redhat.com>
614 * builtins.c: Replace "source_location" with "location_t".
615 * diagnostic-show-locus.c: Likewise.
616 * diagnostic.c: Likewise.
617 * dumpfile.c: Likewise.
618 * gcc-rich-location.h: Likewise.
619 * genmatch.c: Likewise.
620 * gimple.h: Likewise.
621 * gimplify.c: Likewise.
623 * input.h: Likewise. Eliminate the typedef.
624 * omp-expand.c: Likewise.
625 * selftest.h: Likewise.
626 * substring-locations.h (get_source_location_for_substring):
628 (get_location_within_string): ...this.
629 * tree-cfg.c: Replace "source_location" with "location_t".
630 * tree-cfgcleanup.c: Likewise.
631 * tree-diagnostic.c: Likewise.
632 * tree-into-ssa.c: Likewise.
633 * tree-outof-ssa.c: Likewise.
634 * tree-parloops.c: Likewise.
635 * tree-phinodes.c: Likewise.
636 * tree-phinodes.h: Likewise.
637 * tree-ssa-loop-ivopts.c: Likewise.
638 * tree-ssa-loop-manip.c: Likewise.
639 * tree-ssa-phiopt.c: Likewise.
640 * tree-ssa-phiprop.c: Likewise.
641 * tree-ssa-threadupdate.c: Likewise.
642 * tree-ssa.c: Likewise.
643 * tree-ssa.h: Likewise.
644 * tree-vect-loop-manip.c: Likewise.
646 2018-11-13 Michael Ploujnikov <michael.ploujnikov@oracle.com>
648 * doc/extend.texi: Fix typo in the weakref description.
650 2018-11-13 Richard Biener <rguenther@suse.de>
652 PR tree-optimization/86991
653 * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
654 group building until we have successfully detected the SLP
656 (vect_is_simple_reduction): Remove fixup code here.
658 2018-11-13 David Malcolm <dmalcolm@redhat.com>
660 * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
661 (dump_gimple_stmt): Use it.
662 (dump_gimple_stmt_loc): Likewise.
663 (dump_gimple_expr): Likewise.
664 (dump_gimple_expr_loc): Likewise.
665 (dump_generic_expr): Likewise.
666 (dump_generic_expr_loc): Likewise.
667 (dump_printf): Likewise.
668 (dump_printf_loc): Likewise.
669 (dump_dec): Likewise.
670 (dump_dec): Likewise.
671 (dump_hex): Likewise.
672 (dump_symtab_node): Likewise.
673 * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
674 Guard dump call with dump_enabled_p.
675 * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
676 * graphite-optimize-isl.c (optimize_isl): Likewise.
677 * graphite.c (graphite_transform_loops): Likewise.
678 * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
679 * tree-parloops.c (parallelize_loops): Likewise.
680 * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
681 * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
682 (vect_prune_runtime_alias_test_list): Likewise.
683 * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
684 (vect_estimate_min_profitable_iters): Likewise.
685 * tree-vect-slp.c (vect_record_max_nunits): Likewise.
686 (vect_build_slp_tree_2): Likewise.
687 (vect_supported_load_permutation_p): Likewise.
688 (vect_slp_analyze_operations): Likewise.
689 (vect_slp_analyze_bb_1): Likewise.
690 (vect_slp_bb): Likewise.
691 * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
692 * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
693 (pass_slp_vectorize::execute): Likewise.
694 (increase_alignment): Likewise.
696 2018-11-13 David Malcolm <dmalcolm@redhat.com>
699 * ipa-inline.c (report_inline_failed_reason): Guard calls to
700 cl_target_option_print_diff and cl_optimization_print_diff with
703 2018-11-13 David Malcolm <dmalcolm@redhat.com>
705 * doc/invoke.texi (-fsave-optimization-record): Note that the
706 output is compressed.
707 * optinfo-emit-json.cc: Include <zlib.h>.
708 (optrecord_json_writer::write): Compress the output.
710 2018-11-13 Aldy Hernandez <aldyh@redhat.com>
712 * tree-vrp.c (value_range_base::dump): Dump type.
713 Do not use INF nomenclature for 1-bit types.
714 (dump_value_range): Group all variants to common dumping code.
715 (debug): New overloaded functions for value_ranges.
716 (value_range_base::dump): Remove no argument version.
717 (value_range::dump): Same.
719 2018-11-13 Richard Biener <rguenther@suse.de>
721 PR tree-optimization/87931
722 * tree-vect-loop.c (vect_is_simple_reduction): Restrict
723 nested cycles we support to latch computations vectorizable_reduction
726 2018-11-13 Martin Liska <mliska@suse.cz>
728 PR tree-optimization/87885
729 * cfghooks.c (account_profile_record): Rename
731 (profile_record_check_consistency): ... this.
732 Calculate missing num_mismatched_freq_in.
733 (profile_record_account_profile): New function
734 that calculates time and size of a function.
735 * cfghooks.h (struct profile_record): Remove
737 (struct cfg_hooks): Remove after_pass flag.
738 (account_profile_record): Rename to ...
739 (profile_record_check_consistency): ... this.
740 (profile_record_account_profile): New.
741 * cfgrtl.c (rtl_account_profile_record): Remove
743 * passes.c (check_profile_consistency): Do only
745 (account_profile): Calculate size and time of
747 (pass_manager::dump_profile_report): Reformat
749 (execute_one_ipa_transform_pass): Call
750 consistency check before clean upand call account_profile
751 after a clean up is done.
752 (execute_one_pass): Call check_profile_consistency and
753 account_profile instead of using after_pass flag..
754 * tree-cfg.c (gimple_account_profile_record): Likewise.
756 2018-11-13 Martin Liska <mliska@suse.cz>
759 * config/i386/i386.c (ix86_option_override_internal): Error
760 about usage -mabi=ms and -fsanitize={,kernel-}address.
762 2018-11-13 Andre Vieira <andre.simoesdiasvieira@arm.com>
764 * config/aarch64/aarch64.c
765 (aarch64_vectorize_preferred_vector_alignment): Change return type to
767 (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
768 alignment being a poly int.
769 * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
770 return type to poly_uint64.
771 * target.def (default_preferred_vector_alignment): Likewise.
772 * targhooks.c (default_preferred_vector_alignment): Likewise.
773 * targhooks.h (default_preferred_vector_alignment): Likewise.
774 * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
775 (vect_compute_data_ref_alignment): Adapt to vector alignment being a
777 (vect_update_misalignment_for_peel): Likewise.
778 (vect_enhance_data_refs_alignment): Likewise.
779 (vect_find_same_alignment_drs): Likewise.
780 (vect_duplicate_ssa_name_ptr_info): Likewise.
781 (vect_setup_realignment): Likewise.
782 (vect_can_force_dr_alignment_p): Change alignment parameter type to
784 * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
785 mask with a compile time variable vector alignment.
786 (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
788 (vect_do_peeling): Exit early if vector alignment is not constant.
789 * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
791 (vectorizable_store): Likewise.
792 (vectorizable_load): Likweise.
793 * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
795 (vect_known_alignment_in_bytes): Adapt to vector alignment being a
797 (vect_can_force_dr_alignment_p): Change alignment parameter type to
800 2018-11-13 Richard Biener <rguenther@suse.de>
802 PR tree-optimization/87962
803 * tree-vect-loop.c (vect_is_simple_reduction): More reliably
804 detect outer reduction for disqualifying in-loop uses.
806 2018-11-13 Richard Biener <rguenther@suse.de>
808 PR tree-optimization/87967
809 * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
810 for constants for the scalar loop.
812 2018-11-13 Alan Modra <amodra@gmail.com>
814 * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
815 integer constants when -mcmodel=medium.
817 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
819 * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
820 * config/arc/elf.h (PROFILE_HOOK): Define.
821 * config/arc/linux.h (PROFILE_HOOK): Likewise.
823 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
825 * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
826 jump table data in the text section.
828 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
830 * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
831 to fit the eh_return pattern.
832 * config/arc/arc.md (eh_return): Define.
833 (VUNSPEC_ARC_EH_RETURN): Likewise.
834 * config/arc/arc-protos.h (arc_eh_return_address_location): Match
836 * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
839 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
841 * common/config/arc/arc-common.c (arc_option_optimization_table):
842 Millicode optimization is default on for size optimizations.
843 * config/arc/arc-protos.h (arc_check_multi): New function.
844 * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
845 (ENTER_LEAVE_START_REG): Define.
846 (ENTER_LEAVE_END_REG): Likewise.
847 (arc_override_options): Disable millicode when long calls option
849 (arc_frame_info): Change it from int to bool.
850 (arc_compute_frame_size): Clean up.
851 (arc_save_restore): Remove.
852 (frame_save_reg): New function.
853 (frame_restore_reg): Likewise.
854 (arc_enter_leave_p): Likewise.
855 (arc_save_callee_saves): Likewise.
856 (arc_restore_callee_saves): Likewise.
857 (arc_save_callee_enter): Likewise.
858 (arc_restore_callee_leave): Likewise.
859 (arc_save_callee_milli): Likewise.
860 (arc_restore_callee_milli): Likewise.
861 (arc_expand_prologue): Reimplement to emit enter/leave
863 (arc_expand_epilogue): Likewise.
864 (arc_check_multi): New function.
865 * config/arc/arc.md (push_multi_fp): New pattern.
866 (push_multi_fp_blink): Likewise.
867 (pop_multi_fp): Likewise.
868 (pop_multi_fp_blink): Likewise.
869 (pop_multi_fp_ret): Likewise.
870 (pop_multi_fp_blink_ret): Likewise.
871 * config/arc/arc.opt (mmillicode): Update option.
872 (mcode-density-frame): New option.
873 * config/arc/predicates.md (push_multi_operand): New predicate.
874 (pop_multi_operand): Likewise.
875 * doc/invoke.texi (ARC): Update ARC options information.
878 2018-11-13 Claudiu Zissulescu <claziss@synopsys.com>
880 * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
881 * config/arc/arc.c (operands_ok_ldd_std): New function.
882 (mem_ok_for_ldd_std): Likewise.
883 (gen_operands_ldd_std): Likewise.
884 * config/arc/arc.md: Add peephole2 rules for std/ldd.
886 2018-11-13 Eric Botcazou <ebotcazou@adacore.com>
888 * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
889 test on stack_usage_file.
890 (lang_dependent_init): Do not open the .su file if generating LTO.
892 2018-11-13 Jakub Jelinek <jakub@redhat.com>
894 PR rtl-optimization/87918
895 * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
896 simplify_gen_relational rather than simplify_gen_binary.
898 2018-11-13 Richard Biener <rguenther@suse.de>
900 * tree-ssanames.h (set_range_info): Use value_range_base.
901 (get_range_info): Likewise.
902 * tree-ssanames.c (set_range_info): Likewise.
903 (get_range_info): Likewise.
904 * tree-vrp.c (value_range_base::union_helper): Split
905 out common parts of value_range[_base]::union_.
906 (value_range_base::union_): Update.
907 (value_range::union_): Likewise.
908 (determine_value_range_1): Use value_range_base.
909 (determine_value_range): Likewise.
910 * tree-vrp.h (value_range_base::union_helper): Move ...
911 (value_range::union_helper): ... from here.
913 2018-11-13 Alan Modra <amodra@gmail.com>
915 * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
917 (rs6000_secondary_reload_gpr): Don't call find_replacement.
919 2018-11-13 Sandra Loosemore <sandra@codesourcery.com>
922 * doc/invoke.texi (Optimize Options): Clarify that the
923 l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
924 apply to data cache size.
926 2018-11-13 Alan Modra <amodra@gmail.com>
928 * config/rs6000/predicates.md (logical_const_operand),
929 (logical_operand): Correct comment.
930 * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
932 2018-11-13 Alan Modra <amodra@gmail.com>
934 * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
936 2018-11-13 Alan Modra <amodra@gmail.com>
938 * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
940 * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
941 covered by alternative.
942 (movcc_internal1): Ignore h for register preference.
943 (mov<mode>_hardfloat64): Likewise.
944 (mov<mode>_softfloat): Ignore c, l, h for register preference.
946 2018-11-12 Sandra Loosemore <sandra@codesourcery.com>
948 PR preprocessor/47823
949 * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
952 * doc/extend.texi (Pragmas): Note additional pragmas documented
955 2018-11-13 Kugan Vivekanandarajah <kuganv@linaro.org>
959 * tree-scalar-evolution.c (expression_expensive_p): Make
960 BUILTIN POPCOUNT as expensive when backend does not define it.
962 2018-11-12 Fredrik Noring <noring@nocrew.org>
964 * config.gcc: Update with-llsc defaults for MIPS r5900.
966 2018-11-12 Martin Liska <mliska@suse.cz>
968 PR gcov-profile/87442
969 * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
971 * doc/invoke.texi: Document them.
972 * tree-profile.c (parse_profile_filter): New.
973 (parse_profile_file_filtering): Likewise.
974 (release_profile_file_filtering): Likewise.
975 (include_source_file_for_profile): Likewise.
976 (tree_profiling): Filter source files based on the
979 2018-11-12 Martin Liska <mliska@suse.cz>
982 * doc/extend.texi: Add missing values for __builtin_cpu_is and
983 __builtin_cpu_supports for x86 target.
985 2018-11-12 Sandra Loosemore <sandra@codesourcery.com>
988 * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
990 (Arithmetic): Likewise for "lo_sum".
992 2018-11-12 Renlin Li <renlin.li@arm.com>
995 * dse.c (get_stored_val): Add check for compile-time constantness
998 2018-11-12 Sudakshina Das <sudi.das@arm.com>
1000 * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
1001 (ARMv8_5a): New fgroup.
1002 (armv8.5-a): New arch.
1003 (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
1005 * config/arm/arm-tables.opt: Regenerate.
1006 * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
1007 * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
1008 * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
1009 Add matching rules for -march=armv8.5-a and extensions.
1010 * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
1011 Add sb and predres to all armv8-a except armv8.5-a.
1013 2018-11-12 Richard Biener <rguenther@suse.de>
1015 * tree-vrp.h (value_range[_base]::set): Make public. Provide
1016 overload for single value.
1017 (value_range[_base]::set_nonnull): New.
1018 (value_range[_base]::set_null): Likewise.
1019 (value_range): Document bitmap copying behavior, mark
1020 copy constructor and assignment operator deleted.
1021 (value_range::move): New.
1022 (value_range::set_and_canonicalize): Default bitmap to zero.
1023 (set_value_range_to_nonnull): Remove.
1024 (set_value_range_to_null): Likewise.
1025 (set_value_range): Likewise.
1026 (set_value_range_to_value): Likewise.
1027 (extract_range_from_unary_expr): Work on value_range_base.
1028 (extract_range_from_binary_expr_1): Likewise. Rename to...
1029 (extract_range_from_binary_expr): ... this.
1030 * tree-vrp.c (value_range::update): Clear equiv bitmap
1032 (value_range::move): New, move equiv bitmap.
1033 (value_range_base::set_undefined): Avoid assignment.
1034 (value_range::set_undefined): Likewise.
1035 (value_range_base::set_varying): Likewise.
1036 (value_range::set_varying): Likewise.
1037 (set_value_range): Remove.
1038 (value_range_base::set): New overload for value.
1039 (value_range::set): Likewise.
1040 (set_value_range_to_nonnull): Remove.
1041 (value_range_base::set_nonnull): New.
1042 (value_range::set_nonnull): Likewise.
1043 (set_value_range_to_null): Remove.
1044 (value_range_base::set_null): New.
1045 (value_range::set_null): Likewise.
1046 (range_is_null): Work on value_range_base.
1047 (range_is_nonnull): Likewise.
1048 (ranges_from_anti_range): Likewise.
1049 (extract_range_into_wide_ints): Likewise.
1050 (extract_range_from_multiplicative_op): Likewise.
1051 (extract_range_from_binary_expr): Likewise. Update for API changes.
1052 (extract_range_from_unary_expr): Likewise. Remove OBJ_TYPE_REF
1054 (value_range::intersect_helper): Avoid copy and assignment.
1055 (value_range::union_helper): Likewise.
1056 (determine_value_range_1): Adjust.
1057 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
1058 Avoid assignment by using move.
1059 (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
1060 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
1062 * tree-ssanames.c (get_range_info): Likewise.
1063 * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
1064 * vr-values.c (vr_values::get_value_range): Adjust.
1065 (vr_values::update_value_range): Likewise.
1066 (symbolic_range_based_on_p): Work on value_range_base.
1067 (vr_values::extract_range_from_binary_expr): Use value_range_base.
1068 (vr_values::extract_range_from_unary_expr): Likewise.
1069 (vr_values::extract_range_from_cond_expr): Avoid assignment.
1070 (vr_values::extract_range_from_comparison): Adjust.
1071 (vr_values::check_for_binary_op_overflow): Use value_range_base.
1072 (vr_values::extract_range_basic): Adjust.
1073 (vr_values::adjust_range_with_scev): Likewise.
1074 (vr_values::vrp_visit_assignment_or_call): Likewise.
1075 (vr_values::get_vr_for_comparison): Change API to avoid
1076 assignment and copy construction.
1077 (vr_values::compare_name_with_value): Adjust accordingly.
1078 (vr_values::compare_names): Likewise.
1079 (vr_values::extract_range_from_phi_node): Avoid assignment and
1080 bogus in-place modify of equiv bitmap.
1081 (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
1082 * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
1083 for extract_range_from_unary_expr API change.
1084 * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
1086 2018-11-12 Eric Botcazou <ebotcazou@adacore.com>
1088 * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
1090 2018-11-12 Richard Biener <rguenther@suse.de>
1092 * tree-vrp.h (value_range_base::symbolic_p,
1093 value_range_base::constant_p, value_range_base::zero_p,
1094 value_range_base::singleton_p): Move from value_range.
1095 (value_range::dump): Add.
1096 * gimple-ssa-evrp-analyze.c
1097 (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
1098 * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
1099 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
1101 * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
1102 (value_range::constant_p): Likewise.
1103 (value_range::singleton_p): Likewise.
1104 (value_range_base::dump): Add.
1105 (set_value_range_to_undefined): Remove.
1106 (set_value_range_to_varying): Likewise.
1107 (range_int_cst_p): Take value_range_base argument.
1108 (range_int_cst_singleton_p): Likewise.
1109 (value_range_constant_singleton): Likewise.
1110 (vrp_set_zero_nonzero_bits): Likewise.
1111 (extract_range_from_multiplicative_op): Use set_varying.
1112 (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
1113 (extract_range_from_unary_expr): Likewise.
1114 (dump_value_range_base): Change to overload of dump_value_range.
1115 (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
1116 (vrp_prop::visit_stmt): Likewise.
1117 (value_range::intersect_helper): Likewise.
1118 (value_range::union_helper): Likewise.
1119 (determine_value_range_1): Likewise.
1121 2018-11-12 Richard Biener <rguenther@suse.de>
1123 * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
1124 (set_value_range_to_null): Likewise.
1125 * vr-values.c (vr_values::extract_range_from_comparison):
1126 Clear equiv for constant singleton ranges.
1128 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
1130 * config/i386/sse.md: Combine VFIXUPIMM* patterns
1131 (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1132 (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1133 (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
1134 (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1135 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1136 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
1138 2018-11-11 Sandra Loosemore <sandra@codesourcery.com>
1141 * doc/extend.texi (Common Type Attributes): For the align type
1142 attribute, copy language about decreasing alignment from the
1143 corresponding variable attribute.
1145 2018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com>
1147 * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
1150 2018-11-11 Sandra Loosemore <sandra@codesourcery.com>
1153 * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
1154 -frtti and -fno-rtti code.
1156 2018-11-11 Sandra Loosemore <sandra@codesourcery.com>
1159 * doc/extend.texi (Other Builtins): Document probability associated
1160 with __builtin_expect.
1162 2018-11-11 Uros Bizjak <ubizjak@gmail.com>
1165 * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
1166 instead of (TARGET_64BIT && ix86_abi == MS_ABI).
1167 * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
1168 * config/i386/cygming.h (STACK_BOUNDARY): Remove.
1170 2018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com>
1172 * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
1174 2018-11-11 Xianmiao Qu <xianmiao_qu@c-sky.com>
1176 * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
1178 2018-11-11 Richard Biener <rguenther@suse.de>
1180 * tree-vrp.h (class value_range_base): New base class for
1181 value_range containing all but the m_equiv member.
1182 (dump_value_range_base): Add.
1183 (range_includes_zero_p): Work on value_range_base.
1184 * tree-vrp.c (value_range_base::set): Split out base handling
1186 (value_range::set): this.
1187 (value_range::set_equiv): New.
1188 (value_range_base::value_range_base): New constructors.
1189 (value_range_base::check): Split out base handling from...
1190 (value_range::check): this.
1191 (value_range::equal_p): Refactor in terms of
1192 ignore_equivs_equal_p which is now member of the base.
1193 (value_range_base::set_undefined): New.
1194 (value_range_base::set_varying): Likewise.
1195 (value_range_base::dump):Split out base handling from...
1196 (value_range::dump): this.
1197 (value_range_base::set_and_canonicalize): Split out base handling
1199 (value_range::set_and_canonicalize): this.
1200 (value_range_base::union_): New.
1201 * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
1203 * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
1204 instead of value_range everywhere.
1205 (ipcp_vr_lattice::print): Use dump_value_range_base.
1206 (ipcp_vr_lattice::meet_with): Adjust.
1207 (ipcp_vr_lattice::meet_with_1): Likewise.
1208 (ipa_vr_operation_and_type_effects): Likewise.
1209 (propagate_vr_across_jump_function): Likewise.
1210 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
1211 (ipa_get_value_range): Likewise.
1212 (ipa_set_jfunc_vr): Likewise.
1213 (ipa_compute_jump_functions_for_edge): Likewise.
1215 2018-11-10 Sandra Loosemore <sandra@codesourcery.com>
1218 * doc/invoke.texi (Optimize Options): Add @opindex entries
1219 for the positive forms of -fno-xxx and -mno-xxx options
1220 that were lacking them.
1222 2018-11-10 Segher Boessenkool <segher@kernel.crashing.org>
1224 * combine.c (make_more_copies): Only make an intermediate copy if the
1225 dest of a move is a pseudo.
1227 2018-11-09 Maya Rashish <coypu@sdf.org>
1230 * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
1231 (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
1233 2018-11-09 Sandra Loosemore <sandra@codesourcery.com>
1237 * doc/invoke.texi (Optimize Options): Clarify default behavior
1238 for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
1240 2018-11-09 Bernd Edlinger <bernd.edlinger@hotmail.de>
1242 PR tree-optimization/87940
1243 * expr.c (string_constant): Don't strip NOPS in subexpressions.
1244 Fold PLUS_EXPR correctly.
1246 2018-11-09 Ilya Leoshkevich <iii@linux.ibm.com>
1249 * config/s390/s390.c (s390_safe_relative_long_p): New function.
1250 (annotate_constant_pool_refs): Skip insns which support
1251 relative addressing.
1252 (annotate_constant_pool_refs_1): New helper function.
1253 (find_constant_pool_ref): Skip insns which support relative
1255 (find_constant_pool_ref_1): New helper function.
1256 (replace_constant_pool_ref): Skip insns which support
1257 relative addressing.
1258 (replace_constant_pool_ref_1): New helper function.
1259 (s390_mainpool_start): Adapt to the new signature.
1260 (s390_mainpool_finish): Likewise.
1261 (s390_chunkify_start): Likewise.
1262 (s390_chunkify_finish): Likewise.
1263 (pass_s390_early_mach::execute): Likewise.
1264 (s390_prologue_plus_offset): Likewise.
1265 (s390_emit_prologue): Likewise.
1266 (s390_emit_epilogue): Likewise.
1268 2018-11-09 Jakub Jelinek <jakub@redhat.com>
1270 * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
1271 but unsupported lastprivate with conditional modifier.
1273 2018-11-09 Jeff Law <law@redhat.com>
1275 * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
1276 unused argument better. Add gcc_unreachable to silence warning.
1278 2018-11-09 Martin Sebor <msebor@redhat.com>
1281 * attribs.c (has_attribute): New helper function.
1282 (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
1283 * attribs.h (decls_mismatched_attributes): Declare.
1284 * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
1285 (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
1286 * common.opt (-Wattribute-alias): Take an argument.
1287 (-Wno-attribute-alias): New option.
1288 * doc/extend.texi (Common Function Attributes): Document copy.
1289 (Common Variable Attributes): Same.
1290 * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
1291 (-Wattribute-alias): Document new option argument.
1293 2018-11-09 Richard Earnshaw <rearnsha@arm.com>
1295 * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
1296 matching criteria. Remove unused array initializer.
1298 2018-11-09 Bill Schmidt <wschmidt@linux.ibm.com>
1299 Jinsong Ji <jji@us.ibm.com>
1301 * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
1302 constraints by introducing a new temporary.
1303 (_mm_cvtss_si64): Likewise.
1305 2018-11-09 Martin Liska <mliska@suse.cz>
1307 * common.opt: Add -fipa-stack-alignment flag.
1308 * doc/invoke.texi: Document it.
1309 * final.c (rest_of_clean_state): Guard stack
1310 shrinking with flag.
1312 2018-11-09 Martin Liska <mliska@suse.cz>
1314 * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
1316 (ipa_discover_variable_flags): ... this.
1317 * common.opt: Come up with new flag -fipa-reference-addressable.
1318 * doc/invoke.texi: Document it.
1319 * ipa-reference.c (propagate): Call the renamed fn.
1320 * ipa-visibility.c (whole_program_function_and_variable_visibility):
1322 * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
1324 (ipa_discover_variable_flags): ... this. Discover
1325 non-addressable variables only with the newly added flag.
1326 * opts.c: Enable the newly added flag with -O1 and higher
1329 2018-11-09 David Malcolm <dmalcolm@redhat.com>
1331 * json.cc (selftest::test_writing_literals): Fix comment.
1333 2018-11-09 Martin Liska <mliska@suse.cz>
1335 * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
1336 string to a stack buffer.
1337 (aarch64_parse_cpu): Likewise.
1338 (aarch64_parse_tune): Likewise.
1340 2018-11-09 Richard Biener <rguenther@suse.de>
1342 PR tree-optimization/87953
1343 * tree-vect-loop.c (vectorizable_reduction): For analysis
1344 always pass ops[0] to vectorizable_condition.
1346 2018-11-09 Stafford Horne <shorne@gmail.com>
1347 Richard Henderson <rth@twiddle.net>
1348 Joel Sherrill <joel@rtems.org>
1350 * common/config/or1k/or1k-common.c: New file.
1351 * config/or1k/*: New.
1352 * config.gcc (or1k*-*-*): New.
1353 * configure.ac (or1k*-*-*): New test for openrisc tls.
1354 * configure: Regenerated.
1355 * doc/install.texi: Document OpenRISC triplets.
1356 * doc/invoke.texi: Document OpenRISC arguments.
1357 * doc/md.texi: Document OpenRISC.
1359 2018-11-09 Richard Earnshaw <rearnsha@arm.com>
1361 * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
1362 (arm7tdmi-s): Delete CPU.
1363 (arm710t): Add aliases for arm720t and arm740t.
1364 (arm720t, arm740t): Delete CPUs.
1365 (arm920t): Add aliases for arm920, arm922t and arm940t.
1366 (arm920, arm922t, arm940t): Delete CPUs.
1367 (arm10tdmi): Add alias for arm1020t.
1368 (arm1020t): Delete CPU.
1369 (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
1370 (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
1371 (arm10e): Add aliases for arm1020e and arm1022e.
1372 (arm1020e, arm1022e): Delete CPU.
1373 * config/arm/arm.md (generic_sched): Remove entries that are now
1375 (generic_vfp): Likewise.
1376 * config/arm/arm1020e.md: Simplify tuning selection based on alias
1378 * config/arm/arm-tune.md: Regenerated.
1379 * config/arm/arm-tables.opt: Regenerated.
1381 2018-11-09 Richard Biener <rguenther@suse.de>
1383 PR tree-optimization/87621
1384 * tree-vect-loop.c (vectorizable_reduction): Handle reduction
1385 op with only phi inputs.
1386 * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
1387 (ch_base::copy_headers): Run CSE on copied loop headers.
1388 (pass_ch_vect::process_loop_p): Simplify.
1390 2018-11-09 Alexandre Oliva <oliva@adacore.com>
1392 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
1393 for non-w64 x86_64 biarch.
1395 2018-11-09 Alexandre Oliva <aoliva@redhat.com>
1397 PR rtl-optimization/86438
1398 * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
1399 of in_b for the compare if in_b is SET_DEST.
1402 * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
1403 non-toplevel UNSPEC.
1405 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
1407 * tree-vrp.c (value_range::check): Do not access internals
1409 (value_range::singleton_p): Same.
1410 (value_range::type): Same.
1411 (vrp_finalize): Use value_range API.
1413 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
1415 * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
1417 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
1419 * vr-values.c (vr_values::get_value_range): Use value_range API
1420 instead of piecing together ranges.
1421 (vr_values::update_value_range): Same.
1423 2018-11-08 Aldy Hernandez <aldyh@redhat.com>
1425 * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
1426 of performing ad-hoc calculations.
1427 * tree-ssanames.c (set_range_info): New overloaded function
1428 accepting value_range &.
1429 (get_range_info): Same.
1430 * tree-ssanames.h (set_range_info_raw): Remove.
1431 (set_range_info): New prototype.
1432 (get_range_info): Same.
1433 * tree-vrp.h (value_range::null_p): Rename to zero_p.
1434 * tree-vrp.c (value_range::null_p): Same.
1436 2018-11-09 Jan Hubicka <jh@suse.cz>
1438 * tree.c (fld_type_variant_equal_p): Test user align flag.
1439 (flt_type_variant): Copy user align flag.
1440 (fld_incomplete_type_of): Clear it.
1442 2018-11-09 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
1444 * config/arm/neon.md (div<mode>3): New pattern.
1446 2018-11-08 Andi Kleen <ak@linux.intel.com>
1448 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
1449 (OPTION_MASK_ISA_PTWRITE_UNSET): New.
1450 (ix86_handle_option): Handle OPT_mptwrite.
1451 * config/i386/cpuid.h (bit_PTWRITE): Add.
1452 * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
1453 * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
1454 * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
1455 * config/i386/i386.c (ix86_target_string): Handle ptwrite.
1456 (ix86_option_override_internal): Handle PTA_PTWRITE.
1457 (ix86_valid_target_attribute_inner_p): Define ptwrite.
1458 (def_builtin2): Force UINT64 to be 64bit only.
1459 * config/i386/i386.h (TARGET_PTWRITE): Add.
1460 (TARGET_PTWRITE_P): Add.
1462 * config/i386/i386.md: Define ptwrite.
1463 * config/i386/i386.opt: Add -mptwrite.
1464 * config/i386/immintrin.h (_ptwrite64): Add.
1466 * doc/extend.texi: Document __builtin_ia32_ptwrite*.
1467 * doc/invoke.texi: Document -mptwrite.
1469 2018-11-08 Peter Bergner <bergner@linux.ibm.com>
1471 PR rtl-optimization/87600
1472 * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
1473 * lra-constraints.c (process_alt_operands): Skip illegal hard
1474 register usage. Prefer reloading non hard register operands.
1476 2018-11-08 Sandra Loosemore <sandra@codesourcery.com>
1479 * doc/invoke.texi (Optimize Options): Clarify default behavior
1480 for -fno-sched-interblock and -fno-sched-spec.
1482 2018-11-08 Roman Geissler <roman.geissler@amadeus.com>
1484 * collect2.c (linker_select): Add USE_LLD_LD.
1485 (ld_suffixes): Add ld.lld.
1486 (main): Handle -fuse-ld=lld.
1487 * common.opt (-fuse-ld=lld): New option.
1488 * doc/invoke.texi (-fuse-ld=lld): Document.
1489 * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
1491 2018-11-08 Paul Koning <ni1d@arrl.net>
1493 * config/pdp11/constraints.md: Add "Z" series constraints for use
1494 with pre-dec and post-inc addressing.
1495 * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
1496 (pdp11_expand_operands): Add int argument (word count).
1497 (pdp11_sp_frame_offset): Delete.
1498 (pdp11_cmp_length): New function.
1499 (pushpop_regeq): New function.
1500 * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
1502 (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
1504 (pdp11_initial_elimination_offset): Ditto.
1505 (pdp11_expand_operands): Add word count argument. Bugfixes.
1506 (output_move_multiple): Change how pointer adjustment is done.
1507 (pdp11_gen_int_label): Correct format.
1508 (output_ascii): Ditto.
1509 (pdp11_asm_output_var): Add code for DEC assembler case.
1510 (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
1512 (legitimate_const_double_p): Ditto.
1513 (pdp11_register_move_cost): Adjust for new register classes.
1514 (pdp11_regno_reg_class): Ditto.
1515 (expand_block_move): Delete.
1516 (pushpop_regeq): New function.
1517 (pdp11_legitimate_address_p): Bugfix in check for constant
1519 (pdp11_sp_frame_offset): Delete.
1520 (pdp11_reg_save_size): New helper function for new frame layout.
1521 (output_addr_const_pdp11): Remove CONST_DOUBLE case.
1522 (pdp11_expand_shift): Bugfix in check for constant shift count.
1523 (pdp11_shift_length): Ditto.
1524 (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
1525 (pdp11_cmp_length): New function.
1526 * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
1527 some compile options.
1528 (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
1529 (CALL_USED_REGISTERS): Ditto.
1530 (ELIMINABLE_REGS): Ditto.
1531 (REGISTER_NAMES): Ditto.
1532 (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
1534 (REG_CLASS_NAMES): Ditto.
1535 (REG_CLASS_CONTENTS): Ditto. Also remove
1536 HARD_FRAME_POINTER_REGNUM.
1537 (CPU_REG_CLASS): New macro.
1538 (CLASS_MAX_NREGS): Adjust for new register classes.
1539 (FUNCTION_PROFILER): Make no-op.
1540 (may_call_alloca): Remove unused declaration.
1541 (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
1542 (ASM_OUTPUT_SKIP): Fix format.
1543 * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
1544 (HARD_FRAME_POINTER_REGNUM): Remove.
1546 (*rts): Rename. Remove epilogue related checks.
1547 (cmpsi, cmpdi): New insn.
1548 (cbranch<mode>4): Change to apply to SI and DI modes as well.
1549 (mov<mode>): Change constraints to enforce that push/pop
1550 destination cannot use the same register as source.
1551 (*mov<mode><cc_cc>): Ditto.
1552 (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
1553 at assembly output rather than as RTL expander.
1554 (zero_extendqihi2): Bugfix in check for same registers.
1555 (adddi3_nocc): Bugfix in check for constant operand.
1556 (addsi3_nocc): Ditto.
1557 (subdi3_nocc): Ditto.
1558 (subsi3_nocc): Ditto.
1559 (negdi2_nocc): Copy input to pdp11_expand_operands.
1560 (negsi2_nocc): Ditto.
1561 (bswap2_nocc): Ditto.
1562 * config/pdp11/pdp11.opt (mlra): Fix documentation.
1563 * config/pdp11/t-pdp11: Use -Os.
1565 2018-11-08 Richard Earnshaw <rearnsha@arm.com>
1567 * config/arm/parsecpu.awk (/alias/): New parsing rule.
1568 (/begin cpu/): Check that the cpu name hasn't been previously defined.
1569 (gen_comm_data): Print out CPU alias tables.
1570 (check_cpu): Match aliases when checking the CPU name.
1571 * config/arm/arm-protos.h (cpu_alias): New structure.
1572 (cpu_option): Add entry for aliases.
1573 * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
1574 strongarm1100 and strongarm1110.
1575 (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
1576 (config/arm/arm-generic.md): Remove redundant references to
1577 strongarm110, strongarm1100 and strongarm1110.
1578 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1579 Scan aliases for additional hints.
1580 (arm_parse_cpu_option_name): Also match a cpu name against the list
1582 * config/arm/arm-tables.opt: Regenerated.
1583 * config/arm/arm-tune.md: Regenerated.
1585 2018-11-08 Jakub Jelinek <jakub@redhat.com>
1587 * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
1588 BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
1589 BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1590 BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
1591 BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1592 BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
1593 * gengtype.c (open_base_files): Add omp-general.h.
1594 * gimple.c (gimple_build_omp_critical):
1595 (gimple_build_omp_taskgroup): Add CLAUSES argument. Call
1596 gimple_omp_taskgroup_set_clauses.
1597 (gimple_build_omp_atomic_load): Add mo argument, call
1598 gimple_omp_atomic_set_memory_order.
1599 (gimple_build_omp_atomic_store): Likewise.
1600 (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
1601 * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
1603 (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
1604 of GSS_OMP_SINGLE_LAYOUT, adjust comments.
1605 * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
1606 and GF_OMP_ATOMIC_MEMORY_ORDER. Remove GF_OMP_ATOMIC_SEQ_CST, use
1607 different value for GF_OMP_ATOMIC_NEED_VALUE.
1608 (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
1610 (struct gimple_statement_omp_single_layout): And remove here.
1611 (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
1612 than gimple_statement_omp_single_layout.
1613 (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
1615 (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
1616 (gimple_omp_subcode): Formatting fix.
1617 (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
1618 gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
1619 gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
1620 gimple_omp_teams_host, gimple_omp_teams_set_host,
1621 gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
1622 gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
1623 gimple_omp_taskgroup_set_clauses): New inline functions.
1624 (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
1625 (gimple_build_omp_atomic_store): Likewise.
1626 (gimple_omp_atomic_seq_cst_p): Remove.
1627 (gimple_omp_atomic_memory_order): New function.
1628 (gimple_omp_atomic_set_seq_cst): Remove.
1629 (gimple_omp_atomic_set_memory_order): New function.
1630 (gimple_build_omp_taskgroup): Add clauses argument.
1631 * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
1632 (dump_gimple_omp_task): Print taskwait with depend clauses.
1633 (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
1634 dump_omp_atomic_memory_order.
1635 (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
1636 * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
1637 GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
1638 (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
1639 renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
1640 ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
1641 ORT_UNTIED_TASKLOOP enumerators.
1642 (enum gimplify_defaultmap_kind): New.
1643 (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
1644 target_map_pointers_as_0len_arrays members, add defaultmap.
1645 (new_omp_context): Initialize defaultmap member.
1646 (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1647 (maybe_fold_stmt): Don't fold even in host teams regions.
1648 (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
1649 ORT_WORKSHARE. Test ctx->defaultmap[GDMK_SCALAR] instead of
1650 ctx->omp_firstprivatize_variable.
1651 (omp_add_variable): Don't add private/firstprivate for VLAs in
1653 (omp_default_clause): Print "taskloop" rather than "task" if
1655 (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1656 Handle new defaultmap clause kinds.
1657 (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE. Allow simd
1658 iterator to be lastprivate or private. Fix up diagnostics if linear
1659 is used on collapse>1 simd iterator.
1660 (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1661 (gimplify_omp_depend): New function.
1662 (gimplify_scan_omp_clauses): Add shared clause on parallel for
1663 combined parallel master taskloop{, simd} if taskloop has
1664 firstprivate, lastprivate or reduction clause. Handle
1665 OMP_CLAUSE_REDUCTION_TASK diagnostics. Adjust tests for
1666 ORT_COMBINED_TEAMS. Gimplify depend clauses with iterators. Handle
1667 cancel and simd OMP_CLAUSE_IF_MODIFIERs. Handle
1668 OMP_CLAUSE_NONTEMPORAL. Handle new defaultmap clause kinds. Handle
1669 OMP_CLAUSE_{TASK,IN}_REDUCTION. Diagnose invalid conditional
1671 (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL. Handle
1672 GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.
1673 (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL. Handle
1674 OMP_CLAUSE_{TASK,IN}_REDUCTION.
1675 (gimplify_omp_task): Handle taskwait with depend clauses.
1676 (gimplify_omp_for): Add shared clause on parallel for combined
1677 parallel master taskloop{, simd} if taskloop has firstprivate,
1678 lastprivate or reduction clause. Use ORT_TASKLOOP or
1679 ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK. Adjust
1680 tests for ORT_COMBINED_TEAMS. Handle C++ range for loops with
1681 NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS. Firstprivatize
1682 __for_end and __for_range temporaries on OMP_PARALLEL for
1683 distribute parallel for{, simd}. Move OMP_CLAUSE_REDUCTION
1684 and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
1685 sandwiched in between two taskloops.
1686 (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
1687 instead of ctx->omp_firstprivatize_variable.
1688 (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
1689 ORT_COMBINED_HOST_TEAMS if not inside of target construct. If
1690 host teams, use gimplify_and_return_first etc. for body like
1691 for target or target data constructs, and at the end call
1692 gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
1693 (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
1694 of OMP_ATOMIC_SEQ_CST, pass it as new argument to
1695 gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
1696 gimple_omp_atomic_set_seq_cst calls.
1697 (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
1698 case, handle taskgroup clauses.
1699 * lto-streamer-out.c (hash_tree): Handle
1700 OMP_CLAUSE_{TASK,IN}_REDUCTION.
1701 * Makefile.in (GTFILES): Add omp-general.h.
1702 * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
1703 BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
1704 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
1705 BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
1706 BUILT_IN_GOMP_LOOP_DOACROSS_START,
1707 BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
1708 BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1709 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
1710 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
1711 BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
1712 BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
1713 BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
1714 BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1715 BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
1716 BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
1717 BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
1718 BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
1719 BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
1720 BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
1721 BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
1722 * omp-expand.c (workshare_safe_to_combine_p): Return false for
1723 non-worksharing loops.
1724 (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
1725 (determine_parallel_type): Don't combine parallel with worksharing
1726 which has _reductemp_ clause.
1727 (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
1728 GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
1729 if there is nonmonotonic modifier or if there is no modifier and no
1730 ordered clause. For dynamic and guided schedule without monotonic
1731 and nonmonotonic modifier, default to nonmonotonic.
1732 (expand_omp_for): Likewise. Adjust expand_omp_for_generic caller, use
1733 GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
1735 (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
1736 there are any reduction clauses.
1737 (expand_taskwait_call): New function.
1738 (expand_teams_call): New function.
1739 (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
1740 expand_teams_call for it. Formatting fix. Handle taskwait with
1742 (expand_omp_for_generic): Add SCHED_ARG argument. Handle expansion
1743 of worksharing loops with task reductions.
1744 (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
1745 expansion of worksharing loops with task reductions.
1746 (expand_omp_sections): Handle expansion of sections with task
1748 (expand_omp_synch): For host teams call expand_omp_taskreg.
1749 (omp_memory_order_to_memmodel): New function.
1750 (expand_omp_atomic_load, expand_omp_atomic_store,
1751 expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
1752 instead of gimple_omp_atomic_seq_cst_p.
1753 (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
1754 depend clauses as a standalone directive.
1755 * omp-general.c (enum omp_requires): New variable.
1756 (omp_extract_for_data): Initialize have_reductemp member. Allow
1757 NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
1758 GT_EXPR loops depending on incr sign. Formatting fixes.
1759 * omp-general.h (struct omp_for_data): Add have_reductemp member.
1760 (enum omp_requires): New enum.
1761 (omp_requires_mask): Declare.
1762 * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
1764 * omp-low.c (struct omp_context): Add task_reductions and
1765 task_reduction_map fields.
1766 (is_host_teams_ctx): New function.
1767 (is_taskreg_ctx): Return true also if is_host_teams_ctx.
1768 (use_pointer_for_field): Use is_global_var instead of
1769 TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
1771 (build_outer_var_ref): Ignore taskgroup outer contexts.
1772 (delete_omp_context): Release task_reductions and task_reduction_map.
1773 (scan_sharing_clauses): Don't add any fields for reduction clause on
1774 taskloop. Handle OMP_CLAUSE__REDUCTEMP_. Handle
1775 OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
1776 modifier. Don't ignore shared clauses in is_host_teams_ctx contexts.
1777 Handle OMP_CLAUSE_NONTEMPORAL.
1778 (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
1780 (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
1781 clauses with task modifier.
1782 (scan_omp_task): Handle taskwait with depend clauses.
1783 (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
1784 first. Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
1785 Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
1786 (scan_omp_for): Fix comment formatting.
1787 (scan_omp_teams): Handle host teams constructs.
1788 (check_omp_nesting_restrictions): Allow teams with no outer
1789 OpenMP context. Adjust diagnostics for teams strictly nested into
1790 some explicit OpenMP construct other than target. Allow OpenMP atomics
1791 inside of simd regions.
1792 (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
1793 (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
1794 taskreg_nesting_level while scanning host teams construct.
1795 (task_reduction_read): New function.
1796 (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
1797 construct. Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
1798 clauses. Handle OMP_CLAUSE_REDUCTION with task modifier. Remove
1799 second argument create_tmp_var if it is NULL. Don't ignore shared
1800 clauses in is_host_teams_ctx contexts. Handle
1801 OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
1803 (lower_reduction_clauses): Ignore reduction clauses with task
1804 modifier. Remove second argument create_tmp_var if it is NULL.
1805 Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
1806 (lower_send_clauses): Ignore reduction clauses with task modifier.
1807 Handle OMP_CLAUSE__REDUCTEMP_. Don't send anything for
1808 OMP_CLAUSE_REDUCTION on taskloop. Handle OMP_CLAUSE_IN_REDUCTION.
1809 (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
1810 rely that it is the last stmt in body so far. Ignore outer taskgroup
1812 (omp_task_reductions_find_first, omp_task_reduction_iterate,
1813 lower_omp_task_reductions): New functions.
1814 (lower_omp_sections): Handle reduction clauses with taskgroup
1815 modifiers. Adjust maybe_add_implicit_barrier_cancel caller.
1816 (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
1817 (lower_omp_for): Likewise. Handle reduction clauses with taskgroup
1819 (lower_omp_taskgroup): Handle taskgroup reductions.
1820 (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
1821 Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
1822 (lower_depend_clauses): If there are any
1823 OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1824 depend clauses, use a new array format. If OMP_CLAUSE_DEPEND_LAST is
1825 seen, assume lowering is done already and return early. Set kind
1826 on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
1827 (lower_omp_taskreg): Handle reduction clauses with task modifier on
1828 parallel construct. Handle reduction clause on taskloop construct.
1829 Handle taskwait with depend clauses.
1830 (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
1831 for host teams constructs.
1832 * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
1833 nontemporal and _reductemp_ clause entries.
1834 (omp_clause_code_name): Likewise.
1835 (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
1836 OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
1837 * tree-core.h (enum omp_clause_code): Add
1838 OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
1839 (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
1840 (struct tree_base): Add omp_atomic_memory_order field into union.
1841 Remove OMP_ATOMIC_SEQ_CST comment.
1842 (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1843 and OMP_CLAUSE_DEPEND_DEPOBJ.
1844 (struct tree_omp_clause): Add subcode.defaultmap_kind.
1845 * tree.def (OMP_TASKGROUP): Add another operand, move next to other
1846 OpenMP constructs with body and clauses operands.
1847 * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
1848 (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
1849 (OMP_TASKGROUP_CLAUSES): Define.
1850 (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
1851 OMP_CLAUSE__LOOPTEMP_.
1852 (OMP_ATOMIC_SEQ_CST): Remove.
1853 (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
1854 OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
1855 (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
1856 OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
1857 OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
1858 OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
1859 OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
1860 (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
1861 OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
1862 OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
1864 * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
1865 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1866 OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
1867 (convert_local_omp_clauses): Likewise. Remove useless test.
1868 * tree-parloops.c (create_call_for_reduction_1): Pass
1869 OMP_MEMORY_ORDER_RELAXED as new argument to
1870 dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
1871 * tree-pretty-print.c (dump_omp_iterators): New function.
1872 (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
1873 OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION. Print
1874 reduction modifiers. Handle OMP_CLAUSE_DEPEND_DEPOBJ and
1875 OMP_CLAUSE_DEPEND_MUTEXINOUTSET. Print iterators in depend clauses.
1876 Print __internal__ for OMP_CLAUSE_DEPEND_LAST. Handle cancel and
1877 simd OMP_CLAUSE_IF_MODIFIERs. Handle new kinds of
1878 OMP_CLAUSE_DEFAULTMAP. Print conditional: for
1879 OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
1880 (dump_omp_atomic_memory_order): New function.
1881 (dump_generic_node): Use it. Print taskgroup clauses. Print
1882 taskwait with depend clauses.
1883 * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
1884 * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
1885 Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
1886 * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
1887 write_ts_omp_clause_tree_pointers): Likewise.
1889 2018-11-08 David Malcolm <dmalcolm@redhat.com>
1892 * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
1893 "-missed", "-note", and "-all" sub-options.
1894 * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
1896 (can_early_inline_edge_p): Likewise.
1897 (want_early_inline_function_p): Likewise.
1898 (want_inline_self_recursive_call_p): Likewise.
1899 (recursive_inlining): Likewise.
1900 (inline_small_functions): Likewise.
1901 (flatten_function): Likewise.
1902 (ipa_inline): Likewise.
1903 (inline_always_inline_functions): Likewise.
1904 (early_inline_small_functions): Likewise.
1905 (early_inliner): Likewise.
1906 * tree-inline.c (expand_call_inline): Likewise.
1908 2018-11-08 David Malcolm <dmalcolm@redhat.com>
1910 * pretty-print.c (pp_format): Handle %f.
1911 (selftest::test_pp_format): Add test of %f.
1912 * pretty-print.h (pp_double): New macro.
1914 2018-11-08 David Malcolm <dmalcolm@redhat.com>
1916 * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
1917 * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
1918 (dump_pretty_printer::decode_format): Implement "%C" for
1920 (selftest::test_capture_of_dump_calls): Rename "where" to
1921 "stmt_loc". Convert test_decl to a function decl and set its
1922 location. Add a symbol_table_test RAII instance and a
1923 cgraph_node, using it to test "%C" and dump_symtab_node.
1925 2018-11-08 Eric Botcazou <ebotcazou@adacore.com>
1928 * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
1930 2018-11-08 David Malcolm <dmalcolm@redhat.com>
1932 * cgraph.c: Include "selftest.h".
1933 (saved_symtab): New variable.
1934 (selftest::symbol_table_test::symbol_table_test): New ctor.
1935 (selftest::symbol_table_test::~symbol_table_test): New dtor.
1936 (selftest::test_symbol_table_test): New test.
1937 (selftest::cgraph_c_tests): New.
1938 * cgraph.h (saved_symtab): New decl.
1939 (selftest::symbol_table_test): New class.
1940 * selftest-run-tests.c (selftest::run_tests): Call
1941 selftest::cgraph_c_tests.
1942 * selftest.h (selftest::cgraph_c_tests): New decl.
1944 2018-11-08 Richard Biener <rguenther@suse.de>
1946 * tree-data-ref.h (lambda_int): New typedef.
1947 (lambda_vector_gcd): Adjust.
1948 (lambda_vector_new): Likewise.
1949 (lambda_matrix_new): Likewise.
1950 * tree-data-ref.c (print_lambda_vector): Adjust.
1952 2018-11-08 Richard Biener <rguenther@suse.de>
1954 PR tree-optimization/87929
1955 * tree-complex.c (expand_complex_comparison): Clean EH.
1957 2018-11-08 Martin Liska <mliska@suse.cz>
1959 * doc/extend.texi: Reword.
1960 * predict.c (expr_expected_value_1): Likewise.
1962 2018-11-08 Richard Biener <rguenther@suse.de>
1964 PR tree-optimization/87913
1965 * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
1966 of extreme values to ordered comparisons.
1968 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
1971 * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
1973 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
1975 * doc/invoke.texi: Remove leading dash from @opindex entries
1976 throughout the file.
1978 2018-11-07 Sandra Loosemore <sandra@codesourcery.com>
1981 * doc/invoke.texi (Option Summary): Add -e and --entry.
1982 (Link Options): Likewise.
1984 2018-11-07 Nathan Sidwell <nathan@acm.org>
1987 * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
1988 --disable-checking bootstrap.
1990 2018-11-07 Hafiz Abid Qadeer <abidh@codesourcery.com>
1992 * configure: Regenerated.
1994 2018-11-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1997 * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
1998 to the mode of the widest field iff the widest field has mode class
1999 MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
2002 2018-11-07 Nikolai Merinov <n.merinov@inango-systems.com>
2004 * common.opt: Add -Wattribute-warning.
2005 * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
2006 * expr.c (expand_expr_real_1): Add new attribute to warning_at
2007 call to allow user configure behavior of "warning" attribute.
2009 2018-11-07 Segher Boessenkool <segher@kernel.crashing.org>
2011 * target.def: Put @: after every vs., e.g., and i.e. where it is
2012 followed by whitespace.
2013 * doc/extend.texi: Ditto.
2014 * doc/fragments.texi: Ditto.
2015 * doc/gimple.texi: Ditto.
2016 * doc/implement-c.texi: Ditto.
2017 * doc/install.texi: Ditto.
2018 * doc/invoke.texi: Ditto.
2019 * doc/md.texi: Ditto.
2020 * doc/plugins.texi: Ditto.
2021 * doc/rtl.texi: Ditto.
2022 * doc/sourcebuild.texi: Ditto.
2023 * doc/tm.texi.in: Ditto.
2024 * doc/ux.texi: Ditto.
2025 * doc/tm.texi: Regenerate.
2027 2018-11-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2029 * config/arm/arm-cpus.in (ares): New entry.
2030 * config/arm/arm-tables.opt: Regenerate.
2031 * config/arm/arm-tune.md: Likewise.
2032 * doc/invoke.texi (ARM Options): Document ares.
2034 2018-11-07 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2036 * config/aarch64/aarch64-cores.def (ares): Define.
2037 * config/aarch64/aarch64-tune.md: Regenerate.
2038 * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
2040 2018-11-07 Jan Hubicka <jh@suse.cz>
2042 * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
2043 (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
2044 functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
2047 2018-11-07 Richard Biener <rguenther@suse.de>
2049 PR tree-optimization/87914
2050 * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
2052 (vectorizable_reduction): Handle shifts and rotates by dispatching
2053 to vectorizable_shift.
2054 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
2055 in-loop uses of vect_nested_cycle defs. Merge cycle and internal
2057 (vectorizable_shift): Export and handle being called as
2059 (vect_analyze_stmt): Call vectorizable_shift after
2060 vectorizable_reduction.
2061 * tree-vectorizer.h (vectorizable_shift): Declare.
2063 2018-11-07 Jan Hubicka <jh@suse.cz>
2065 * ipa-devirt.c (odr_types_equivalent_p): Expect constants
2066 than const decls in TREE_VALUE of enum.
2067 (dump_type_inheritance_graph): Improve duplicate dumping.
2068 (free_enum_values): New.
2069 (build_type_inheritance_graph): Use it.
2070 * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
2071 which are not main variants or not ODR types.
2072 (verify_type_variant): Expect variants to have no TYPE_VALUES.
2074 2018-11-07 Richard Biener <rguenther@suse.de>
2076 * ipa-inline.c (want_inline_small_function_p): Compute
2077 big_speedup_p lazily and last.
2079 2018-11-07 Jan Hubicka <jh@suse.cz>
2081 * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
2082 building incomplete variant of complete type.
2083 (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
2084 variant of complete type.
2086 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
2088 * config/mips/mips.c: Fix typo in documentation of
2089 mips_loongson_ext2_prefetch_cookie.
2090 (mips_option_override): fix brain twister logical.
2091 * config/mips/mips.h: Fix typo in documentation of
2092 ISA_HAS_CTZ_CTO and define pattern.
2093 * config/mips/mips.md (prefetch): Hoist EXT2 above
2095 (prefetch_indexed): Hoist EXT2 above the EXT block.
2097 2018-11-07 Jan Hubicka <jh@suse.cz>
2099 * tree.c (free_lang_data_in_type): Add fld parameter; simplify
2100 return and parameter types of function and method types.
2101 (free_lang_data_in_cgraph): Update.
2103 2018-11-07 Martin Liska <mliska@suse.cz>
2105 PR rtl-optimization/87868
2106 * postreload-gcse.c (eliminate_partially_redundant_load): Set
2107 threshold to max_count if we would overflow.
2108 * profile-count.h: Make max_count a public constant.
2110 2018-11-07 Martin Liska <mliska@suse.cz>
2112 * mem-stats.h: Fix GNU coding style.
2114 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
2116 * config/mips/gs264e.md: New.
2117 * config/mips/mips-cpus.def: Define gs264e.
2118 * config/mips/mips-tables.opt: Regenerate.
2119 * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
2121 (mips_issue_rate): Add support for gs264e.
2122 (mips_multipass_dfa_lookahead): Likewise.
2123 * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
2124 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
2125 (MIPS_ASE_MSA_SPEC): New.
2126 (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
2127 (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
2128 (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
2129 * config/mips/mips.md: Include gs264e.md.
2130 (processor): Add gs264e.
2131 * config/mips/mips.opt (MSA): Use Mask instead of Var.
2132 * doc/invoke.texi: Add gs264e to supported architectures.
2134 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
2136 * config/mips/gs464e.md: New.
2137 * config/mips/mips-cpus.def: Define gs464e.
2138 * config/mips/mips-tables.opt: Regenerate.
2139 * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
2141 (mips_issue_rate): Add support for gs464e.
2142 (mips_multipass_dfa_lookahead): Likewise.
2143 (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
2144 * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
2145 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
2146 (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
2147 (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
2148 * config/mips/mips.md: Include gs464e.md.
2149 (processor): Add gs464e.
2150 * doc/invoke.texi: Add gs464e to supported architectures.
2152 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
2154 * config/mips/loongson3a.md: Rename to ...
2155 * config/mips/gs464.md: ... here.
2156 * config/mips/mips-cpus.def: Define gs464; Add loongson3a
2157 as an alias of gs464 processor.
2158 * config/mips/mips-tables.opt: Regenerate.
2159 * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
2160 instead of PROCESSOR_LOONGSON_3A.
2161 (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
2163 (mips_option_override): Enable MMI and EXT for gs464.
2164 * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
2165 Rename TUNE_LOONGSON_3A to TUNE_GS464.
2166 (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
2167 (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
2168 ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
2170 * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
2171 (processor): Add gs464;
2172 * doc/invoke.texi: Add gs464 to supported architectures.
2174 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
2176 * config/mips/mips-protos.h
2177 (mips_loongson_ext2_prefetch_cookie): New prototype.
2178 * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
2179 (mips_option_override): Enable TARGET_LOONGSON_EXT when
2180 TARGET_LOONGSON_EXT2 is true.
2181 * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
2182 __mips_loongson_ext2, __mips_loongson_ext_rev=2.
2183 (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
2184 (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
2185 TARGET_LOONGSON_EXT2.
2186 (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
2187 (define_insn "ctz<mode>2"): New insn pattern.
2188 (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
2189 (define_insn "prefetch_indexed_<mode>"): Include
2190 TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
2191 * config/mips/mips.opt (-mloongson-ext2): Add option.
2192 * gcc/doc/invoke.texi (-mloongson-ext2): Document.
2194 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
2196 * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
2197 __mips_loongson_ext.
2198 (MIPS_ASE_LOONGSON_EXT_SPEC): New.
2199 (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
2201 (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
2202 * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
2203 <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
2204 instead of TARGET_LOONGSON_3A.
2205 * config/mips/mips.opt (-mloongson-ext): Add option.
2206 * gcc/doc/invoke.texi (-mloongson-ext): Document.
2208 2018-11-07 Chenghua Xu <paul.hua.gm@gmail.com>
2210 * config.gcc (extra_headers): Add loongson-mmiintrin.h.
2211 * config/mips/loongson.md: Move to ...
2212 * config/mips/loongson-mmi.md: here; Adjustment.
2213 * config/mips/loongson.h: Move to ...
2214 State as deprecated. Include loongson-mmiintrin.h for back
2215 compatibility and warning.
2216 * config/mips/loongson-mmiintrin.h: ... here.
2217 * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
2218 mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
2219 TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
2220 (mips_option_override): Make sure MMI use hard float;
2221 (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
2222 mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
2223 mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
2224 TARGET_LOONGSON_VECTORS.
2225 * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
2226 (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
2227 (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
2228 (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
2230 (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
2231 TARGET_LOONGSON_VECTORS.
2232 * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
2233 TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
2234 (Loongson MMI patterns): Include loongson-mmi.md instead of
2236 * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
2237 * gcc/doc/invoke.texi (-mloongson-mmi): Document.
2239 2018-11-07 Richard Biener <rguenther@suse.de>
2242 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
2243 BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
2245 2018-11-07 Alexandre Oliva <aoliva@redhat.com>
2247 PR rtl-optimization/87874
2248 * lra.c (lra_substitute_pseudo): Do not create a subreg for
2251 2018-11-06 Aaron Sawdey <acsawdey@linux.ibm.com>
2253 * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
2254 if not in indexed or indirect form.
2255 (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
2256 (bswap<mode>2_store): Ditto.
2258 2018-11-06 Richard Earnshaw <rearnsha@arm.com>
2260 * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
2263 2018-11-06 Richard Biener <rguenther@suse.de>
2265 PR tree-optimization/86850
2266 * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
2267 instead of src.m_vec.
2269 2018-11-06 Jan Hubicka <jh@suse.cz>
2271 * tree.c (fld_simplified_type_name): Break out form ...
2272 (free_lang_data_in_type): ... here.
2273 (fld_type_variant_equal_p): Use it.
2275 2018-11-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2277 * config/default-d.c: Include memmodel.h.
2279 * config/sol2-d.c: New file.
2280 * config/t-sol2 (sol2-d.o): New rule.
2281 * config.gcc <*-*-solaris2*>: Set d_target_objs,
2282 target_has_targetdm.
2284 2018-11-06 Jan Hubicka <jh@suse.cz>
2286 * tree.c (fld_type_variant): Also copy alignment; be sure that
2287 new variant is equal.
2289 2018-11-06 Ilya Leoshkevich <iii@linux.ibm.com>
2292 * config/s390/s390.md: Add relative_long attribute.
2294 2018-11-06 Jan Hubicka <jh@suse.cz>
2296 * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
2297 * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
2298 * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
2299 stream TYPE_NEEDS_CONSTRUCTING.
2300 * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
2301 * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
2303 2018-11-06 Richard Biener <rguenther@suse.de>
2305 * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
2307 (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
2309 2018-11-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2311 * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
2312 REG_CLASS_CONTENTS[GEN_REGS].
2313 (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
2315 2018-11-06 Jan Hubicka <jh@suse.cz>
2317 * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
2319 2018-11-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2322 * config/sol2.h (ASAN_CC1_SPEC): Define.
2323 (LD_WHOLE_ARCHIVE_OPTION): Define.
2324 (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
2325 (ASAN_REJECT_SPEC): Provide default.
2326 (LIBASAN_EARLY_SPEC): Define.
2327 (LIBTSAN_EARLY_SPEC): Define.
2328 (LIBLSAN_EARLY_SPEC): Define.
2329 * config/i386/sol2.h (CC1_SPEC): Redefine.
2330 (ASAN_REJECT_SPEC): Define.
2332 * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
2333 (TARGET_ASAN_SHADOW_OFFSET): Define.
2334 (sparc_asan_shadow_offset): New function.
2335 * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
2336 (ASAN_REJECT_SPEC): Define.
2338 2018-11-06 Jan Hubicka <jh@suse.cz>
2340 * tree.c (fld_type_variant): Copy canonical type.
2341 (fld_incomplete_type_of): Check that canonical types looks sane;
2342 copy canonical type.
2343 (verify_type): Accept when incomplete type has complete canonical type.
2345 2018-11-06 Jan Hubicka <jh@suse.cz>
2347 * tree.c (free_lang_data): Reset overwite_assembler_name,
2348 print_xnode, print_decl, print_type and print_identifier of
2351 2018-11-06 Richard Biener <rguenther@suse.de>
2353 PR tree-optimization/87889
2354 * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
2355 Do nothing if old and new arg are the same
2357 2018-11-06 Andreas Krebbel <krebbel@linux.ibm.com>
2360 * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
2361 attributes for operands 3 and 4.
2363 2018-11-06 Richard Biener <rguenther@suse.de>
2366 * simplify-rtx.c (simplify_binary_operation_1): Add pattern
2367 matching bitfield insertion.
2369 2018-11-06 Alexandre Oliva <aoliva@redhat.com>
2371 * auto-inc-dec.c: Include valtrack.h. Improve comments.
2372 (reg_next_debug_use): New.
2373 (attempt_change): Propagate adjusted expression into affected
2375 (merge_in_block): Track uses in debug insns.
2376 (pass_inc_dec::execute): Allocate and release
2379 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
2381 * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
2382 (_mm512_fixupimm_round_pd): Update parameters and builtin.
2383 (_mm512_maskz_fixupimm_round_pd): Ditto.
2384 (_mm512_fixupimm_round_ps): Ditto.
2385 (_mm512_maskz_fixupimm_round_ps): Ditto.
2386 (_mm_fixupimm_round_sd): Ditto.
2387 (_mm_maskz_fixupimm_round_sd): Ditto.
2388 (_mm_fixupimm_round_ss): Ditto.
2389 (_mm_maskz_fixupimm_round_ss): Ditto.
2390 (_mm512_fixupimm_pd): Ditto.
2391 (_mm512_maskz_fixupimm_pd): Ditto.
2392 (_mm512_fixupimm_ps): Ditto.
2393 (_mm512_maskz_fixupimm_ps): Ditto.
2394 (_mm_fixupimm_sd): Ditto.
2395 (_mm_maskz_fixupimm_sd): Ditto.
2396 (_mm_fixupimm_ss): Ditto.
2397 (_mm_maskz_fixupimm_ss): Ditto.
2398 (_mm512_mask_fixupimm_round_pd): Update builtin.
2399 (_mm512_mask_fixupimm_round_ps): Ditto.
2400 (_mm_mask_fixupimm_round_sd): Ditto.
2401 (_mm_mask_fixupimm_round_ss): Ditto.
2402 (_mm512_mask_fixupimm_pd): Ditto.
2403 (_mm512_mask_fixupimm_ps): Ditto.
2404 (_mm_mask_fixupimm_sd): Ditto.
2405 (_mm_mask_fixupimm_ss): Ditto.
2406 * config/i386/avx512vlintrin.h:
2407 (_mm256_fixupimm_pd): Update parameters and builtin.
2408 (_mm256_maskz_fixupimm_pd): Ditto.
2409 (_mm256_fixupimm_ps): Ditto.
2410 (_mm256_maskz_fixupimm_ps): Ditto.
2411 (_mm_fixupimm_pd): Ditto.
2412 (_mm_maskz_fixupimm_pd): Ditto.
2413 (_mm_fixupimm_ps): Ditto.
2414 (_mm_maskz_fixupimm_ps): Ditto.
2415 (_mm256_mask_fixupimm_pd): Update builtin.
2416 (_mm256_mask_fixupimm_ps): Ditto.
2417 (_mm_mask_fixupimm_pd): Ditto.
2418 (_mm_mask_fixupimm_ps): Ditto.
2419 * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
2420 * config/i386/i386-builtin.def: Update builtin definitions.
2421 * config/i386/i386.c: Handle new builtin types and remove useless ones.
2422 * config/i386/sse.md: Update VFIXUPIMM* patterns.
2423 (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2424 (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2425 (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
2426 (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2427 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2428 (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
2429 * config/i386/subst.md:
2430 (round_saeonly_sd_mask_operand4): Add new subst_attr.
2431 (round_saeonly_sd_mask_op4): Ditto.
2432 (round_saeonly_expand_operand5): Ditto.
2433 (round_saeonly_expand): Update.
2435 2018-11-05 Max Filippov <jcmvbkbc@gmail.com>
2437 * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
2439 2018-11-05 Segher Boessenkool <segher@kernel.crashing.org>
2441 PR rtl-optimization/87871
2442 * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
2444 2018-11-05 Paul Koning <ni1d@arrl.net>
2446 * doc/sourcebuild.texi (target attributes): Document new "inf"
2447 effective target keyword.
2449 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
2451 * config/s390/s390.c (s390_register_move_cost): Increase costs for
2452 moves involving the CC reg.
2454 2018-11-05 Richard Biener <rguenther@suse.de>
2456 PR tree-optimization/87873
2457 * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
2459 * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
2460 * tree-vect-loop.c (vect_transform_loop): When splitting the
2461 loop exit also create forwarder PHIs for constants.
2462 * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
2463 Handle constant to_arg, add extra checking we match up the correct
2466 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
2468 * config/s390/s390.md: QImode and HImode for load on condition.
2470 2018-11-05 Robin Dapp <rdapp@linux.ibm.com>
2472 * config/s390/predicates.md: Fix typo.
2473 * config/s390/s390.md: Allow immediates for load on condition.
2475 2018-11-05 Martin Liska <mliska@suse.cz>
2477 * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
2478 * bitmap.h (struct bitmap_usage): Likewise.
2479 * ggc-common.c (SCALE): Remove.
2481 (struct ggc_usage): Use SIZE_AMOUNT. And update
2483 * ggc-page.c (SCALE): Remove.
2484 (STAT_LABEL): Remove.
2485 (ggc_print_statistics): Use SIZE_AMOUNT.
2486 * gimple.h (SCALE): Remove.
2488 * input.c (ONE_K): Remove.
2491 (STAT_LABEL): Likewise.
2492 (FORMAT_AMOUNT): Likewise.
2493 (dump_line_table_statistics): Use SIZE_AMOUNT.
2494 * mem-stats.h (struct mem_usage): Likewise.
2495 * rtl.c (dump_rtx_statistics): Likewise.
2496 (rtx_alloc_counts): Change type to size_t.
2497 (rtx_alloc_sizes): Likewise.
2498 (rtx_count_cmp): New.
2499 (dump_rtx_statistics): Sort first based on counts.
2500 * tree.c (tree_nodes_cmp): New.
2501 (tree_codes_cmp): New.
2502 (dump_tree_statistics): Sort first based on counts.
2503 * system.h (ONE_K): New.
2505 (SIZE_SCALE): Likewise.
2506 (SIZE_LABEL): Likewise.
2507 (SIZE_AMOUNT): Likewise.
2508 * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
2509 * tree-dfa.c (dump_dfa_stats): Likewise.
2510 * tree-phinodes.c (phinodes_print_statistics): Likewise.
2511 * tree-ssanames.c (ssanames_print_statistics): Likewise.
2512 * tree.c (dump_tree_statistics): Likewise.
2513 * vec.c (struct vec_usage): Likewise.
2514 * trans-mem.c (tm_mangle): Enlarge buffer in order to not
2515 trigger a -Werror=format-overflow with
2516 --enable-gather-detailed-stats.
2518 2018-11-05 Martin Liska <mliska@suse.cz>
2520 * mem-stats.h (mem_alloc_description::release_instance_overhead):
2522 * vec.c (struct vec_usage): Register m_element_size.
2523 (vec_prefix::register_overhead): New arguments: elements and
2525 (vec_prefix::release_overhead): Subtract elements.
2526 * vec.h (struct vec_prefix): Change signature.
2527 (va_heap::reserve): Pass proper arguments.
2528 (va_heap::release): Likewise.
2530 2018-11-05 Martin Liska <mliska@suse.cz>
2532 * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
2536 2018-11-05 Richard Biener <rguenther@suse.de>
2538 * tree-scalar-evolution.h (final_value_replacement_loop): Update
2540 * tree-scalar-evolution.c (final_value_replacement_loop): Return
2541 whether anything was done.
2542 (scev_const_prop): Remove constant propagation part, fold
2544 * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
2545 (pass_data_scev_cprop): TODO_cleanup_cfg is now done
2548 2018-11-05 Jakub Jelinek <jakub@redhat.com>
2550 PR tree-optimization/87859
2551 * gimple-ssa-store-merging.c (struct merged_store_group): Add
2552 only_constants and first_nonmergeable_order members.
2553 (merged_store_group::merged_store_group): Initialize them.
2554 (merged_store_group::do_merge): Clear only_constants member if
2555 adding something other than INTEGER_CST store.
2556 (imm_store_chain_info::coalesce_immediate_stores): Don't merge
2557 stores with order >= first_nonmergeable_order. Use
2558 merged_store->only_constants instead of always recomputing it.
2559 Set merged_store->first_nonmergeable_order if we've skipped any
2560 stores. Attempt to merge overlapping INTEGER_CST stores that
2561 we would otherwise skip.
2564 * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
2566 2018-11-05 Xuepeng Guo <xuepeng.guo@intel.com>
2569 * config/i386/emmintrin.h (__v16qs): New to cope with option
2571 (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
2572 (_mm_cmplt_epi8): Likewise.
2573 (_mm_cmpgt_epi8): Likewise.
2575 2018-11-05 Richard Biener <rguenther@suse.de>
2577 PR rtl-optimization/87852
2578 * fwprop.c (use_killed_between): Only consider single-defs of the
2579 use whose definition statement dominates the use.
2581 2018-11-05 Martin Liska <mliska@suse.cz>
2584 * doc/invoke.texi: Remove options that are
2585 not disabled with -Os.
2587 2018-11-05 Martin Liska <mliska@suse.cz>
2590 * doc/extend.texi: Update constrain about the last argument
2591 of __builtin_expect_with_probability.
2593 2018-11-05 Martin Liska <mliska@suse.cz>
2596 * predict.c (expr_expected_value_1): Verify
2597 that last argument is a real constants and emit
2600 2018-11-05 Martin Liska <mliska@suse.cz>
2602 PR gcov-profile/77698
2603 * ipa-profile.c (ipa_profile): Adjust hotness threshold
2606 2018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
2608 PR tree-optimization/86572
2609 * builtins.c (c_strlen): Handle negative offsets in a safe way.
2611 2018-11-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
2613 PR tree-optimization/87672
2614 * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
2615 * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
2617 2018-11-04 Uros Bizjak <ubizjak@gmail.com>
2620 * cfgexpand.c (pass_expand::execute): Move the call to
2621 finish_eh_generation in front of the call to expand_stack_alignment.
2623 2018-11-04 Venkataramanan Kumar <venkataramanan.kumar@amd.com>
2625 * common/config/i386/i386-common.c (processor_alias_table): Add
2627 * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
2628 (case ${target}): Add znver2.
2629 * config/i386/driver-i386.c: (host_detect_local_cpu): Let
2630 -march=native recognize znver2 processors.
2631 * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
2632 * config/i386/i386.c (m_znver2): New definition.
2633 (m_ZNVER): New definition.
2634 (m_AMD_MULTIPLE): Includes m_znver2.
2635 (processor_cost_table): Add znver2 entry.
2636 (processor_target_table): Add znver2 entry.
2637 (get_builtin_code_for_version): Set priority for
2639 (processor_model): Add M_AMDFAM17H_ZNVER2.
2640 (arch_names_table): Ditto.
2641 (ix86_reassociation_width): Include znver2.
2642 * config/i386/i386.h (TARGET_znver2): New definition.
2643 (struct ix86_size_cost): Add TARGET_ZNVER2.
2644 (enum processor_type): Add PROCESSOR_ZNVER2.
2645 * config/i386/i386.md (define_attr "cpu"): Add znver2.
2646 * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
2647 * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
2648 (ix86_adjust_cost): Add znver2.
2649 * config/i386/x86-tune.def: Replace m_ZNVER1 by m_ZNVER.
2650 * gcc/doc/extend.texi: Add details about znver2.
2651 * gcc/doc/invoke.texi: Add details about znver2.
2653 2018-11-03 Sandra Loosemore <sandra@codesourcery.com>
2657 * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
2660 2018-11-02 Aaron Sawdey <acsawdey@linux.ibm.com>
2662 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
2663 attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
2665 2018-11-02 Richard Earnshaw <rearnsha@arm.com>
2667 * config/aarch64/aarch64.c ((aarch64_override_options): Disable
2668 shrink-wrapping when -mtrack-speculation.
2670 2018-11-02 Richard Biener <rguenther@suse.de>
2672 * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
2673 (pop_cost_one_pair): Do not free pair.
2674 (pop_best_coalesce): Likewise.
2675 (create_coalesce_list): Initialize obstack.
2676 (delete_coalesce_list): Free obstack.
2677 (find_coalesce_pair): Obstack-allocate coalesce pairs.
2678 (add_cost_one_coalesce): Likewise.
2679 (struct live_track): Remove bitmap pointer indirections.
2680 (new_live_track): Adjust.
2681 (delete_live_track): Likewise.
2682 (live_track_remove_partition): Likewise.
2683 (live_track_add_partition): Likewise.
2684 (live_track_live_p): Likewise.
2685 (live_track_process_def): Likewise.
2686 (live_track_clear_base_vars): Likewise.
2688 2018-11-02 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2690 * configure.ac (gcc_cv_as_sparc_register_op): Remove.
2691 * configure: Regenerate.
2692 * config.in: Regenerate.
2693 * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
2694 HAVE_AS_REGISTER_PSEUDO_OP guard.
2695 * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
2696 !HAVE_AS_REGISTER_PSEUDO_OP support.
2697 (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
2700 2018-11-02 Richard Biener <rguenther@suse.de>
2702 * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
2705 2018-11-02 Richard Biener <rguenther@suse.de>
2707 PR tree-optimization/87776
2708 * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
2709 executable when iterating but running into rpo-vn-max-loop-depth
2710 and not eliding the iteration.
2712 2018-11-30 Jan Hubicka <jh@suse.cz>
2714 * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
2717 2018-11-01 Aaron Sawdey <acsawdey@linux.ibm.com>
2719 * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
2721 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
2722 Combine with rs6000_address_for_fpconvert.
2723 (rs6000_address_for_fpconvert): Combine with
2724 rs6000_force_indexed_or_indirect_mem.
2725 (rs6000_expand_vector_init): Change function call from
2726 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2727 * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
2728 rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2729 (floatsi<mode>2_lfiwax_mem): Ditto.
2730 (floatunssi<mode>2_lfiwzx): Ditto.
2731 (floatunssi<mode>2_lfiwzx_mem): Ditto.
2732 (float<QHI:mode><FP_ISA3:mode>2): Ditto.
2733 (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
2734 (fix_trunc<mode>si2_stfiwx): Ditto.
2735 (fixuns_trunc<mode>si2_stfiwx): Ditto.
2736 (float_<mode>si2_hw): Ditto.
2737 (floatuns_<mode>si2_hw): Ditto.
2738 * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
2739 (vsx_splat_<mode>): Ditto.
2741 2018-11-01 Joseph Myers <joseph@codesourcery.com>
2743 * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
2744 at end of assembler input text.
2745 * configure: Regenerate.
2747 2018-11-01 Jakub Jelinek <jakub@redhat.com>
2749 PR tree-optimization/87826
2750 * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
2751 negative or larger or equal to type's precision.
2753 2018-10-31 Alexandre Oliva <aoliva@redhat.com>
2755 * opts.c (default_options_table): Do not enable
2756 OPT_fdelayed_branch at -Og.
2757 * doc/invoke.texi (-fdelayed-branch): Document it.
2759 2018-10-31 Richard Henderson <richard.henderson@linaro.org>
2761 * optabs-libfuncs.c (build_libfunc_function_visibility):
2762 New, split out from...
2763 (build_libfunc_function): ... here.
2764 (init_one_libfunc_visibility): New, split out from ...
2765 (init_one_libfunc): ... here.
2767 * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
2768 scratch register need not be early-clobber. Document the reason
2769 why we cannot use ST<OP>.
2771 2018-10-31 Joseph Myers <joseph@codesourcery.com>
2774 * configure.ac: Remove AC_PREREQ. Use AC_LANG_SOURCE. Use single
2775 line for second argument of AC_DEFINE_UNQUOTED.
2776 * doc/install.texi (Tools/packages necessary for modifying GCC):
2777 Update to autoconf 2.69 and automake 1.15.1.
2778 * aclocal.m4, config.in, configure: Regenerate.
2780 2018-10-31 Pat Haugen <pthaugen@us.ibm.com>
2782 * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
2785 2018-10-31 Martin Liska <mliska@suse.cz>
2788 * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
2789 Add new argument invalid_extension.
2790 (aarch64_get_all_extension_candidates): New function.
2791 (aarch64_rewrite_selected_cpu): Add NULL to function call.
2792 * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
2794 (aarch64_get_all_extension_candidates): New function.
2795 * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
2796 argument invalid_extension.
2797 (aarch64_parse_cpu): Likewise.
2798 (aarch64_print_hint_for_extensions): New function.
2799 (aarch64_validate_mcpu): Provide hint about invalid extension.
2800 (aarch64_validate_march): Likewise.
2801 (aarch64_handle_attr_arch): Pass new argument.
2802 (aarch64_handle_attr_cpu): Provide hint about invalid extension.
2803 (aarch64_handle_attr_isa_flags): Likewise.
2805 2018-10-31 Richard Biener <rguenther@suse.de>
2809 * tree-outof-ssa.c (insert_backedge_copies): Restrict
2810 copy generation to useful cases. Place the copy before
2811 the definition of the backedge value when possible.
2813 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
2815 * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
2816 * config/arc/arc.c (arc_active_insn): New function.
2817 (check_store_cacheline_hazard): Likewise.
2818 (workaround_arc_anomaly): Use check_store_cacheline_hazard.
2819 (arc_override_options): Disable delay slot scheduler for older
2821 (arc_store_addr_hazard_p): New implementation, old one renamed to
2823 (arc_store_addr_hazard_internal_p): Renamed.
2824 (arc_reorg): Don't combine into brcc instructions which are part
2825 of hardware hazard solution.
2826 * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
2827 (tune_arc700): Likewise.
2828 * config/arc/arc.opt (arc7xx): New tune value.
2829 * config/arc/arc700.md: Improve A7 scheduler.
2831 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
2833 * config/arc/arc.c (arc_override_options): Remove
2834 TARGET_COMPACT_CASESI.
2835 * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
2836 (CASE_VECTOR_MODE): Likewise.
2837 (CASE_VECTOR_PC_RELATIVE): Likewise.
2838 (CASE_VECTOR_SHORTEN_MODE): Likewise.
2839 (CASE_VECTOR_SHORTEN_MODE1): Delete.
2840 (ADDR_VEC_ALIGN): Update.
2841 (ASM_OUTPUT_CASE_LABEL): Undefine.
2842 (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
2843 (TARGET_BI_BIH): Define.
2844 (DEFAULT_BRANCH_INDEX): Likewise.
2845 * config/arc/arc.md (casesi): Rework to accept BI/BIH
2846 instructions, remove compact_casesi use case.
2847 (casesi_compact_jump): Remove.
2848 (casesi_dispatch): New pattern.
2849 * config/arc/arc.opt: Add mbranch-index option. Deprecate
2850 compact_casesi option.
2851 * doc/invoke.texi: Document mbranch-index option.
2853 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
2855 * config/arc/arc.c (arc_get_tp): Remove function.
2856 (arc_emit_call_tls_get_addr): Likewise.
2857 (arc_call_tls_get_addr): New function.
2858 (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
2859 * config/arc/arc.md (tls_load_tp_soft): Remove.
2860 (tls_gd_get_addr): Likewise.
2862 2018-10-31 Claudiu Zissulescu <claziss@synopsys.com>
2864 * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
2865 (mulsi3_600_lib): Remove pattern.
2866 (umulsi3_highpart_600_lib_le): Likewise.
2867 (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
2868 (umulsidi3): Remove call to umulsidi3_600_lib.
2869 (umulsidi3_600_lib): Remove pattern.
2870 (peephole2): Remove peephole using the above deprecated patterns.
2872 2018-10-31 Thomas Preud'homme <thomas.preudhomme@linaro.org>
2875 * config/arm/arm.c (arm_option_check_internal): Disable the combined
2876 use of -mslow-flash-data and -mword-relocations.
2877 (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
2878 * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
2880 * doc/invoke.texi (-mword-relocations): Mention conflict with
2882 (-mslow-flash-data): Reciprocally.
2884 2018-10-31 Richard Henderson <richard.henderson@linaro.org>
2886 * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
2887 16-byte modes held in GP registers to use an even regno.
2889 * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
2890 (aarch64_atomic_ldop_supported_p): Remove.
2891 (aarch64_gen_atomic_ldop): Remove.
2892 * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
2893 Fully expand LSE operations here.
2894 (atomic_fetch_<atomic_optab><ALLI>): Likewise.
2895 (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
2896 (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
2897 and use ATOMIC_LDOP instead; use register_operand for the input;
2898 drop the split and emit insns directly.
2899 (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
2900 (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
2901 (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
2903 * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
2904 (aarch64_gen_atomic_ldop): Don't call it.
2905 * config/aarch64/atomics.md (atomic_exchange<ALLI>):
2906 Use aarch64_reg_or_zero.
2907 (aarch64_atomic_exchange<ALLI>): Likewise.
2908 (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
2909 operand 0; use aarch64_reg_or_zero for input; merge ...
2910 (@aarch64_atomic_swp<ALLI>): ... this and remove.
2912 * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
2913 (aarch64_split_compare_and_swap): Use it.
2914 (aarch64_expand_compare_and_swap): Likewise. Remove convert_modes;
2915 test oldval against the proper predicate.
2916 * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
2917 Use nonmemory_operand for expected.
2918 (cas_short_expected_pred): New.
2919 (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
2920 (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
2921 * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
2922 (aarch64_plushi_operand): New.
2924 * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
2925 Force oldval into the rval register for TARGET_LSE; emit the compare
2926 during initial expansion so that it may be deleted if unused.
2927 (aarch64_gen_atomic_cas): Remove.
2928 * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
2929 Change =&r to +r for operand 0; use match_dup for operand 2;
2930 remove is_weak and mod_f operands as unused. Drop the split
2932 (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
2933 (@aarch64_compare_and_swap<GPI>_lse): Similarly.
2934 (@aarch64_atomic_cas<GPI>): Similarly.
2936 2018-10-31 Richard Biener <rguenther@suse.de>
2938 * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
2941 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2943 * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
2944 * config/aarch64/aarch64.md: Include saphira.md
2945 * config/aarch64/saphira.md: New file for pipeline description.
2947 2018-10-30 Martin Sebor <msebor@redhat.com>
2950 * gimple-ssa-sprintf.c (format_directive): Use %G to include
2952 (sprintf_dom_walker::compute_format_length):
2953 Avoid setting POSUNDER4K here.
2954 (get_destination_size): Handle null argument values.
2955 (get_user_idx_format): New function.
2956 (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
2957 functions, including user-defined with attribute format printf.
2958 Use %G to include inlining context.
2959 Set POSUNDER4K here.
2961 2018-10-30 Jan Hubicka <jh@suse.cz>
2963 * params.def (lto-partitions): Bump from 32 to 128.
2965 2018-10-30 Jan Hubicka <jh@suse.cz>
2968 (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
2970 (free_lang_data_in_type): Forward declare.
2971 (fld_type_variant_equal_p): New function.
2972 (fld_type_variant): New function
2973 (fld_incomplete_types): New hash.
2974 (fld_incomplete_type_of): New function
2975 (fld_simplfied-type): New function.
2976 (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
2977 (free_lang_data): Allocate and free fld_incomplete_type; update call
2978 of free_lang_data_in_decl.
2980 2018-10-30 Eric Botcazou <ebotcazou@adacore.com>
2982 * gcov.c (output_lines): Remove duplicate line.
2984 2018-10-30 Aaron Sawdey <acsawdey@linux.ibm.com>
2986 * config/rs6000/rs6000.md (bswapdi2): Force address into register
2987 if not in indexed or indirect form.
2988 (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
2989 (bswapdi2_store): Ditto.
2990 * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
2992 * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
2993 Prototype for helper function.
2995 2018-10-30 Martin Sebor <msebor@redhat.com>
2997 * doc/extend.texi (optimize): Clarify/expand attribute documentation.
2998 (target, pragma GCC optimize, pragma GCC target): Ditto.
3000 2018-10-30 Jonathan Wakely <jwakely@redhat.com>
3002 * doc/extend.texi: Fix prototype and description of
3003 __builtin_expect_with_probability.
3005 2018-10-30 Michael Ploujnikov <michael.ploujnikov@oracle.com>
3007 * cgraph.h (clone_function_name_1): Replaced by new
3008 clone_function_name_numbered that takes name as string; for
3009 privatize_symbol_name_1 use only.
3010 (clone_function_name): Renamed to
3011 clone_function_name_numbered to be explicit about numbering.
3012 (clone_function_name): New two-argument function that does
3013 not number its output.
3014 (clone_function_name): New three-argument function that
3015 takes a number to append to its output.
3016 * cgraphclones.c (duplicate_thunk_for_node):
3017 (clone_function_name_1): Renamed.
3018 (clone_function_name_numbered): Two new functions.
3019 (clone_function_name): Improved documentation.
3020 (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
3021 * config/rs6000/rs6000.c (make_resolver_func): Ditto.
3022 * final.c (final_scan_insn_1): Use the new clone_function_name
3024 * multiple_target.c (create_dispatcher_calls): Ditto.
3025 (create_target_clone): Ditto.
3026 * omp-expand.c (grid_expand_target_grid_body): Ditto.
3027 * omp-low.c (create_omp_child_function_name): Ditto.
3028 * omp-simd-clone.c (simd_clone_create): Ditto.
3029 * symtab.c (simd_symtab_node::noninterposable_alias): Use the
3030 new clone_function_name without numbering.
3032 2018-10-30 Richard Earnshaw <rearnsha@arm.com>
3034 * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
3035 Assert that the allocation size is not zero.
3037 2018-10-30 Richard Biener <rguenther@suse.de>
3039 PR tree-optimization/87800
3040 * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
3041 non-induction or reduction PHIs.
3043 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
3045 * config/aarch64/falkor-tag-collision-avoidance.c
3046 (execute_tag_collision_avoidance): Call df_note_add_problem.
3048 2018-10-30 Martin Liska <mliska@suse.cz>
3050 * doc/extend.texi: Fix typo in documentation
3051 of __builtin_expect_with_probability.
3053 2018-10-29 David Malcolm <dmalcolm@redhat.com>
3056 * input.c (get_substring_ranges_for_loc): Detect if
3057 linemap_resolve_location gives us a NULL map, and reject
3060 2018-10-29 Iain Buclaw <ibuclaw@gdcproject.org>
3062 * config.gcc (xstormy16-*-elf): Set tm_d_file.
3064 2018-10-29 David Malcolm <dmalcolm@redhat.com>
3065 Martin Sebor <msebor@redhat.com>
3066 Manuel Lopez-Ibanez <manu@gcc.gnu.org>
3068 * doc/ux.texi (Quoting): New subsection, adapted from material at
3069 https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
3070 MartinSebor and ManuelLopezIbanez.
3071 (Fix-it hints): Note that fix-it hints shouldn't be marked for
3074 2018-10-29 Kugan Vivekanandarajah <kuganv@linaro.org>
3077 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
3080 2018-10-29 Paul A. Clarke <pc@us.ibm.com>
3082 * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
3084 2018-10-29 Paul Koning <ni1d@arrl.net>
3086 * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
3088 2018-10-29 Paul A. Clarke <pc@us.ibm.com>
3090 * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
3091 _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
3092 _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
3093 _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
3094 _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
3095 _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
3096 _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
3097 _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
3098 Change 'vector' to '__vector'.
3099 * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
3100 _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
3101 _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
3102 _mm_avg_pu16): Likewise. And, whitespace corrections.
3104 2018-10-29 Richard Biener <rguenther@suse.de>
3106 PR tree-optimization/87785
3107 * tree-vect-slp.c (vect_gather_slp_loads): Only gather
3110 2018-10-29 Olivier Hainque <hainque@adacore.com>
3112 * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
3114 2018-10-29 Olivier Hainque <hainque@adacore.com>
3116 * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
3117 ports configuration macro, defaults to "gnu".
3118 (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
3119 TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
3121 2018-10-29 Olivier Hainque <hainque@adacore.com>
3123 * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
3125 2018-10-29 Richard Biener <rguenther@suse.de>
3127 PR tree-optimization/87790
3128 * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
3129 (vect_make_slp_decision): Adjust.
3130 (vect_slp_analyze_bb_1): Likewise.
3131 (vect_detect_hybrid_slp_stmts): Properly union SLP type over
3134 2018-10-29 Richard Biener <rguenther@suse.de>
3136 PR tree-optimization/87785
3137 * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
3139 (vect_build_slp_tree): Likewise.
3140 (vect_gather_slp_loads): New function.
3141 (vect_analyze_slp_instance): Gather loads separately from the
3144 2018-10-29 Martin Liska <mliska@suse.cz>
3146 * Makefile.in: Make dependency to json.o.
3147 * doc/gcov.texi: Document new JSON format, remove
3148 old intermediate format documentation.
3149 * gcov.c (struct function_info): Come up with m_name and
3151 (function_info::function_info): Initialize it.
3152 (function_info::~function_info): Release it.
3153 (main): Rename flag_intermediate_format to flag_json_format.
3154 (print_usage): Describe --json-format.
3155 (process_args): Set flag_json_format.
3156 (output_intermediate_line): Remove.
3157 (output_intermediate_json_line): Likewise.
3158 (get_gcov_intermediate_filename): Return new extension
3160 (output_intermediate_file): Implement JSON emission.
3161 (output_json_intermediate_file): Implement JSON emission.
3162 (generate_results): Use ::get_name for function name.
3163 Handle JSON output file.
3164 (read_graph_file): Use ::get_name instead of cplus_demangle.
3165 (read_count_file): Likewise.
3166 (solve_flow_graph): Likewise.
3167 (add_line_counts): Likewise.
3168 (accumulate_line_counts): Use new flag_json_format.
3169 (output_function_details): Use ::get_name instead of cplus_demangle.
3170 (output_lines): Likewise.
3171 * json.cc (test_writing_literals): Add new tests.
3172 * json.h (class literal): Add new boolean constructor.
3174 2018-10-29 Segher Boessenkool <segher@kernel.crashing.org>
3176 PR rtl-optimization/87701
3177 PR rtl-optimization/87780
3178 * combine.c (make_more_copies): Rewrite.
3180 2018-10-28 Kugan Vivekanandarajah <kuganv@linaro.org>
3182 * doc/generic.texi (ABSU_EXPR): Document.
3183 * match.pd (absu(x)*absu(x) -> x*x): Handle.
3184 (absu(absu(X)) -> absu(X)): Likewise.
3185 (absu(-X) -> absu(X)): Likewise.
3186 (absu(X) where X is nonnegative -> X): Likewise.
3188 2018-10-28 Iain Buclaw <ibuclaw@gdcproject.org>
3190 * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
3191 (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
3192 (tm_d.h, cs-tm_d.h, default-d.o): New rules.
3193 (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
3194 (s-tm-texi): Also check timestamp on d-target.def.
3195 (generated_files): Add TM_D_H and d-target-hooks-def.h.
3196 (build/genhooks.o): Also depend on D_TARGET_DEF.
3197 * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
3199 * config/aarch64/aarch64-d.c: New file.
3200 * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
3202 * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
3204 * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
3205 * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
3206 * config/arm/arm-d.c: New file.
3207 * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
3208 * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
3209 * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3210 * config/arm/t-arm (arm-d.o): New rule.
3211 * config/default-d.c: New file.
3212 * config/glibc-d.c: New file.
3213 * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3214 * config/i386/i386-d.c: New file.
3215 * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
3216 * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
3217 * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3218 (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
3219 * config/i386/t-i386 (i386-d.o): New rule.
3220 * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3221 * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3222 * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
3223 * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3224 * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3225 * config/mips/mips-d.c: New file.
3226 * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
3227 * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
3228 * config/mips/t-mips (mips-d.o): New rule.
3229 * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3230 * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3231 * config/powerpcspe/powerpcspe-d.c: New file.
3232 * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
3234 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
3235 Support GNU D by using 0 as the language type.
3236 * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
3237 * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
3238 * config/riscv/riscv-d.c: New file.
3239 * config/riscv/riscv-protos.h (riscv_d_target_versions): New
3241 * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
3242 * config/riscv/t-riscv (riscv-d.o): New rule.
3243 * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3244 * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3245 * config/rs6000/rs6000-d.c: New file.
3246 * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
3248 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
3249 Support GNU D by using 0 as the language type.
3250 * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
3251 * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
3252 * config/s390/s390-d.c: New file.
3253 * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
3254 * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
3255 * config/s390/t-s390 (s390-d.o): New rule.
3256 * config/sparc/sparc-d.c: New file.
3257 * config/sparc/sparc-protos.h (sparc_d_target_versions): New
3259 * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
3260 * config/sparc/t-sparc (sparc-d.o): New rule.
3261 * config/t-glibc (glibc-d.o): New rule.
3262 * configure: Regenerated.
3263 * configure.ac (tm_d_file): New variable.
3264 (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
3265 * doc/contrib.texi (Contributors): Add self for the D frontend.
3266 * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
3267 * doc/install.texi (Configuration): Mention libphobos as an option for
3268 --enable-shared. Mention d as an option for --enable-languages.
3269 (Testing): Mention check-d as a target.
3270 * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
3271 name suffixes. Mention d as a -x option.
3272 * doc/sourcebuild.texi (Top Level): Mention libphobos.
3273 * doc/standards.texi (Standards): Add section on D language.
3274 * doc/tm.texi: Regenerated.
3275 * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
3276 TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
3277 * dwarf2out.c (is_dlang): New function.
3278 (gen_compile_unit_die): Use DW_LANG_D for D.
3279 (declare_in_namespace): Return module die for D, instead of adding
3280 extra declarations into the namespace.
3281 (gen_namespace_die): Generate DW_TAG_module for D.
3282 (gen_decl_die): Handle CONST_DECLSs for D.
3283 (dwarf2out_decl): Likewise.
3284 (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
3285 (prune_unused_types_walk): Handle DW_tag_interface_type same as other
3286 kinds of aggregates.
3287 * gcc.c (default_compilers): Add entries for .d, .dd and .di.
3288 * genhooks.c: Include d/d-target.def.
3290 2018-10-28 Iain Sandoe <iain@sandoe.co.uk>
3293 * config/rs6000/darwin.h (STACK_BOUNDARY): New.
3294 (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
3295 (STACK_DYNAMIC_OFFSET): Likewise.
3297 2018-10-27 Sandra Loosemore <sandra@codesourcery.com>
3300 * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
3303 2018-10-26 Tulio Magno Quites Machado Filho <tuliom@linux.ibm.com>
3305 * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
3306 return type and other typos.
3308 2018-10-26 Aaron Sawdey <acsawdey@linux.ibm.com>
3310 * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
3311 a shorter sequence with fewer branches.
3312 (emit_final_str_compare_gpr): Ditto.
3314 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
3316 * config/rs6000/tmmintrin.h: New file.
3317 * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
3319 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
3321 * config/rs6000/mmintrin.h: Enable 32bit compilation.
3322 * config/rs6000/xmmintrin.h: Likewise.
3324 2018-10-26 Paul A. Clarke <pc@us.ibm.com>
3326 * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
3328 2018-10-26 Richard Biener <rguenther@suse.de>
3330 * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
3332 (vect_mark_slp_stmts_relevant): Likewise.
3333 (vect_detect_hybrid_slp_stmts): Likewise.
3334 (vect_bb_slp_scalar_cost): Likewise.
3335 (vect_remove_slp_scalar_calls): Likewise.
3337 2018-10-26 Jan Hubicka <jh@suse.cz>
3339 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
3340 (warn_types_mismatch): Fix walk of DECL_NAME.
3341 (odr_types_equivalent_p): Fix overactive assert.
3343 2018-10-26 Richard Biener <rguenther@suse.de>
3345 PR tree-optimization/87105
3346 * tree-vectorizer.h (_slp_tree::refcnt): New member.
3347 * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
3349 (vect_create_new_slp_node): Initialize refcnt to one.
3351 (scalar_stmts_set_t, bst_fail): Remove.
3352 (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
3353 (vect_build_slp_tree): Add bst_map argument and lookup
3354 already created SLP nodes.
3355 (vect_print_slp_tree): Handle a SLP graph, print SLP node
3357 (vect_slp_rearrange_stmts): Handle a SLP graph.
3358 (vect_analyze_slp_instance): Adjust and free SLP nodes from
3359 the CSE map. Fix indenting.
3360 (vect_schedule_slp_instance): Add short-cut.
3362 2018-10-26 Martin Liska <mliska@suse.cz>
3365 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
3366 addr_expr and not with pointers.
3368 2018-10-26 Jan Hubicka <jh@suse.cz>
3370 * tree.c (free_lang_data_in_type): Only check main variants.
3371 * ipa-devirt.c (warn_odr): Make static.
3372 (types_same_for_odr): Drop strict variant.
3373 (types_odr_comparable): Likewise.
3374 (odr_or_derived_type_p): Look for main variants.
3375 (odr_name_hasher::equal): Cleanup comment.
3376 (odr_subtypes_equivalent): Add warn and warned arguments; check main
3378 (type_variants_equivalent_p): break out from ...
3379 (odr_types_equivalent): ... here; go for main variants where needed.
3380 (warn_odr): ... here; turn static.
3381 (warn_types_mismatch): Compare mangled names of main variants.
3382 * ipa-utils.h (types_odr_comparable): Drop strict parameter.
3383 (type_with_linkage_p): Sanity check that we look at main variant.
3384 * lto.c (lto_read_decls): Only consider main variant to be ODR type.
3385 * tree.h (types_same_for_odr): Drop strict argument.
3387 2018-10-26 Richard Biener <rguenther@suse.de>
3389 PR tree-optimization/87746
3390 * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
3391 Simplify and fix WRT strided store groups with size not
3392 equal to step in element count.
3393 (vect_analyze_group_access_1): Dump the whole group.
3395 2018-10-25 Carl Love <cel@us.ibm.com>
3397 * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
3398 P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
3399 P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name. Add quad
3400 precicion entry for each overloaded builtin.
3401 * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
3402 VSCEDPUO): Rename overloaded name.
3403 (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
3404 VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
3405 * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
3406 define_expand for xscmpexqp instruction.
3407 (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
3409 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
3410 Jinsong Ji <jji@us.ibm.com>
3412 * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
3413 function with vec_sl.
3414 (_mm_slli_epi32): Likewise.
3415 (_mm_slli_epi64): Likewise.
3416 (_mm_srai_epi16): Replace deprecated function with vec_sra.
3417 (_mm_srai_epi32): Likewise.
3418 (_mm_srli_epi16): Replace deprecated function with vec_sr.
3419 (_mm_srli_epi32): Likewise.
3420 (_mm_srli_epi64): Likewise.
3421 (_mm_sll_epi16): Replace deprecated function with vec_sl.
3422 (_mm_sll_epi32): Likewise.
3423 (_mm_sll_epi64): Likewise.
3424 (_mm_sra_epi16): Replace deprecated function with vec_sra.
3425 (_mm_sra_epi32): Likewise.
3426 (_mm_srl_epi16): Replace deprecated function with vec_sr.
3427 (_mm_srl_epi32): Likewise.
3428 (_mm_srl_epi64): Likewise.
3430 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
3431 Jinsong Ji <jji@us.ibm.com>
3433 * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
3434 comparison operators with vec_cmp* for compatibility due to
3435 unfortunate history; clean up formatting and use types more
3437 (_mm_sll_epi32): Likewise.
3438 (_mm_sll_epi64): Likewise.
3439 (_mm_srl_epi16): Likewise.
3440 (_mm_srl_epi32): Likewise.
3441 (_mm_srl_epi64): Likewise.
3443 2018-10-25 Bill Schmidt <wschmidt@linux.ibm.com>
3444 Jinsong Ji <jji@us.ibm.com>
3446 * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
3447 * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
3448 __vector __bool int. Use vec_cmpgt in preference to deprecated
3449 function vec_vcmpgtfp.
3450 (_mm_max_ps): Likewise.
3452 2018-10-25 Jeff Law <law@redhat.com>
3454 * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
3457 2018-10-25 Martin Sebor <msebor@redhat.com>
3459 * doc/extend.texi (aligned): Expand attribute description.
3460 (Alignment): Rename section. Discuss function arguments.
3462 2018-10-25 Jan Hubicka <jh@suse.cz>
3464 * ipa-devirt.c (main_odr_variant): Remove.
3465 (hash_odr_name, types_same_for_odr, types_odr_comparable,
3466 odr_name_hasher::equal, odr_subtypes_equivalent_p):
3467 Drop use of main_odr_variant.
3468 (add_type_duplicate): Silence confused warnings on integer types.
3469 (get_odr_type): Always look for main variant.
3470 (register_odr_type): Simplify.
3472 2018-10-25 Richard Biener <rguenther@suse.de>
3474 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
3475 Initialize ng to silence error with release checking bootstrap.
3477 2018-10-25 Richard Biener <rguenther@suse.de>
3479 * tree-if-conv.c: Include tree-ssa-sccvn.h.
3480 (tree_if_conversion): Run CSE on the if-converted loop body.
3482 2018-10-25 Ilya Leoshkevich <iii@linux.ibm.com>
3484 * config/s390/constraints.md (ZL): New constraint.
3485 * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
3487 * config/s390/s390.md (movdi_larl): Remove.
3488 (movdi_64): Add the LARL alternative.
3490 2018-10-25 Ilya Leoshkevich <iii@linux.ibm.com>
3493 * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
3494 (RTX_CODE_HWINT_P): New macro.
3495 (rtx_code_size): Use RTX_CODE_HWINT_P ().
3497 2018-10-25 Jan Hubicka <jh@suse.cz>
3499 * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
3502 2018-10-25 Richard Biener <rguenther@suse.de>
3504 PR tree-optimization/87665
3505 PR tree-optimization/87745
3506 * tree-vectorizer.h (get_earlier_stmt): Remove.
3507 (get_later_stmt): Pick up UID from the original non-pattern stmt.
3509 2018-10-25 Sam Tebbs <sam.tebbs@arm.com>
3511 * options.texi (Deprecated): Move list to Var section.
3513 2018-10-24 Bill Schmidt <wschmidt@linux.ibm.com>
3514 Jinsong Ji <jji@us.ibm.com>
3516 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
3517 __vector long to __vector long long.
3518 (_mm_cvtpd_ps): Likewise.
3519 (_mm_cvttpd_epi32): Likewise.
3520 (_mm_cvtpi32_pd): Likewise.
3521 (_mm_unpackhi_epi64): Likewise.
3522 (_mm_unpacklo_epi64): Likewise.
3524 2018-10-24 Segher Boessenkool <segher@kernel.crashing.org>
3526 PR rtl-optimization/87720
3527 * combine.c (make_more_copies): Skip if the dest is pc_rtx.
3529 2018-10-24 Alexandre Oliva <aoliva@redhat.com>
3531 * gimple-ssa-isolate-paths.c
3532 (find_implicit_erroneous_behavior): Do not change code if the
3533 pass is running for warnings only.
3534 (find_explicit_erroneous_behavior): Likewise.
3536 2018-10-24 Michael Meissner <meissner@linux.ibm.com>
3538 * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
3539 Define as rs6000_mangle_decl_assembler_name.
3540 (rs6000_mangle_decl_assembler_name): If the user switched from IBM
3541 long double to IEEE long double, switch the names of the long
3542 double built-in functions to be <func>f128 instead of <func>l.
3544 2018-10-24 Martin Sebor <msebor@redhat.com>
3546 * doc/extend.texi (nonnull): List no-argument form. Reference
3547 -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
3549 2018-10-24 Richard Biener <rguenther@suse.de>
3551 * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
3553 2018-10-24 Martin Liska <mliska@suse.cz>
3555 PR tree-optimization/84436
3556 * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
3558 (switch_conversion::contains_linear_function_p): New.
3559 (switch_conversion::build_one_array): Support linear
3560 transformation on input.
3561 * tree-switch-conversion.h (struct switch_conversion): Add
3562 contains_linear_function_p declaration.
3564 2018-10-24 Richard Biener <rguenther@suse.de>
3566 * varasm.c (const_hash_1): Return hash of ADDR_EXPR
3567 if its argument is CONSTANT_CLASS_P.
3569 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
3571 * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
3572 it is wrong for forward declarations.
3574 2018-10-24 Ilya Leoshkevich <iii@linux.ibm.com>
3576 * config/s390/s390.c (s390_check_qrst_address): Add the missing
3577 SYMBOL_REF_P () check.
3579 2018-10-24 Richard Biener <rguenther@suse.de>
3581 PR tree-optimization/87105
3582 * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
3583 dump classification.
3584 (vect_analyze_data_ref_accesses): Handle duplicate loads and
3585 stores by splitting the affected group after the fact.
3586 * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
3587 fail the SLP build because of size constraints.
3589 2018-10-24 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
3591 * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
3592 * configure: Regenerate.
3593 * config.in: Regenerate.
3594 * varasm.c (mergeable_string_section): Use readonly_data_section
3595 if linker doesn't support SHF_MERGE with alignment > 8.
3596 (mergeable_constant_section): Likewise.
3598 2018-10-24 Richard Biener <rguenther@suse.de>
3600 PR tree-optimization/84013
3601 * tree-ssa-structalias.c (struct msdi_data): New struct for
3602 marshalling data to walk_stmt_load_store_ops.
3603 (maybe_set_dependence_info): Refactor as callback for
3604 walk_stmt_load_store_ops.
3605 (compute_dependence_clique): Set restrict info on all stmt kinds.
3607 2018-10-24 Martin Liska <mliska@suse.cz>
3609 * cgraph.c (cgraph_node::dump):
3610 Remove reduntant dumps and make tp_first_run dump more compact.
3612 2018-10-24 Richard Biener <rguenther@suse.de>
3614 PR tree-optimization/87665
3615 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
3618 2018-10-12 Jeff Law <law@redhat.com>
3620 * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
3623 2018-10-23 Richard Biener <rguenther@suse.de>
3625 * tree-vrp.c (add_assert_info): Guard dump_printf with
3627 * gimple-ssa-evrp-analyze.c
3628 (evrp_range_analyzer::record_ranges_from_incoming_edge):
3629 Use value_range::ignore_equivs_equal_p.
3631 2018-10-23 Richard Biener <rguenther@suse.de>
3633 PR tree-optimization/87105
3634 PR tree-optimization/87608
3635 * passes.def (pass_all_early_optimizations): Add early phi-opt
3637 * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
3638 addition to debug stmts.
3639 (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
3640 and abs replacement early.
3641 * tree-cfg.c (gimple_empty_block_p): Likewise.
3643 2018-10-23 Richard Earnshaw <rearnsha@arm.com>
3646 * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
3647 specified to configure.
3648 (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
3650 2018-10-23 Richard Biener <rguenther@suse.de>
3652 PR tree-optimization/87700
3653 * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
3655 2018-10-23 Jakub Jelinek <jakub@redhat.com>
3658 * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
3659 second argument from __mmask16 to __mmask8.
3660 * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
3661 _mm_mask_packs_epi32): Likewise.
3662 * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
3664 (_mm512_mask_prefetch_i64scatter_pd): Likewise. Formatting fix.
3666 2018-10-23 Richard Biener <rguenther@suse.de>
3668 * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
3670 2018-10-23 Richard Biener <rguenther@suse.de>
3672 PR tree-optimization/86144
3673 * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
3674 over simd attribute.
3676 2018-10-23 Richard Biener <rguenther@suse.de>
3678 PR tree-optimization/87693
3679 * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
3680 the case we do not find the taken edge.
3682 2018-10-22 Bill Schmidt <wschmidt@linux.ibm.com>
3683 Jinsong Ji <jji@us.ibm.com>
3685 * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
3686 (_mm_store_pd): Use unaligned vector type for pointer cast.
3687 (_mm_maskmoveu_si128): Likewise.
3688 * config/rs6000/xmmintrin.h (__m128_u): New typedef.
3689 (_mm_store_ps): Use unaligned vector type for pointer cast.
3691 2018-10-22 Paul Koning <ni1d@arrl.net>
3693 * symtab.c (symtab_node::increase_alignment): Correct max
3696 2018-10-22 Yury Gribov <tetra2005@gmail.com>
3698 PR tree-optimization/87633
3699 * match.pd: Do not generate unordered integer comparisons.
3701 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
3703 PR rtl-optimization/87600
3704 * combine.c: Add include of expr.h.
3705 (cant_combine_insn_p): Do not combine moves from any hard non-fixed
3706 register to a pseudo.
3707 (make_more_copies): New function, add a copy to a new pseudo after
3708 the moves from hard registers into pseudos.
3709 (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
3710 later. Call make_more_copies.
3712 2018-10-22 Andrew Stubbs <ams@codesourcery.com>
3714 * lra-constraints.c (process_alt_operands): New local array,
3715 matching_early_clobber. Check matching_early_clobber before
3716 decrementing reject, and set matching_early_clobber after.
3718 2018-10-22 Segher Boessenkool <segher@kernel.crashing.org>
3721 * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
3722 call output_addr_const and hope for the best.
3724 2018-10-22 Richard Biener <rguenther@suse.de>
3726 * gimple-ssa-evrp-analyze.c
3727 (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
3728 smarter about what ranges to use.
3729 * tree-vrp.c (add_assert_info): Dump here.
3730 (register_edge_assert_for_2): Instead of here at multiple but
3733 * gcc.dg/tree-ssa/evrp12.c: New testcase.
3734 * gcc.dg/predict-6.c: Adjust.
3735 * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
3736 * gcc.dg/tree-ssa/vrp02.c: Likewise.
3737 * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
3739 2018-10-22 Steven Bosscher <steven@gcc.gnu.org>
3740 Richard Biener <rguenther@suse.de>
3742 * bitmap.h: Update data structure documentation, including a
3743 description of bitmap views as either linked-lists or splay trees.
3744 (struct bitmap_element_def): Update comments for splay tree bitmaps.
3745 (struct bitmap_head_def): Likewise.
3746 (bitmap_list_view, bitmap_tree_view): New prototypes.
3747 (bitmap_initialize_stat): Initialize a bitmap_head's indx and
3749 (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
3750 (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
3751 * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
3752 released bitmap element here.
3753 (bitmap_element_free): Remove.
3754 (bitmap_elt_clear_from): Work on splay tree bitmaps.
3755 (bitmap_list_link_element): Renamed from bitmap_element_link. Move
3756 this function similar ones such that linked-list bitmap implementation
3757 functions are grouped.
3758 (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
3759 and moved for grouping.
3760 (bitmap_list_insert_element_after): Renamed from
3761 bitmap_elt_insert_after, and moved for grouping.
3762 (bitmap_list_find_element): New function spliced from bitmap_find_bit.
3763 (bitmap_tree_link_left, bitmap_tree_link_right,
3764 bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
3765 bitmap_tree_link_element, bitmap_tree_unlink_element,
3766 bitmap_tree_find_element): New functions for splay-tree bitmap
3768 (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
3769 Renamed and moved, see above entries.
3770 (bitmap_tree_listify_from): New function to convert part of a splay
3771 tree bitmap to a linked-list bitmap.
3772 (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
3773 (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
3774 (bitmap_find_bit): Remove.
3775 (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
3776 bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
3777 Handle splay tree bitmaps.
3778 (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
3779 bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
3780 bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
3781 bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
3782 bitmap_intersect_compl_p, bitmap_ior_and_compl,
3783 bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
3784 bitmap_hash): Reject trying to act on splay tree bitmaps. Make
3785 corresponding changes to use linked-list specific bitmap_element
3786 manipulation functions as applicable for efficiency.
3787 (bitmap_tree_to_vec): New function.
3788 (debug_bitmap_elt_file): New function split out from ...
3789 (debug_bitmap_file): ... here. Handle splay tree bitmaps.
3790 (bitmap_print): Likewise.
3792 PR tree-optimization/63155
3793 * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
3795 * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
3798 2018-10-22 Martin Liska <mliska@suse.cz>
3800 PR tree-optimization/87686
3802 2018-08-29 Martin Liska <mliska@suse.cz>
3804 * tree-switch-conversion.c (switch_conversion::expand):
3805 Strenghten assumption about gswitch statements.
3807 2018-10-22 Martin Liska <mliska@suse.cz>
3809 * ipa-icf.c (sem_item::compare_attributes): Remove.
3810 (sem_item::compare_referenced_symbol_properties): Use
3811 attribute_list_equal instead.
3812 (sem_function::equals_wpa): Likewise.
3813 * ipa-icf.h: Remove compare_attributes.
3815 2018-10-22 Richard Biener <rguenther@suse.de>
3818 * mem-stats.h (mem_usage::operator==): Fix pasto.
3820 2018-10-22 Richard Biener <rguenther@suse.de>
3822 PR tree-optimization/87640
3823 * tree-vrp.c (set_value_range_with_overflow): Decompose
3825 (extract_range_from_binary_expr_1): Adjust.
3827 2018-10-22 Martin Jambor <mjambor@suse.cz>
3829 * tree-eh.h (stmt_could_throw_p): Add function parameter.
3830 (stmt_can_throw_external): Likewise.
3831 (stmt_can_throw_internal): Likewise.
3832 * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
3833 (lower_eh_constructs_2): Likewise.
3834 (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
3835 (stmt_can_throw_external): Likewise.
3836 (stmt_can_throw_internal): Likewise.
3837 (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
3838 (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
3839 (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
3840 (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
3841 (pass_lower_eh_dispatch::execute): Pass cfun to
3842 stmt_can_throw_external.
3843 (cleanup_empty_eh): Likewise.
3844 (verify_eh_edges): Pass cfun to stmt_could_throw_p.
3845 * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
3846 stmt_can_throw_external instead of pushing it to cfun.
3847 (symbol_table::create_edge): Likewise.
3848 * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
3849 stmt_can_throw_internal.
3850 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
3851 to stmt_could_throw_p.
3852 * gimple-ssa-store-merging.c (handled_load): Pass cfun to
3853 stmt_can_throw_internal.
3854 (pass_store_merging::execute): Likewise.
3855 * gimple-ssa-strength-reduction.c
3856 (find_candidates_dom_walker::before_dom_children): Pass cfun to
3858 * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
3859 stmt_can_throw_internal.
3860 * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
3861 to stmt_can_throw_external.
3862 (check_stmt): Pass cfun to stmt_could_throw_p.
3863 (check_stmt): Pass cfun to stmt_can_throw_external.
3864 (pass_nothrow::execute): Likewise.
3865 * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
3866 * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
3867 stmt_can_throw_internal.
3868 (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
3869 (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
3870 (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
3871 * tree-complex.c (expand_complex_libcall): Pass cfun to
3872 stmt_could_throw_p and to stmt_can_throw_internal.
3873 (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
3874 * tree-inline.c (copy_edges_for_bb): Likewise.
3875 (maybe_move_debug_stmts_to_successors): Likewise.
3876 * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
3878 * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
3879 * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
3880 * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
3881 stmt_can_throw_internal.
3882 * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
3883 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
3885 (mark_aliased_reaching_defs_necessary_1): Pass cfun to
3886 stmt_can_throw_internal.
3887 * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
3888 * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
3890 * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
3891 (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
3892 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3893 (convert_mult_to_fma_1): Likewise.
3894 (convert_to_divmod): Likewise.
3895 * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
3896 * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
3897 * tree-ssa-propagate.c
3898 (substitute_and_fold_dom_walker::before_dom_children): Likewise.
3899 * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
3900 (maybe_optimize_range_tests): Likewise.
3901 (linearize_expr_tree): Likewise.
3902 (reassociate_bb): Likewise.
3903 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3904 * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
3905 * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
3906 (handle_char_store): Likewise.
3907 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
3908 stmt_can_throw_internal.
3909 * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
3911 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3912 (vectorizable_call): Pass cfun to stmt_can_throw_internal.
3913 (vectorizable_simd_clone_call): Likewise.
3914 * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
3915 (gimple_stringop_fixed_value): Likewise.
3917 2018-10-22 Ilya Leoshkevich <iii@linux.ibm.com>
3919 * config/s390/s390.c (s390_loadrelative_operand_p): Accept
3920 literal pool references.
3921 (s390_check_qrst_address): Adapt to the new behavior of
3922 s390_loadrelative_operand_p ().
3924 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
3927 * config/i386/sse.md (*andnot<mode>3_bcst): New.
3929 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
3932 * config/i386/sse.md (*<code><mode>3_bcst): New.
3934 2018-10-22 H.J. Lu <hongjiu.lu@intel.com>
3937 * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
3938 V4DI, V16SI and V8DI.
3939 (*sub<mode>3<mask_name>_bcst): New.
3940 (*add<mode>3<mask_name>_bcst): Likewise.
3942 2018-10-21 Bill Schmidt <wschmidt@linux.ibm.com>
3943 Jinsong Ji <jji@us.ibm.com>
3945 * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
3946 __m64 with __vector unsigned long long for compatibility.
3947 (_mm_movemask_epi8): Likewise.
3948 * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
3949 (_mm_cvttps_pi32): Likewise.
3950 (_mm_cvtpi32_ps): Likewise.
3951 (_mm_cvtps_pi16): Likewise.
3952 (_mm_loadh_pi): Likewise.
3953 (_mm_storeh_pi): Likewise.
3954 (_mm_movehl_ps): Likewise.
3955 (_mm_movelh_ps): Likewise.
3956 (_mm_loadl_pi): Likewise.
3957 (_mm_storel_pi): Likewise.
3958 (_mm_movemask_ps): Likewise.
3959 (_mm_shuffle_pi16): Likewise.
3961 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
3964 * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
3965 __builtin_ia32_vfnmsubpd512_mask.
3966 (_mm512_mask_fnmsub_round_pd): Likewise.
3967 (_mm512_fnmsub_pd): Likewise.
3968 (_mm512_mask_fnmsub_pd): Likewise.
3969 (_mm512_maskz_fnmsub_round_pd): Use
3970 __builtin_ia32_vfnmsubpd512_maskz.
3971 (_mm512_maskz_fnmsub_pd): Likewise.
3972 (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
3973 (_mm512_mask_fnmsub_round_ps): Likewise.
3974 (_mm512_fnmsub_ps): Likewise.
3975 (_mm512_mask_fnmsub_ps): Likewise.
3976 (_mm512_maskz_fnmsub_round_ps): Use
3977 __builtin_ia32_vfnmsubps512_maskz.
3978 (_mm512_maskz_fnmsub_ps): Likewise.
3979 * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
3980 __builtin_ia32_vfnmsubpd256_mask.
3981 (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
3982 (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3983 (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
3984 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3985 (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3986 (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
3987 (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
3988 (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
3989 * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
3990 __builtin_ia32_vfnmsubpd.
3991 (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
3992 (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
3993 (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
3994 (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
3995 (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
3996 * config/i386/i386-builtin.def: Add
3997 __builtin_ia32_vfnmsubpd256_mask,
3998 __builtin_ia32_vfnmsubpd256_maskz,
3999 __builtin_ia32_vfnmsubpd128_mask,
4000 __builtin_ia32_vfnmsubpd128_maskz,
4001 __builtin_ia32_vfnmsubps256_mask,
4002 __builtin_ia32_vfnmsubps256_maskz,
4003 __builtin_ia32_vfnmsubps128_mask,
4004 __builtin_ia32_vfnmsubps128_maskz,
4005 __builtin_ia32_vfnmsubpd512_mask,
4006 __builtin_ia32_vfnmsubpd512_maskz,
4007 __builtin_ia32_vfnmsubps512_mask,
4008 __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
4009 __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
4010 __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
4011 __builtin_ia32_vfnmsubpd256.
4012 * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
4013 (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
4014 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
4016 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
4018 (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
4020 (fmai_vmfnmsub_<mode><round_name>): Likewise.
4022 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
4025 * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
4026 __builtin_ia32_vfnmaddpd512_mask.
4027 (_mm512_mask_fnmadd_round_pd): Likewise.
4028 (_mm512_fnmadd_pd): Likewise.
4029 (_mm512_mask_fnmadd_pd): Likewise.
4030 (_mm512_maskz_fnmadd_round_pd): Use
4031 __builtin_ia32_vfnmaddpd512_maskz.
4032 (_mm512_maskz_fnmadd_pd): Likewise.
4033 (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
4034 (_mm512_mask_fnmadd_round_ps): Likewise.
4035 (_mm512_fnmadd_ps): Likewise.
4036 (_mm512_mask_fnmadd_ps): Likewise.
4037 (_mm512_maskz_fnmadd_round_ps): Use
4038 __builtin_ia32_vfnmaddps512_maskz.
4039 (_mm512_maskz_fnmadd_ps): Likewise.
4040 * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
4041 __builtin_ia32_vfnmaddpd256_mask.
4042 (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
4043 (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
4044 (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
4045 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
4046 (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
4047 (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
4048 (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
4049 (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
4050 * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
4051 __builtin_ia32_vfnmaddpd.
4052 (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
4053 (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
4054 (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
4055 (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
4056 (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
4057 * config/i386/i386-builtin.def: Add
4058 __builtin_ia32_vfnmaddpd256_mask,
4059 __builtin_ia32_vfnmaddpd256_maskz,
4060 __builtin_ia32_vfnmaddpd128_mask,
4061 __builtin_ia32_vfnmaddpd128_maskz,
4062 __builtin_ia32_vfnmaddps256_mask,
4063 __builtin_ia32_vfnmaddps256_maskz,
4064 __builtin_ia32_vfnmaddps128_mask,
4065 __builtin_ia32_vfnmaddps128_maskz,
4066 __builtin_ia32_vfnmaddpd512_mask,
4067 __builtin_ia32_vfnmaddpd512_maskz,
4068 __builtin_ia32_vfnmaddps512_mask,
4069 __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
4070 __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
4071 __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
4072 __builtin_ia32_vfnmaddpd256.
4073 * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
4074 (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
4075 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
4077 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
4079 (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
4081 (fmai_vmfnmadd_<mode><round_name>): Likewise.
4083 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
4086 * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
4087 __builtin_ia32_vfmsubpd512_mask.
4088 (_mm512_mask_fmsub_round_pd): Likewise.
4089 (_mm512_fmsub_pd): Likewise.
4090 (_mm512_mask_fmsub_pd): Likewise.
4091 (_mm512_maskz_fmsub_round_pd): Use
4092 __builtin_ia32_vfmsubpd512_maskz.
4093 (_mm512_maskz_fmsub_pd): Likewise.
4094 (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
4095 (_mm512_mask_fmsub_round_ps): Likewise.
4096 (_mm512_fmsub_ps): Likewise.
4097 (_mm512_mask_fmsub_ps): Likewise.
4098 (_mm512_maskz_fmsub_round_ps): Use
4099 __builtin_ia32_vfmsubps512_maskz.
4100 (_mm512_maskz_fmsub_ps): Likewise.
4101 * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
4102 __builtin_ia32_vfmsubpd256_mask.
4103 (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
4104 (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
4105 (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
4106 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
4107 (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
4108 (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
4109 (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
4110 (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
4111 * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
4112 __builtin_ia32_vfmsubpd.
4113 (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
4114 (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
4115 (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
4116 (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
4117 (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
4118 * config/i386/i386-builtin.def: Add
4119 __builtin_ia32_vfmsubpd256_mask,
4120 __builtin_ia32_vfmsubpd256_maskz,
4121 __builtin_ia32_vfmsubpd128_mask,
4122 __builtin_ia32_vfmsubpd128_maskz,
4123 __builtin_ia32_vfmsubps256_mask,
4124 __builtin_ia32_vfmsubps256_maskz,
4125 __builtin_ia32_vfmsubps128_mask,
4126 __builtin_ia32_vfmsubps128_maskz,
4127 __builtin_ia32_vfmsubpd512_mask,
4128 __builtin_ia32_vfmsubpd512_maskz,
4129 __builtin_ia32_vfmsubps512_mask,
4130 __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
4131 __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
4132 __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
4133 __builtin_ia32_vfmsubpd256.
4134 * config/i386/sse.md (fma4i_fmsub_<mode>): New.
4135 (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
4136 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
4138 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
4140 (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
4142 (fmai_vmfmsub_<mode><round_name>): Likewise.
4144 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
4146 * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
4147 Remove plus. Renamed to ...
4148 (*sub<mode>3<mask_name>_bcst): This.
4149 (*add<mode>3<mask_name>_bcst_2): Renamede to ...
4150 (*add<mode>3<mask_name>_bcst): This.
4152 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
4155 * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
4157 2018-10-21 H.J. Lu <hongjiu.lu@intel.com>
4160 * i386/avx512vlintrin.h (_mm256_or_epi32): New.
4161 (_mm_or_epi32): Likewise.
4162 (_mm256_xor_epi32): Likewise.
4163 (_mm_xor_epi32): Likewise.
4164 (_mm256_or_epi64): Likewise.
4165 (_mm_or_epi64): Likewise.
4166 (_mm256_xor_epi64): Likewise.
4167 (_mm_xor_epi64): Likewise.
4169 2018-10-20 H.J. Lu <hongjiu.lu@intel.com>
4172 * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
4174 2018-10-20 Jakub Jelinek <jakub@redhat.com>
4177 * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
4179 2018-10-20 Andreas Schwab <schwab@linux-m68k.org>
4181 * doc/ux.texi: Move @section directly after @node.
4183 2018-10-19 Jakub Jelinek <jakub@redhat.com>
4187 * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
4188 depend closely nested inside of loop with ordered clause with
4191 2018-10-19 David Malcolm <dmalcolm@redhat.com>
4193 * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
4194 * doc/gccint.texi: Include ux.texi and use it in top-level menu.
4195 * doc/ux.texi: New file.
4197 2018-10-19 Segher Boessenkool <segher@kernel.crashing.org>
4199 * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
4200 be the first CR field allocated.
4202 2018-10-19 Richard Biener <rguenther@suse.de>
4205 * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
4206 TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
4208 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
4211 * config/i386/sse.md
4212 (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
4213 (*add<mode>3<mask_name>_bcst_2): Likewise.
4215 2018-10-19 H.J. Lu <hongjiu.lu@intel.com>
4217 * config/i386/sse.md
4218 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
4219 Replace nonimmediate_operand with register_operand.
4220 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
4222 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
4225 2018-10-19 Ilya Leoshkevich <iii@linux.ibm.com>
4227 PR rtl-optimization/87596
4228 * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
4229 lra_get_insn_recog_data () instead of lra_insn_recog_data[]
4230 for instructions in FROM..TO range.
4232 2018-10-19 Eric Botcazou <ebotcazou@adacore.com>
4234 * cfgexpand.c (expand_one_var): Use specific wording in error message
4235 for non-local frame variables.
4236 * stor-layout.c (layout_decl): Do not issue a warning for them.
4238 2018-10-19 Robin Dapp <rdapp@linux.ibm.com>
4240 * haifa-sched.c (priority): Add force_recompute parameter.
4241 (apply_replacement): Call priority () with force_recompute = true.
4242 (restore_pattern): Likewise.
4244 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
4246 * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
4247 HOST_BITS_PER_WIDE_INT.
4248 (test_vector_ops_duplicate): Likewise.
4250 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
4253 * config/i386/sse.md (VF_AVX512): New.
4254 (avx512bcst): Likewise.
4255 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
4257 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
4259 (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
4262 2018-10-18 Jonathan Wakely <jwakely@redhat.com>
4264 * doc/invoke.texi (-dumpversion): Improve grammar.
4265 (-dumpfullversion): Make more consistent with -dumpversion.
4267 2018-10-18 Uros Bizjak <ubizjak@gmail.com>
4269 * config/i386/i386.c (ix86_emit_fp_unordered_jump):
4270 Set JUMP_LABEL to the jump insn.
4271 (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
4272 Predict emitted jump and add label to jump insn.
4274 2018-10-18 David Malcolm <dmalcolm@redhat.com>
4276 PR tree-optimization/87562
4277 * input.c (get_substring_ranges_for_loc): Use
4278 LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
4279 getting the linemap for the endpoint. Verify that it's either
4280 in the same linemap as the start point's spelling location, or
4281 at least in the same file.
4283 2018-10-18 Richard Biener <rguenther@suse.de>
4285 * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
4286 feed width-specific load/store costs through ix86_vec_cost.
4287 * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
4288 (k8_cost): Likewise.
4289 (bdver_cost): Likewise.
4290 (znver1_cost): Likewise.
4291 (btver1_cost): Likewise.
4292 (btver2_cost): Likewise.
4294 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
4296 * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
4297 to simplify subreg of vec_merge.
4299 2018-10-18 Richard Biener <rguenther@suse.de>
4301 * config/i386/i386.c: Fix costing of vector FMA.
4303 2018-10-18 Richard Biener <rguenther@suse.de>
4305 * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
4307 (ix86_builtin_vectorization_cost): For vec_construct properly
4308 cost insertion into SSE regs.
4309 (...): Adjust calls to ix86_vec_cost.
4311 2018-10-18 Richard Biener <rguenther@suse.de>
4315 2018-02-07 Richard Biener <rguenther@suse.de>
4317 PR tree-optimization/84204
4318 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4321 2018-10-18 H.J. Lu <hongjiu.lu@intel.com>
4324 * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
4326 (test_vector_ops_duplicate): Add test for a scalar subreg of a
4327 VEC_MERGE of a VEC_DUPLICATE.
4329 2018-10-17 Joseph Myers <joseph@codesourcery.com>
4331 * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
4332 * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
4333 * doc/standards.texi (C Language): Document C2X.
4334 * dwarf2out.c (highest_c_language), config/rl78/rl78.c
4335 (rl78_option_override): Handle "GNU C2X" language name.
4337 2018-10-17 Joseph Myers <joseph@codesourcery.com>
4339 * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
4340 Document C17 as published in 2018.
4342 2018-10-17 Eric Botcazou <ebotcazou@adacore.com>
4345 * fold-const.c (fold_truth_andor_1): If the right side is not constant,
4346 bail out if both sides do not have the same storage order.
4348 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
4350 * bitmap.c (bitmap_head::dump): New.
4351 * bitmap.h (bitmap_head): Add dump().
4352 * gimple-ssa-evrp-analyze.c
4353 (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
4354 (evrp_range_analyzer::set_ssa_range_info): Same.
4355 (evrp_range_analyzer::record_ranges_from_phis): Same.
4356 (evrp_range_analyzer::record_ranges_from_stmt): Same.
4357 * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
4358 * gimple-ssa-sprintf.c (get_int_range): Same.
4359 (format_integer): Same.
4360 (sprintf_dom_walker::handle_gimple_call): Same.
4361 * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
4362 (ipcp_vr_lattice::top_p): Same.
4363 (ipcp_vr_lattice::bottom_p): Same.
4364 (ipcp_vr_lattice::set_to_bottom): Same.
4365 (ipa_vr_operation_and_type_effects): Same.
4366 (propagate_vr_across_jump_function): Same.
4367 (ipcp_store_vr_results): Same.
4368 * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
4369 (ipa_print_node_jump_functions_for_edge): Same.
4370 (ipa_get_value_range): Same.
4371 (ipa_compute_jump_functions_for_edge): Same.
4372 (ipa_write_jump_function): Same.
4373 * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
4374 * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
4376 * vr-values.c (set_value_range_to_nonnegative): Same.
4377 (set_value_range_to_truthvalue): Same.
4378 (vr_values::get_value_range): Same.
4379 (vr_values::set_defs_to_varying): Same.
4380 (vr_values::update_value_range): Same.
4381 (symbolic_range_based_on_p): Same.
4382 (vr_values::op_with_boolean_value_range_p): Same.
4383 (vr_values::extract_range_for_var_from_comparison_expr): Same.
4384 (vr_values::extract_range_from_ssa_name): Same.
4385 (vr_values::extract_range_from_binary_expr): Same.
4386 (vr_values::extract_range_from_unary_expr): Same.
4387 (vr_values::extract_range_from_cond_expr): Same.
4388 (vr_values::extract_range_from_comparison): Same.
4389 (vr_values::check_for_binary_op_overflow): Same.
4390 (vr_values::extract_range_basic): Same.
4391 (vr_values::extract_range_from_assignment): Same.
4392 (compare_ranges): Same.
4393 (compare_range_with_value): Same.
4394 (vr_values::adjust_range_with_scev): Same.
4395 (vrp_valueize): Same.
4396 (vrp_valueize_1): Same.
4397 (vr_values::get_vr_for_comparison): Same.
4398 (vr_values::compare_name_with_value): Same.
4399 (vr_values::compare_names): Same.
4400 (vr_values::vrp_evaluate_conditional): Same.
4401 (find_case_label_ranges): Same.
4402 (vr_values::vrp_visit_switch_stmt): Same.
4403 (vr_values::extract_range_from_phi_node): Same.
4404 (vr_values::simplify_div_or_mod_using_ranges): Same.
4405 (vr_values::simplify_bit_ops_using_ranges): Same.
4406 (test_for_singularity): Same.
4407 (range_fits_type_p): Same.
4408 (vr_values::simplify_cond_using_ranges_1): Same.
4409 (vr_values::simplify_switch_using_ranges): Same.
4410 (vr_values::simplify_float_conversion_using_ranges): Same.
4411 (vr_values::two_valued_val_range_p): Same.
4412 (vr_values::add_equivalence): Move to value_range::equiv_add.
4413 * vr-values.h (vr_values::add_equivalence): Remove.
4414 (VR_INITIALIZER): Remove.
4415 * tree-vrp.c (value_range::set): New.
4416 (value_range::equiv_add): New.
4417 (value_range::value_range): New.
4418 (value_range::deep_copy): New.
4419 (value_range::check): New.
4420 (value_range::equal_p): New.
4421 (value_range::ignore_equivs_equal_p): New.
4422 (value_range::operator==): New.
4423 (value_range::operator!=): New.
4424 (value_range::symbolic_p): New.
4425 (value_range::numeric_p): New.
4426 (value_range::set_undefined): New.
4427 (value_range::set_varying): New.
4428 (value_range::may_contain_p): New.
4429 (value_range::equiv_clear): New.
4430 (value_range::singleton_p): New.
4431 (value_range::intersect): New.
4432 (value_range::dump): New.
4433 (value_range::set_and_canonicalize): New.
4434 (set_value_range): Adjust for value_range API.
4435 (set_value_range_to_undefined): Same.
4436 (set_value_range_to_varying): Same.
4437 (set_and_canonicalize_value_range): Same.
4438 (set_value_range_to_nonnull): Same.
4439 (set_value_range_to_null): Same.
4440 (range_is_null): Same.
4441 (range_is_nonnull): Same.
4442 (range_int_cst_p): Same.
4443 (range_int_cst_singleton_p): Same.
4444 (symbolic_range_p): Same.
4445 (range_includes_zero_p): Same.
4446 (value_range_constant_singleton): Same.
4447 (vrp_set_zero_nonzero_bits): Same.
4448 (ranges_from_anti_range): Same.
4449 (extract_range_into_wide_ints): Same.
4450 (extract_range_from_multiplicative_op): Same.
4451 (set_value_range_with_overflow): Same.
4452 (extract_range_from_binary_expr_1): Same.
4453 (extract_range_from_unary_expr): Same.
4454 (dump_value_range): Same.
4455 (debug_value_range): Same.
4456 (vrp_prop::check_array_ref): Same.
4457 (vrp_prop::check_mem_ref): Same.
4458 (vrp_prop::vrp_initialize): Same.
4459 (vrp_prop::visit_stmt): Same.
4460 (intersect_ranges): Same.
4461 (vrp_prop::visit_phi): Same.
4462 (vrp_prop::vrp_finalize): Same.
4463 (determine_value_range_1): Same.
4464 (determine_value_range): Same.
4465 (vrp_intersect_ranges_1): Rename to...
4466 (vrp_intersect_1): this.
4467 (vrp_intersect_ranges): Rename to...
4468 (value_range::intersect_helper): ...this.
4469 (vrp_meet_1): Rename to...
4470 (value_range::union_helper): ...this.
4471 (vrp_meet): Rename to...
4472 (value_range::union_): ...this.
4473 (copy_value_range): Remove.
4474 * tree-vrp.h (struct value_range): Rewrite into a proper class.
4475 (value_range::vrtype): New.
4476 (value_range::type): New.
4477 (value_range::equiv): New.
4478 (value_range::min): New.
4479 (value_range::max): New.
4480 (value_range::varying_p): New.
4481 (value_range::undefined_p): New.
4482 (value_range::null_p): New.
4483 (value_range::equiv_add): New.
4484 (copy_value_range): Remove.
4486 2018-10-17 David Malcolm <dmalcolm@redhat.com>
4488 * Makefile.in (SELFTEST_TARGETS): New.
4489 (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
4490 (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
4491 (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
4493 (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
4494 (selftest-c++-gdb, selftest-c++-valgrind): Move to
4496 * configure: Regenerate.
4497 * configure.ac (selftest_languages): New.
4499 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
4501 * tree-vrp.c (extract_range_from_multiplicative_op): Remove
4502 overflow wraps argument.
4503 (extract_range_from_binary_expr_1): Do not pass overflow wraps to
4504 wide_int_range_multiplicative_op.
4505 * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
4506 overflow wraps argument.
4507 (wide_int_range_multiplicative_op): Same.
4508 (wide_int_range_lshift): Same.
4509 (wide_int_range_div): Same.
4510 * wide-int-range.h (wide_int_range_multiplicative_op): Same.
4511 (wide_int_range_lshift): Same.
4512 (wide_int_range_div): Same.
4514 2018-10-17 Aldy Hernandez <aldyh@redhat.com>
4516 * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
4517 use sign as argument.
4518 * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
4519 wide_int_range_shift_undefined_p.
4521 2018-10-16 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4523 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
4525 (@despeculate_copy<ALLI_TI:mode>): ... This.
4526 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
4529 2018-10-16 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
4531 * config.gcc: Obsolete *-*-solaris2.10*.
4532 * doc/install.texi (Specific, *-*-solaris2*): Document it.
4534 2018-10-12 Jeff Law <law@redhat.com>
4536 * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
4537 reg + sym +- const_int addressing modes.
4539 2018-10-15 David Malcolm <dmalcolm@redhat.com>
4541 * common.opt (fdiagnostics-minimum-margin-width=): New option.
4542 * diagnostic-show-locus.c (layout::layout): Apply the minimum
4544 (layout::start_annotation_line): Only print up to 3 of the
4545 margin character, to avoid touching the left-hand side.
4546 (selftest::test_diagnostic_show_locus_fixit_lines): Update for
4547 minimum margin width, as set by test_diagnostic_context's ctor.
4548 (selftest::test_fixit_insert_containing_newline): Likewise.
4549 (selftest::test_fixit_insert_containing_newline_2): Likewise.
4550 (selftest::test_line_numbers_multiline_range): Clear
4551 dc.min_margin_width.
4552 * diagnostic.c (diagnostic_initialize): Initialize
4554 * diagnostic.h (struct diagnostic_context): Add field
4556 * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
4557 * opts.c (common_handle_option): Handle
4558 OPT_fdiagnostics_minimum_margin_width_.
4559 * selftest-diagnostic.c
4560 (selftest::test_diagnostic_context::test_diagnostic_context):
4561 Initialize min_margin_width to 6.
4562 * toplev.c (general_init): Initialize global_dc->min_margin_width.
4564 2018-10-15 David Malcolm <dmalcolm@redhat.com>
4566 * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4567 Fix usage of "error_at_rich_loc" in the comment.
4569 2018-10-15 Renlin Li <renlin.li@arm.com>
4572 * tree-vectorizer.c (try_vectorize_loop_1): Don't use
4573 if-conversioned loop when it contains ifn with types not
4574 supported by backend.
4575 * internal-fn.c (expand_direct_optab_fn): Add an assert.
4576 (direct_internal_fn_supported_p): New helper function.
4577 * internal-fn.h (direct_internal_fn_supported_p): Declare.
4579 2018-10-15 Jakub Jelinek <jakub@redhat.com>
4582 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4583 Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4584 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4586 2018-10-15 Bin Cheng <bin.cheng@linux.alibaba.com>
4588 PR tree-optimization/87022
4589 * tree-loop-distribution.c (pg_add_dependence_edges): Check all
4590 bits in dist vector rather than the first one.
4592 2018-10-15 Richard Biener <rguenther@suse.de>
4595 * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
4596 (visit_loadstore): When a used restrict tag escaped verify that
4597 the points-to solution of "other" pointers do not include
4599 (compute_dependence_clique): If a used restrict tag escaped
4600 communicated that down to visit_loadstore.
4602 2018-10-15 Andreas Krebbel <krebbel@linux.ibm.com>
4604 * config/s390/s390.c (s390_expand_vec_init): Force vector element
4605 into reg if it isn't a general operand.
4607 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
4610 * config/i386/sse.md (*vec_dupv2di): Add register source to
4613 2018-10-14 H.J. Lu <hongjiu.lu@intel.com>
4616 * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4617 Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
4618 OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4619 OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4621 2018-10-13 Eric Botcazou <ebotcazou@adacore.com>
4623 * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
4624 (notice_args_size): Set it in the current trace if no insn that can
4625 throw internally has been seen yet.
4626 (connect_traces): When connecting args_size between traces, allow the
4627 incoming values not to match if there is an insn setting it before the
4628 first insn that can throw internally; in that case, force the creation
4629 of a CFI note on this latter insn.
4631 2018-10-13 Jonathan Wakely <jwakely@redhat.com>
4633 * opt-problem.h (opt_wrapper): Use template-argument-list when naming
4634 the base class, because using the injected-class-name was not clearly
4635 specified until DR 176.
4637 2018-10-12 Paul Koning <ni1d@arrl.net>
4639 * config/pdp11/pdp11.md (doloop_end): New expander.
4640 (doloop_end_insn): renamed from "doloop_end".
4641 (addqi3): New pattern.
4642 (subqi3): New pattern.
4643 * config/pdp11/predicates.md (incdec_operand): New predicate.
4645 2018-10-12 Yury Gribov <tetra2005@gmail.com>
4648 * real.c (format_helper::can_represent_integral_type_p): New function
4649 * real.h (format_helper::can_represent_integral_type_p): Ditto.
4650 * match.pd: New pattern.
4652 2018-10-12 Alexandre Oliva <oliva@adacore.com>
4654 * configure.ac: Introduce --enable-large-address-aware
4655 to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
4656 * doc/install.texi: Document it.
4657 * configure, config.in: Rebuilt.
4658 * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
4659 based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
4660 (LINK_SPEC): Insert it.
4661 * config/i386/mingw-w64.h: Likewise.
4663 * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
4665 2018-10-12 Peter Bergner <bergner@linux.ibm.com>
4667 PR rtl-optimization/87600
4668 * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
4670 2018-10-12 Paul Koning <ni1d@arrl.net>
4672 * doc/md.texi (doloop_end): Document that the pattern code may
4673 need to check operand mode.
4675 2018-10-12 Wilco Dijkstra <wdijkstr@arm.com>
4677 * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
4678 to zero-extend between int and floating-point registers.
4679 (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
4680 ldp into floating-point registers. Add type and arch attributes.
4681 (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
4682 Use f_loads for type attribute.
4684 2018-10-11 Martin Sebor <msebor@redhat.com>
4686 * doc/extend.texi (attribute packed): Correct typos.
4688 2018-10-11 Martin Sebor <msebor@redhat.com>
4690 * doc/extend.texi (attribute flatten): Mention interaction with
4693 2018-10-11 Jan Hubicka <hubicka@ucw.cz>
4696 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4697 Set new_decl virtual flag to zero.
4699 2018-10-11 Martin Sebor <msebor@redhat.com>
4702 * doc/extend.texi (attribute format_arg): Discuss using multiple
4703 attributes on a single function.
4705 2018-10-11 Giuliano Belinassi <giuliano.belinassi@usp.br>
4707 PR tree-optimization/86829
4708 * match.pd (sin (atan (x))): New simplification rules.
4709 (cos (atan (x))): Likewise.
4710 * real.c (build_sinatan_real): New function.
4711 * real.h (build_sinatan_real): Prototype.
4713 2018-10-11 Will Schmidt <will_schmidt@vnet.ibm.com>
4715 * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
4717 (fold_mergeeo_helper): New helper function.
4718 (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
4719 intrinsics. Correct some whitespace indentation issues.
4721 2018-10-11 Wilco Dijkstra <wdijkstr@arm.com>
4724 * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
4725 Use HOST_WIDE_INT_1U for shift.
4727 2018-10-11 Doug Rupp <rupp@adacore.com>
4728 Olivier Hainque <hainque@adacore.com>
4730 * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
4731 Pass --relax to the linker for RTPs.
4732 (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
4734 2018-10-11 Andrew Stubbs <ams@codesourcery.com>
4735 Jan Hubicka <jh@suse.cz>
4736 Martin Jambor <mjambor@suse.cz>
4738 * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
4739 the same elements are repeated rather than printing all of them.
4740 * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
4741 "repeated" elements.
4742 * read-rtl-function.c (test_loading_repeat): New function.
4743 (read_rtl_function_c_tests): Call test_loading_repeat.
4744 * rtl-tests.c (test_dumping_repeat): New function.
4745 (rtl_tests_c_tests): Call test_dumping_repeat.
4747 2018-10-11 Richard Biener <rguenther@suse.de>
4749 * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
4750 bdver?_cost): Unify to ...
4751 (bdver_memcpy, bdver_memset, bdver_cost): ... this.
4752 * config/i386/i386.c (processor_cost_table): Adjust.
4754 2018-10-10 Eric Botcazou <ebotcazou@adacore.com>
4757 * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
4758 the thunk when expanding to GIMPLE.
4760 2018-10-10 Bernd Edlinger <bernd.edlinger@hotmail.de>
4762 * varasm.c (mergeable_string_section): Don't try to move zero-length
4763 strings to the merge section.
4765 2018-10-10 Uros Bizjak <ubizjak@gmail.com>
4768 * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
4770 2018-10-10 Jakub Jelinek <jakub@redhat.com>
4773 * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
4774 to special_args set.
4776 2018-10-10 Richard Biener <rguenther@suse.de>
4778 * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
4779 reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
4780 reduc_plus_scal_v4sf): Merge into pattern reducing to half width
4781 and recursing and pattern terminating the recursion on SSE
4782 vector width using ix86_expand_reduc.
4783 (reduc_sminmax_scal_<mode>): Split into part reducing to half
4784 width and recursing and SSE2 vector variant doing the final
4785 reduction with ix86_expand_reduc.
4786 (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
4787 with terminating the recursion at AVX level, splitting that
4790 2018-10-09 David Malcolm <dmalcolm@redhat.com>
4792 * genmatch.c (error_cb): Rename to...
4793 (diagnostic_cb): ...this, converting int params to enums.
4794 (fatal_at): Update for renaming.
4795 (warning_at): Likewise.
4797 * input.c (selftest::ebcdic_execution_charset::apply):
4798 Update for renaming of...
4799 (selftest::ebcdic_execution_charset::on_error): ...this, renaming
4801 (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
4802 converting level and reason to enums.
4803 (class selftest::lexer_error_sink): Rename to...
4804 (class selftest::lexer_test_options): ...this, renaming field
4805 "m_errors" to "m_diagnostics".
4806 (selftest::lexer_test_options::apply): Update for renaming of...
4807 (selftest::lexer_test_options::on_error): ...this, renaming to...
4808 (selftest::lexer_test_options::on_diagnostic): ...this
4809 converting level and reason to enums.
4810 (selftest::test_lexer_string_locations_raw_string_unterminated):
4811 Update for renamings.
4812 * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
4815 2018-10-09 Paul A. Clarke <pc@us.ibm.com>
4817 * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
4818 * config/rs6000/pmmintrin.h: New file.
4820 2018-10-09 Eric Botcazou <ebotcazou@adacore.com>
4822 PR tree-optimization/86659
4823 * gimple-match.h (gimple_match_op constructors): Initialize reverse.
4825 2018-10-09 Richard Biener <rguenther@suse.de>
4827 PR tree-optimization/63155
4828 * tree-ssa-structalias.c: Include tree-ssa.h.
4829 (get_constraint_for_ssa_var): For undefs return nothing_id.
4830 (find_func_aliases): Cleanup PHI handling.
4832 2018-10-09 Richard Biener <rguenther@suse.de>
4834 * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
4837 2018-10-09 Martin Liska <mliska@suse.cz>
4839 * asan.c (asan_emit_stack_protection): If a stack variable
4840 is located in a same file as current function, then emit
4841 line info into variable definition string.
4843 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
4845 * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
4848 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
4850 * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
4853 2018-10-08 Eric Botcazou <ebotcazou@adacore.com>
4855 PR tree-optimization/86659
4856 * gimple-match.h (struct gimple_match_op): Add reverse field.
4857 (gimple_match_op::set_op): New overloaded method.
4858 * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
4859 the REF_REVERSE_STORAGE_ORDER flag on the value.
4860 (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
4861 REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
4863 2018-10-08 Richard Sandiford <richard.sandiford@arm.com>
4866 * gimple-ssa-backprop.c (backprop::intersect_uses): Use
4867 FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
4869 2018-10-08 H.J. Lu <hongjiu.lu@intel.com>
4872 * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
4873 Defined with __builtin_ia32_vfmaddsubpd512_mask.
4875 2018-10-08 Richard Biener <rguenther@suse.de>
4877 * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
4878 cost the same as AVX128 ones.
4880 2018-10-08 Paul Koning <ni1d@arrl.net>
4882 * config/pdp11/pdp11-protos.h (output_block_move): Remove.
4883 (expand_block_move): New function.
4884 * config/pdp11/pdp11.c (output_block_move): Remove.
4885 (expand_block_move): New function.
4886 * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
4887 * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
4888 (*movmemhi1): Remove.
4890 2018-10-08 Robin Dapp <rdapp@linux.ibm.com>
4892 * config/s390/2827.md: Increase latencies for some FP instructions.
4894 2018-10-08 Richard Biener <rguenther@suse.de>
4896 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4898 * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
4899 * tree-vectorizer.h (dump_stmt_cost): Adjust.
4900 (add_stmt_cost): Dump return value of the hook.
4902 2018-10-08 Richard Biener <rguenther@suse.de>
4904 PR tree-optimization/63155
4905 * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
4906 (ssa_propagation_engine::ssa_propagate): Remove redundant
4907 bitmap bit clearing.
4909 2018-10-05 Peter Bergner <bergner@linux.ibm.com>
4911 PR rtl-optimization/86939
4912 PR rtl-optimization/87479
4913 * ira.h (non_conflicting_reg_copy_p): New prototype.
4914 * ira-lives.c (ignore_reg_for_conflicts): New static variable.
4915 (make_hard_regno_dead): Don't add conflicts for register
4916 ignore_reg_for_conflicts.
4917 (make_object_dead): Likewise.
4918 (non_conflicting_reg_copy_p): New function.
4919 (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
4920 Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
4921 * lra-lives.c (ignore_reg_for_conflicts): New static variable.
4922 (make_hard_regno_dead): Don't add conflicts for register
4923 ignore_reg_for_conflicts. Remove special conflict handling of
4924 REAL_PIC_OFFSET_TABLE_REGNUM. Remove now unused argument
4925 check_pic_pseudo_p and update callers.
4926 (mark_pseudo_dead): Don't add conflicts for register
4927 ignore_reg_for_conflicts.
4928 (process_bb_lives): Set ignore_reg_for_conflicts for copies.
4930 2018-10-05 Andrew Waterman <andrew@sifive.com>
4931 Jim Wilson <jimw@sifive.com>
4933 * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
4934 Add define_expand. Add ! HONOR_SNANS check to current pattern. Add
4935 new pattern using HONOR_SNANS that emits one extra instruction.
4937 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
4939 * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
4940 patterns): Merge SI and DI patterns to a GPR pattern.
4941 (unnamed define_insn and define_split for record form of that): Merge
4942 to a single define_insn_and_split pattern.
4944 2018-10-05 David Malcolm <dmalcolm@redhat.com>
4947 * input.c (expand_location_to_spelling_point): Add param "aspect"
4948 and use rather than hardcoding LOCATION_ASPECT_CARET.
4949 (get_substring_ranges_for_loc): Handle the case of a single token
4950 within a macro expansion.
4951 * input.h (expand_location_to_spelling_point): Add "aspect" param,
4952 defaulting to LOCATION_ASPECT_CARET.
4954 2018-10-05 Paul Koning <ni1d@arrl.net>
4956 * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
4957 (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
4958 (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4959 (pdp11_guard_type): New function.
4961 2018-10-05 Paul Koning <ni1d@arrl.net>
4963 * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
4964 * config/pdp11/pdp11.opt (mfloat32): Remove.
4966 * doc/invoke.texi (pdp11 -mfloat32): Remove:
4967 (pdp11 -mfloat64): Remove.
4969 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
4971 * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
4972 (*cmp<mode>_cc_i387): Ditto.
4973 (*cmpu<mode>_cc_i387): Ditto.
4974 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4975 * config/i386/i386.c (ix86_expand_fp_compare): Remove
4977 <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
4978 Emit x86_sahf_1 pattern.
4979 (ix86_expand_compare): Update call to ix86_expand_fp_compare.
4980 (ix86_expand_carry_flag_compare): Ditto.
4982 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
4984 * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
4985 to reg_or_0_operand. Add "C" constraint.
4986 (*cmpxf_cc_i387): Ditto.
4987 (*cmp<mode>_i387): Change operand 2 predicate
4988 to nonimm_or_0_operand. Add "C" constraint.
4989 (*cmp<mode>_cc_i387): Ditto.
4990 (*cmp<mode>_0_i387): Remove insn pattern.
4991 (*cmp<mode>_0_cc_i387): Ditto.
4993 2018-10-05 Uros Bizjak <ubizjak@gmail.com>
4995 * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
4996 * config/i386/predicates.md (nonimm_or_0_operand): Rename
4997 from vector_move_operand. Update all uses.
4999 2018-10-05 Martin Sebor <msebor@redhat.com>
5001 PR tree-optimization/87490
5002 * builtins.c (expand_builtin_strnlen): Handle a null data.decl
5005 2018-10-05 Richard Biener <rguenther@suse.de>
5007 PR tree-optimization/63155
5008 * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
5009 vertical space in dumpfiles.
5010 * tree-ssa-propagate.h
5011 (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
5012 * tree-ssa-propagate.c (cfg_blocks_back): New global.
5013 (ssa_edge_worklist_back): Likewise.
5014 (curr_order): Likewise.
5015 (cfg_blocks_get): Remove abstraction.
5016 (cfg_blocks_add): Likewise.
5017 (cfg_blocks_empty_p): Likewise.
5018 (add_ssa_edge): Add to current or next worklist based on
5020 (add_control_edge): Likewise.
5021 (ssa_propagation_engine::process_ssa_edge_worklist): Fold
5023 (ssa_propagation_engine::ssa_propagate): ... here. Unify
5024 iteration from CFG and SSA edge worklist so we process
5025 everything in RPO order, prioritizing forward progress
5027 (ssa_prop_init): Allocate new worklists, do not dump
5029 (ssa_prop_fini): Free new worklists.
5031 2018-10-05 Richard Biener <rguenther@suse.de>
5033 * tree-core.h (tree_block::abstract_flag): Remove.
5034 (tree_block::block_num): Make full 32bits.
5035 * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
5036 * tree.h (BLOCK_ABSTRACT): Remove.
5037 * dwarf2out.c (gen_lexical_block_die): Remove dead code
5038 resulting from BLOCK_ABSTRACT being always false.
5039 (gen_inlined_subroutine_die): Likewise.
5040 (gen_block_die): Likewise.
5041 * tree.c (block_ultimate_origin): Likewise.
5042 * tree-pretty-print.c (dump_block_node): Remove code dealing
5043 with BLOCK_ABSTRACT.
5044 * tree-ssa-live.c (dump_scope_block): Likewise.
5045 * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
5046 * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
5048 2018-10-05 Richard Biener <rguenther@suse.de>
5050 * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
5051 is asked for initialize mode to the component mode of the
5054 2018-10-05 H.J. Lu <hongjiu.lu@intel.com>
5057 * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
5058 assembler for -mavx.
5059 * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
5061 2018-10-05 Segher Boessenkool <segher@kernel.crashing.org>
5064 * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
5066 * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
5067 to be DImode. When using mffscrn, force the operand to a register.
5069 2018-10-04 Uros Bizjak <ubizjak@gmail.com>
5071 * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
5072 from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
5073 X87MODEF mode iterator.
5074 (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
5075 *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
5076 X87MODEF mode iterator.
5078 2018-10-04 Vinay Kumar <vinay.kumar@blackfigtech.com>
5080 * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
5081 -Wno-prio-ctor-dtor.
5083 2018-10-04 David Malcolm <dmalcolm@redhat.com>
5085 * Makefile.in (OBJS): Add opt-problem.o.
5086 * dump-context.h: Include "selftest.h.
5087 (selftest::temp_dump_context): New forward decl.
5088 (class dump_context): Make friend of class
5089 selftest::temp_dump_context.
5090 (dump_context::dump_loc_immediate): New decl.
5091 (class dump_pretty_printer): Move here from dumpfile.c.
5092 (class temp_dump_context): Move to namespace selftest.
5093 (temp_dump_context::temp_dump_context): Add param
5094 "forcibly_enable_dumping".
5095 (selftest::verify_dumped_text):
5096 (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
5097 (selftest::verify_item):
5098 (ASSERT_IS_TEXT): Move here from dumpfile.c.
5099 (ASSERT_IS_TREE): Likewise.
5100 (ASSERT_IS_GIMPLE): Likewise.
5101 * dumpfile.c (dump_context::dump_loc): Move immediate dumping
5103 (dump_context::dump_loc_immediate): ...this new function.
5104 (class dump_pretty_printer): Move to dump-context.h.
5105 (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
5106 (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
5107 (temp_dump_context::temp_dump_context): Move to "selftest"
5108 namespace. Add param "forcibly_enable_dumping", and use it to
5109 conditionalize the use of m_pp;
5110 (selftest::verify_dumped_text): Make non-static.
5111 (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
5112 (selftest::verify_item): Make non-static.
5113 (ASSERT_IS_TEXT): Move to dump-context.h.
5114 (ASSERT_IS_TREE): Likewise.
5115 (ASSERT_IS_GIMPLE): Likewise.
5116 (selftest::test_capture_of_dump_calls): Pass "true" for new
5117 param of temp_dump_context.
5118 * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
5119 it to MSG_ALL_PRIORITIES. Update values of TDF_COMPARE_DEBUG and
5121 * opt-problem.cc: New file.
5122 * opt-problem.h: New file.
5123 * optinfo-emit-json.cc
5124 (selftest::test_building_json_from_dump_calls): Pass "true" for
5125 new param of temp_dump_context.
5126 * optinfo.cc (optinfo_kind_to_dump_flag): New function.
5127 (optinfo::emit_for_opt_problem): New function.
5128 (optinfo::emit): Clarity which emit_item is used.
5129 * optinfo.h (optinfo::get_dump_location): New accessor.
5130 (optinfo::emit_for_opt_problem): New decl.
5131 (optinfo::emit): Make const.
5132 * selftest-run-tests.c (selftest::run_tests): Call
5133 selftest::opt_problem_cc_tests.
5134 * selftest.h (selftest::opt_problem_cc_tests): New decl.
5135 * tree-data-ref.c (dr_analyze_innermost): Convert return type from
5136 bool to opt_result, converting fprintf messages to
5137 opt_result::failure_at calls. Add "stmt" param for use by the
5139 (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
5140 (runtime_alias_check_p): Convert return type from bool to
5141 opt_result, converting dump_printf calls to
5142 opt_result::failure_at, using the statement DDR_A for their
5144 (find_data_references_in_stmt): Convert return type from bool to
5145 opt_result, converting "return false" to opt_result::failure_at
5147 * tree-data-ref.h: Include "opt-problem.h".
5148 (dr_analyze_innermost): Convert return type from bool to opt_result,
5149 and add a const gimple * param.
5150 (find_data_references_in_stmt): Convert return type from bool to
5152 (runtime_alias_check_p): Likewise.
5153 * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
5154 dr_analyze_innermost.
5155 * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
5156 Convert return type from bool to opt_result, adding a message for
5157 the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
5158 (vect_analyze_data_ref_dependence): Convert return type from bool
5159 to opt_result. Change sense of return type from "false"
5160 effectively meaning "no problems" to "false" meaning a problem,
5161 so that "return false" becomes "return opt_result::success".
5162 Convert "return true" calls to opt_result::failure_at, using
5163 the location of statement A rather than vect_location.
5164 (vect_analyze_data_ref_dependences): Convert return type from bool
5166 (verify_data_ref_alignment): Likewise, converting dump_printf_loc
5167 calls to opt_result::failure_at, using the stmt location rather
5169 (vect_verify_datarefs_alignment): Convert return type from bool
5171 (vect_enhance_data_refs_alignment): Likewise. Split local "stat"
5172 into multiple more-tightly-scoped copies.
5173 (vect_analyze_data_refs_alignment): Convert return type from bool
5175 (vect_analyze_data_ref_accesses): Likewise, converting a
5176 "return false" to a "return opt_result::failure_at", adding a
5178 (vect_prune_runtime_alias_test_list): Convert return type from
5179 bool to opt_result, converting dump_printf_loc to
5180 opt_result::failure_at. Add a %G to show the pertinent statement,
5181 and use the stmt's location rather than vect_location.
5182 (vect_find_stmt_data_reference): Convert return type from
5183 bool to opt_result, converting dump_printf_loc to
5184 opt_result::failure_at, using stmt's location.
5185 (vect_analyze_data_refs): Convert return type from bool to
5186 opt_result. Convert "return false" to "return
5187 opt_result::failure_at", adding messages as needed.
5188 * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
5189 type from bool to opt_result.
5190 (vect_determine_vf_for_stmt): Likewise.
5191 (vect_determine_vectorization_factor): Likewise, converting
5192 dump_printf_loc to opt_result::failure_at, using location of phi
5193 rather than vect_location.
5194 (vect_analyze_loop_form_1): Convert return type from bool to
5195 opt_result, converting dump_printf_loc calls, retaining the use of
5197 (vect_analyze_loop_form): Convert return type from loop_vec_info
5198 to opt_loop_vec_info.
5199 (vect_analyze_loop_operations): Convert return type from bool to
5200 opt_result, converting dump_printf_loc calls, using the location
5201 of phi/stmt rather than vect_location where available. Convert
5202 various "return false" to "return opt_result::failure_at" with
5203 "unsupported phi" messages.
5204 (vect_get_datarefs_in_loop): Convert return type from bool to
5205 opt_result. Add a message for the
5206 PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
5207 (vect_analyze_loop_2): Convert return type from bool to
5208 opt_result. Ensure "ok" is set to a opt_result::failure_at before
5209 each "goto again;", adding new messages where needed.
5210 Add "unsupported grouped {store|load}" messages.
5211 (vect_analyze_loop): Convert return type from loop_vec_info to
5213 * tree-vect-slp.c (vect_analyze_slp): Convert return type from
5215 * tree-vect-stmts.c (process_use): Likewise, converting
5216 dump_printf_loc call and using stmt location, rather than
5218 (vect_mark_stmts_to_be_vectorized): Likeise.
5219 (vect_analyze_stmt): Likewise, adding a %G.
5220 (vect_get_vector_types_for_stmt): Convert return type from bool to
5221 opt_result, converting dump_printf_loc calls and using stmt
5222 location, rather than vect_location.
5223 (vect_get_mask_type_for_stmt): Convert return type from tree to
5224 opt_tree, converting dump_printf_loc calls and using stmt location.
5225 * tree-vectorizer.c: Include "opt-problem.h.
5226 (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
5227 MSG_PRIORITY_INTERNALS. Convert local "loop_vinfo" from
5228 loop_vec_info to opt_loop_vec_info. If if fails, and dumping is
5229 enabled, use it to report at the top level "couldn't vectorize
5230 loop" followed by the problem.
5231 * tree-vectorizer.h (opt_loop_vec_info): New typedef.
5232 (vect_mark_stmts_to_be_vectorized): Convert return type from bool
5234 (vect_analyze_stmt): Likewise.
5235 (vect_get_vector_types_for_stmt): Likewise.
5236 (tree vect_get_mask_type_for_stmt): Likewise.
5237 (vect_analyze_data_ref_dependences): Likewise.
5238 (vect_enhance_data_refs_alignment): Likewise.
5239 (vect_analyze_data_refs_alignment): Likewise.
5240 (vect_verify_datarefs_alignment): Likewise.
5241 (vect_analyze_data_ref_accesses): Likewise.
5242 (vect_prune_runtime_alias_test_list): Likewise.
5243 (vect_find_stmt_data_reference): Likewise.
5244 (vect_analyze_data_refs): Likewise.
5245 (vect_analyze_loop): Convert return type from loop_vec_info to
5247 (vect_analyze_loop_form): Likewise.
5248 (vect_analyze_slp): Convert return type from bool to opt_result.
5250 2018-10-04 David Malcolm <dmalcolm@redhat.com>
5252 * doc/invoke.texi (-fopt-info): Document new "internals"
5254 * dump-context.h (dump_context::apply_dump_filter_p): New decl.
5255 * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
5257 (optinfo_verbosity_options): Add "internals".
5258 (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
5259 (dump_context::apply_dump_filter_p): New member function.
5260 (dump_context::dump_loc): Use apply_dump_filter_p rather than
5261 explicitly masking the dump_kind.
5262 (dump_context::begin_scope): Increment the scope depth first. Use
5263 apply_dump_filter_p rather than explicitly masking the dump_kind.
5264 (dump_context::emit_item): Use apply_dump_filter_p rather than
5265 explicitly masking the dump_kind.
5266 (dump_dec): Likewise.
5267 (dump_hex): Likewise.
5268 (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
5269 (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
5270 (opt_info_switch_p): Update handling of default
5271 MSG_OPTIMIZED_LOCATIONS to cope with default of
5272 MSG_PRIORITY_USER_FACING.
5273 (dump_basic_block): Use apply_dump_filter_p rather than explicitly
5274 masking the dump_kind.
5275 (selftest::test_capture_of_dump_calls): Update test_dump_context
5276 instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
5277 than MSG_ALL. Generalize scope test to be run at all four
5278 combinations of with/without MSG_PRIORITY_USER_FACING and
5279 MSG_PRIORITY_INTERNALS, adding examples of explicit priority
5280 for each of the two values.
5281 * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
5282 Rename MSG_ALL to MSG_ALL_KINDS. Add MSG_PRIORITY_USER_FACING,
5283 MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
5284 values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
5285 (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
5286 with MSG_PRIORITY_*.
5287 * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
5288 dump messages as MSG_PRIORITY_USER_FACING.
5289 * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
5290 about the interaction with MSG_PRIORITY_*.
5292 2018-10-04 Bernd Edlinger <bernd.edlinger@hotmail.de>
5294 * varasm.c (output_constant): Add new parameter merge_strings.
5295 Make strings properly zero terminated in merge string sections.
5296 (mergeable_string_section): Don't fail if the last char is non-zero.
5297 (assemble_variable_contents): Handle merge string sections.
5298 (assemble_variable): Likewise.
5299 (assemble_constant_contents): Likewise.
5300 (output_constant_def_contents): Likewise.
5301 (output_constructor_array_range,
5302 output_constructor_regular_field): Adjust call to output_constant.
5303 (output_object_block): Adjust call to assemble_constant_contents
5304 and assemble_variable_contents.
5306 2018-10-04 Martin Liska <mliska@suse.cz>
5309 * cgraphunit.c (process_function_and_variable_attributes):
5310 Warn about a function with alias attribute and a body.
5312 2018-10-04 Martin Liska <mliska@suse.cz>
5315 * multiple_target.c (redirect_to_specific_clone): New function.
5316 (ipa_target_clone): Use it.
5317 * tree-inline.c: Fix comment.
5319 2018-10-04 David Malcolm <dmalcolm@redhat.com>
5321 * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
5323 (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
5324 (gcc::dump_manager::register_pass): New member function, adapted
5325 from loop body in gcc::pass_manager::register_pass, adding a
5326 call to update_dfi_for_opt_info.
5327 (gcc::dump_manager::opt_info_enable_passes): Store the
5328 -fopt-info options into the new fields. Move the loop
5330 (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
5332 * dumpfile.h (struct opt_pass): New forward decl.
5333 (gcc::dump_manager::register_pass): New decl.
5334 (gcc::dump_manager::update_dfi_for_opt_info): New decl.
5335 (class gcc::dump_manager): Add fields "m_optgroup_flags",
5336 "m_optinfo_flags", and "m_optinfo_filename".
5337 * passes.c (gcc::pass_manager::register_pass): Move all of the
5338 dump-handling code to gcc::dump_manager::register_pass.
5340 2018-10-04 Peter Bergner <bergner@linux.ibm.com>
5342 PR rtl-optimization/87466
5343 * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
5344 * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
5345 * doc/tm.texi: Regenerate.
5346 * ira-lives.c (process_bb_node_lives): Use the new target hook.
5347 * lra-lives.c (process_bb_lives): Likewise.
5348 * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
5351 2018-10-04 Tamar Christina <tamar.christina@arm.com>
5353 * params.c (add_params): Fix initialization.
5355 2018-10-04 Martin Liska <mliska@suse.cz>
5357 PR gcov-profile/84107
5358 * tree-profile.c (init_ic_make_global_vars):
5359 Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
5360 Come up with new ic_tuple* variables. Emit
5361 __gcov_indirect_call{,_topn} variables.
5362 (gimple_gen_ic_profiler): Access the variable
5364 (gimple_gen_ic_func_profiler): Access
5365 __gcov_indirect_call.callee field.
5366 (gimple_init_gcov_profiler): Use ptr_type_node.
5367 * value-prof.c (gimple_ic): Use ptr_type_node.
5369 2018-10-04 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5371 PR tree-optimization/85787
5372 * ipa-pure-const.c (malloc_candidate_p_1): Move most of
5373 malloc_candidate_p into this function and add support for
5374 detecting multiple phis.
5375 (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
5377 2018-10-04 Martin Liska <mliska@suse.cz>
5380 * ipa-inline.c (inline_to_all_callers_1):
5381 Call ultimate_alias_target for node being inlined.
5383 2018-10-03 Jeff Law <law@redhat.com>
5385 * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
5387 * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
5388 * tree.h (get_typenode_from_name): Prototype.
5390 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
5392 * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
5393 Change operand 2 predicate to nonimmediate_operand.
5394 (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
5396 2018-10-03 Martin Sebor <msebor@redhat.com>
5397 Jeff Law <law@redhat.com>
5399 * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
5401 (get_string_length): Detect unterminated arrays.
5402 (format_string): Same.
5403 (format_directive): Warn about unterminated arrays.
5404 (handle_gimple_call): Mark statements with no_warning as needed.
5406 2018-10-03 Jim Wilson <jimw@sifive.com>
5408 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
5409 also define __riscv_abi_rve. Delete trailing white space.
5411 2018-10-03 Paul Koning <ni1d@arrl.net>
5413 Enable LRA register allocator for PDP11.
5414 * config/pdp11/constraints.md (Q): Use define_memory_constraint.
5417 * config/pdp11/pdp11.c (pdp11_lra_p): New function.
5418 * config/pdp11/pdp11.opt (-mlra): New option.
5419 * doc/invoke.texi (PDP-11 Options): Document -mlra.
5421 2018-10-03 Uros Bizjak <ubizjak@gmail.com>
5423 * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
5424 (*<absneg:code>extend<mode>xf2): Ditto.
5426 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
5428 PR tree-optimization/87415
5429 * tree-vrp.c (set_value_range_with_overflow): Special case one bit
5432 2018-10-02 Jeff Law <law@redhat.com>
5434 * gimple-fold.c (get_range_strlen): Only set *nonstr when
5435 an unterminated string is discovered. Bubble up range
5436 even for unterminated strings.
5437 (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
5438 indicates the string was not terminated via NONSTR.
5440 2018-10-03 Aldy Hernandez <aldyh@redhat.com>
5442 * tree-vrp.c (extract_range_from_unary_expr): Special case all
5443 pointer conversions.
5444 Do not do anything special for anti-ranges.
5446 2018-10-03 Jérôme Lambourg <lambourg@adacore.com>
5448 * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
5449 DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
5451 2018-10-03 Martin Liska <mliska@suse.cz>
5453 PR gcov-profile/86109
5454 * coverage.c (coverage_begin_function): Do not
5455 mark lambdas as artificial.
5456 * tree-core.h (struct GTY): Remove tm_clone_flag
5457 and introduce new lambda_function.
5458 * tree.h (DECL_LAMBDA_FUNCTION): New macro.
5460 2018-10-02 Aaron Sawdey <acsawdey@linux.ibm.com>
5463 * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
5464 P8_VECTOR and VSX are enabled.
5466 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
5468 * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
5469 0x3907 as CPU model number.
5471 2018-10-02 Andreas Krebbel <krebbel@linux.ibm.com>
5473 * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
5474 * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
5475 PF_Z14. Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
5476 TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
5477 TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
5478 * config/s390/s390.md: Likewise. Rename also the cpu attribute
5479 value from arch12 to z14.
5481 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
5483 * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
5485 (isinf<mode>2): Ditto.
5487 2018-10-02 Uros Bizjak <ubizjak@gmail.com>
5489 * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
5490 before emitting fxam. Perform calculations in XFmode.
5492 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
5494 * match.pd (((X /[ex] A) +- B) * A): New transformation.
5496 2018-10-02 Marc Glisse <marc.glisse@inria.fr>
5499 * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
5500 * tree.c (signed_or_unsigned_type_for): Handle complex.
5502 2018-10-02 Jeff Law <law@redhat.com>
5504 * gimple-fold.c (get_range_strlen): Remove dead code.
5506 2018-10-02 Martin Sebor <msebor@redhat.com>
5507 Jeff Law <law@redhat.com>
5509 * builtins.c (unterminated_array): Add new arguments.
5510 If argument is not terminated, bubble up size and exact
5512 (expand_builtin_strnlen): Detect, avoid expanding
5513 and diagnose unterminated arrays.
5514 (c_strlen): Fill in offset of start of unterminated strings.
5515 * builtins.h (unterminated_array): Update prototype.
5517 2018-10-02 Richard Biener <rguenther@suse.de>
5519 * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
5520 of haddv4df, first reduce to SSE width and exploit the fact
5521 that we only need element zero with the reduction result.
5522 (reduc_plus_scal_v2df): Likewise.
5524 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
5526 * dojump.h (do_jump): Delete.
5527 (do_jump_1): Likewise.
5528 (split_comparison): Move around.
5529 * dojump.c (do_jump): Make static.
5530 (do_jump_1): Likewise.
5531 (jumpifnot): Move around.
5532 (jumpifnot_1): Likewise.
5534 (jumpif_1): Likewise.
5535 * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
5537 2018-10-02 Eric Botcazou <ebotcazou@adacore.com>
5539 * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
5540 insns in the delay slot and add_insn_after for the jump insn.
5542 2018-10-02 Richard Biener <rguenther@suse.de>
5544 * tree-inline.c (expand_call_inline): Use the location of
5545 the callee declaration for the inline-entry marker.
5546 * final.c (notice_source_line): Remove special-casing of
5547 NOTE_INSN_INLINE_ENTRY.
5549 2018-10-01 Carl Love <cel@us.ibm.com>
5552 * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
5553 (__builtin_mtfsb0): New.
5554 (__builtin_mtfsb1): New.
5555 ( __builtin_set_fpscr_rn): New.
5556 (__builtin_set_fpscr_drn): New.
5557 * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
5558 (rs6000_expand_set_fpscr_rn_builtin): Add.
5559 (rs6000_expand_set_fpscr_drn_builtin): Add.
5560 (rs6000_expand_builtin): Add case statement entries for
5561 RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
5562 RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
5563 RS6000_BUILTIN_MFFSL.
5564 (rs6000_init_builtins): Add ftype initialization and def_builtin
5565 calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
5566 __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
5567 * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
5568 rs6000_mffscdrn): Add define_insn.
5569 (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
5570 * doc/extend.texi: Add documentation for the builtins.
5572 2018-10-01 Richard Biener <rguenther@suse.de>
5574 PR tree-optimization/87465
5575 * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
5576 causing branch miscounts.
5578 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5580 * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
5581 aarch64_option_default_param): New.
5582 (params.h): Include.
5583 (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
5584 * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
5585 stack-clash protection validation code.
5587 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5589 * params.c (validate_param): New.
5590 (add_params): Use it.
5591 (set_param_value): Refactor param validation into validate_param.
5592 (diagnostic.h): Include.
5593 * diagnostic.h (diagnostic_ready_p): New.
5595 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5597 * params.c (set_param_value):
5598 Add index of parameter being validated.
5599 * common/common-target.def (option_validate_param): New.
5600 * common/common-targhooks.h (default_option_validate_param): New.
5601 * common/common-targhooks.c (default_option_validate_param): New.
5602 * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
5603 * doc/tm.texi: Regenerate.
5605 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5608 * config/aarch64/aarch64.c (aarch64_override_options_internal):
5609 Add validation for stack-clash parameters and set defaults.
5611 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5614 * configure.ac: Add stack-clash-protection-guard-size.
5615 * doc/install.texi: Document it.
5616 * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
5617 * params.def: Update comment for guard-size.
5618 (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
5619 PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
5620 * configure: Regenerate.
5622 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5625 * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
5626 STACK_DYNAMIC_OFFSET): New.
5627 * config/aarch64/aarch64.c (aarch64_layout_frame):
5628 Update outgoing args size.
5629 (aarch64_stack_clash_protection_alloca_probe_range,
5630 TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5632 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5635 * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
5637 * target.def (stack_clash_protection_alloca_probe_range): New.
5638 (stack_clash_protection_final_dynamic_probe): Remove.
5639 * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
5640 (default_stack_clash_protection_final_dynamic_probe): Remove.
5641 * targhooks.c: Likewise.
5642 * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5643 (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
5644 * doc/tm.texi: Regenerate.
5646 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5649 * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
5650 * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
5651 aarch64_clamp_to_uimm12_shift): New.
5652 (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
5653 * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
5655 2018-10-01 Tamar Christina <tamar.christina@arm.com>
5658 * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
5660 2018-10-01 Jeff Law <law@redhat.com>
5661 Richard Sandiford <richard.sandiford@linaro.org>
5662 Tamar Christina <tamar.christina@arm.com>
5665 * config/aarch64/aarch64.md
5666 (probe_stack_range): Add k (SP) constraint.
5667 * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
5668 STACK_CLASH_MAX_UNROLL_PAGES): New.
5669 * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
5670 stack probes for stack clash.
5671 (aarch64_allocate_and_probe_stack_space): New.
5672 (aarch64_expand_prologue): Use it.
5673 (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
5674 (aarch64_sub_sp): Add emit_move_imm optional param.
5676 2018-10-01 MCC CS <deswurstes@users.noreply.github.com>
5678 PR tree-optimization/87261
5679 * match.pd: Remove trailing whitespace.
5680 Add (x & y) | ~(x | y) -> ~(x ^ y),
5681 (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
5683 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
5685 * config/arc/arc.md (*add_n): Clean up pattern, update instruction
5687 (ashlsi3_insn): Update instruction constraints.
5688 (ashrsi3_insn): Likewise.
5689 (rotrsi3): Likewise.
5690 (add_shift): Likewise.
5691 * config/arc/constraints.md (Csz): New 32 bit constraint. It
5692 avoids placing in the limm field small constants which, otherwise,
5693 could end into a small instruction.
5695 2018-10-01 Claudiu Zissulescu <claziss@synopsys.com>
5697 * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
5698 destination register is not odd-even.
5699 (umaddsidi4_split): Likewise.
5701 2018-10-01 Richard Biener <rguenther@suse.de>
5703 * tree-inline.c (expand_call_inline): Store origin of fn
5704 in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
5705 * tree.c (block_ultimate_origin): Simplify and do some
5708 2018-09-30 Uros Bizjak <ubizjak@gmail.com>
5710 * config/i386/mmx.md (EMMS): New int iterator.
5711 (emms): New int attribute.
5712 (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
5713 EMMS int iterator. Explicitly declare clobbers.
5714 (mmx_emms): Remove expander.
5716 * config/i386/predicates.md (emms_operation): Remove predicate.
5717 (vzeroall_pattern): New predicate.
5718 (vzeroupper_pattern): Rename from vzeroupper_operation.
5719 * config/i386/i386.c (ix86_avx_u128_mode_after): Use
5720 vzeroupper_pattern and vzeroall_pattern predicates.
5722 2018-09-30 Peter Bergner <bergner@linux.ibm.com>
5724 PR rtl-optimization/86939
5725 * ira-lives.c (make_hard_regno_born): Rename from this...
5726 (make_hard_regno_live): ... to this. Remove update to conflict
5727 information. Update function comment.
5728 (make_hard_regno_dead): Add conflict information update. Update
5730 (make_object_born): Rename from this...
5731 (make_object_live): ... to this. Remove update to conflict information.
5732 Update function comment.
5733 (make_object_dead): Add conflict information update. Update function
5735 (mark_pseudo_regno_live): Call make_object_live.
5736 (mark_pseudo_regno_subword_live): Likewise.
5737 (mark_hard_reg_dead): Update function comment.
5738 (mark_hard_reg_live): Call make_hard_regno_live.
5739 (process_bb_node_lives): Likewise.
5740 * lra-lives.c (make_hard_regno_born): Rename from this...
5741 (make_hard_regno_live): ... to this. Remove update to conflict
5742 information. Remove now uneeded check_pic_pseudo_p argument.
5743 Update function comment.
5744 (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
5745 to conflict information. Update function comment.
5746 (mark_pseudo_live): Remove update to conflict information. Update
5748 (mark_pseudo_dead): Add conflict information update.
5749 (mark_regno_live): Call make_hard_regno_live.
5750 (mark_regno_dead): Call make_hard_regno_dead with new arguement.
5751 (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
5753 2018-09-29 H.J. Lu <hongjiu.lu@intel.com>
5756 * config/i386/i386.c (construct_container): Use TImode for
5757 BLKmode values in 2 integer registers.
5759 2018-09-29 Jeff Law <law@redhat.com>
5761 * builtins.c (unterminated_array): Pass in c_strlen_data * to
5762 c_strlen rather than just a tree *.
5763 (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
5764 Update recursive calls appropriately. If caller did not provide a
5765 suitable data pointer, create a local one. When a non-terminated
5766 string is discovered, bubble up information about the string via the
5767 c_strlen_data object.
5768 * builtins.h (c_strlen): Update prototype.
5769 (c_strlen_data): New structure.
5770 * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
5771 For a type 2 call, if c_strlen indicates a non-terminated string
5772 use the length of the non-terminated string.
5773 (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
5775 2018-09-29 Jakub Jelinek <jakub@redhat.com>
5778 * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
5779 __m512d type for __A argument rather than __m512.
5781 2018-09-28 John David Anglin <danglin@gcc.gnu.org>
5783 * match.pd (simple_comparison): Don't optimize if either operand is
5784 a function pointer when target needs function pointer canonicalization.
5786 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
5788 * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
5789 power5 .. power9 to remove indirection.
5790 * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
5791 ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
5792 ASM_CPU_476_SPEC): Delete.
5793 (ASM_CPU_SPEC): Adjust.
5794 (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
5795 asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
5797 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
5799 * config.in: Delete HAVE_AS_DCI.
5800 * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
5801 * config/rs6000/rs6000.h: Ditto.
5802 * configure.ac: Delete HAVE_AS_DCI.
5803 * configure: Regenerate.
5805 2018-09-28 Segher Boessenkool <segher@kernel.crashing.org>
5807 * config.in (HAVE_AS_LWSYNC): Delete.
5808 * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5809 * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
5811 * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5812 * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
5814 * configure.ac: Delete HAVE_AS_LWSYNC.
5815 * configure: Regenerate.
5817 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
5818 Pierre-Marie de Rodat <derodat@adacore.com>
5820 * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
5821 * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
5822 (cgraph_node::create_thunk): Add indirect_offset parameter.
5823 (thunk_adjust): Likewise.
5824 * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
5825 and initialize the corresponding field with it.
5826 (cgraph_node::dump): Dump indirect_offset field.
5827 * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
5828 * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
5829 (thunk_adjust): Add indirect_offset parameter and deal with it.
5830 (cgraph_node::expand_thunk): Deal with the indirect_offset field and
5831 pass it to thunk_adjust. Do not call the target hook if it's non-zero
5832 or if the thunk is external or local. Fix formatting. Do not chain
5833 the RESULT_DECL to BLOCK_VARS. Pass the static chain to the target,
5834 if any, in the GIMPLE representation.
5835 * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
5836 * lto-cgraph.c (lto_output_node): Write indirect_offset field.
5837 (input_node): Read indirect_offset field.
5838 * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
5839 call to thunk_adjust.
5840 * tree-nested.c (struct nesting_info): Add thunk_p field.
5841 (create_nesting_tree): Set it.
5842 (convert_all_function_calls): Copy static chain from targets to thunks.
5843 (finalize_nesting_tree_1): Return early for thunks.
5844 (unnest_nesting_tree_1): Do not finalize thunks.
5845 (gimplify_all_functions): Do not gimplify thunks.
5847 2018-09-28 David Malcolm <dmalcolm@redhat.com>
5849 * opt-suggestions.c (option_proposer::build_option_suggestions):
5850 Release "option_values".
5852 2018-09-28 David Malcolm <dmalcolm@redhat.com>
5854 * coverage.c (get_coverage_counts): Convert problem-reporting dump
5855 messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
5856 * dumpfile.c (kind_as_string): New function.
5857 (dump_loc): Rather than a hardcoded prefix of "note: ", use
5858 kind_as_string to vary the prefix based on dump_kind.
5859 (selftest::test_capture_of_dump_calls): Update for above.
5861 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
5863 * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
5864 (GET_SSE_REGNO): Rename from SSE_REGNO. Update all uses for rename.
5866 2018-09-28 Uros Bizjak <ubizjak@gmail.com>
5868 * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
5869 * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
5870 INVALID_REGNUM instead of FPSR_REG.
5871 (ix86_md_asm_adjust): Do not clobber FPSR_REG.
5872 * config/i386/i386.md: Update comment of FP compares.
5873 (fldenv): Do not clobber FPSR_REG.
5875 2018-09-28 Richard Biener <rguenther@suse.de>
5877 * tree.h (BLOCK_ORIGIN): New.
5878 * omp-expand.c (grid_expand_target_grid_body): Assign
5879 BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
5880 * tree-inline.c (remap_block): Likewise.
5881 * auto-profile.c (get_function_decl_from_block): Simplify
5882 by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
5883 * langhooks.c (lhd_print_error_function): Likewise.
5884 * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
5886 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5887 * tree.c (block_nonartificial_location): Likewise.
5888 (block_ultimate_origin): Likewise.
5889 * tree-pretty-print.c (percent_K_format): Likewise. Remove
5890 no longer needed LTO case.
5892 2018-09-28 Andrew Stubbs <ams@codesourcery.com>
5893 Jan Hubicka <jh@suse.cz>
5894 Martin Jambor <mjambor@suse.cz>
5896 * simplify-rtx.c (simplify_merge_mask): New function.
5897 (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
5898 same masks are used in op1 or op2.
5899 (test_vec_merge): New function.
5900 (test_vector_ops): Call test_vec_merge.
5902 2018-09-28 Eric Botcazou <ebotcazou@adacore.com>
5904 * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
5905 * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
5906 * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
5907 (cypress_costs): Set it.
5908 (supersparc_costs): Likewise.
5909 (hypersparc_costs): Likewise.
5910 (leon_cost): Likewise.
5911 (leon3_costs): Likewise.
5912 (sparclet_costs): Likewise.
5913 (ultrasparc_costs): Likewise.
5914 (ultrasparc_costs): Likewise.
5915 (niagara_costs): Likewise.
5916 (niagara2_costs): Likewise.
5917 (niagara3_costs): Likewise.
5918 (niagara4_costs): Likewise.
5919 (niagara7_costs): Likewise.
5920 (m8_costs): Likewise.
5921 (TARGET_CAN_FOLLOW_JUMP): Define.
5922 (pass_work_around_errata::gate): Minor tweak.
5923 (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
5924 Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
5925 Automaitcally clear MASK_FSMULD mask for V7 processors.
5926 (sparc_can_follow_jump): New static function.
5927 (output_ubranch): Deal with CROSSING_JUMP_P.
5928 (sparc_use_sched_lookahead): Rewrite using switch statement.
5929 (sparc_issue_rate): Reorder.
5930 (sparc_branch_cost): New function.
5932 2018-09-27 Martin Sebor <msebor@redhat.com>
5934 * tree.h (tree_to_shwi): Add attribute nonnull and pure.
5935 (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
5936 (int_fits_type_p): Same.
5938 2018-09-27 Uros Bizjak <ubizjak@gmail.com>
5940 * config/i386/i386.md (FPCR_REG): Remove.
5941 (UNSPEC_FLDCW): Remove.
5942 (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
5943 (x86_fldcw_1): Remove insn pattern.
5944 (fnstenv): Do not clobber FPCR_REG.
5946 * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
5947 (CALL_USED_REGISTERS): Ditto.
5948 (REG_ALLOC_ORDER): Ditto.
5949 (REG_CLASS_CONTENTS): Ditto.
5950 (HI_REGISTER_NAMES): Ditto.
5951 (ADDITIONAL_REGISTER_NAMES): Use defines instead
5952 of numerical constants.
5953 * config/i386/i386.c (regclass_map): Remove fpsr register.
5954 (dbx_register_map): Ditto.
5955 (dbx64_register_map): Ditto.
5956 (svr4_dbx_register_map): Ditto.
5957 (print_reg): Do not handle FPCR_REG.
5959 2018-09-27 Segher Boessenkool <segher@kernel.crashing.org>
5962 * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
5963 HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
5964 Delete, always treat as true.
5965 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
5966 Ditto. Simplify remaining code.
5967 * config/powerpcspe/powerpcspe.h: Ditto.
5968 * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
5969 Simplify remaining code.
5970 (rs6000_expand_builtin): Ditto.
5971 * config/rs6000/rs6000.h: Ditto.
5972 * configure.ac: Ditto.
5973 * configure: Regenerate.
5975 2018-09-27 Martin Liska <mliska@suse.cz>
5977 * coverage.c (get_coverage_counts): Revert the formatting
5978 of missing profile opt info.
5980 2018-09-27 Richard Biener <rguenther@suse.de>
5984 * dwarf2out.c (set_block_origin_self): Do not mark outermost
5985 block as we do not output that.
5986 (gen_inlined_subroutine_die): Elide the originally outermost
5987 block, matching what we do for concrete instances.
5988 (decls_for_scope): Add parameter specifying whether to recurse
5991 2018-09-27 Andrew Stubbs <ams@codesourcery.com>
5992 Tom de Vries <tom@codesourcery.com>
5996 * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
5997 STORE_FLAG_VALUE == 1.
5999 2018-09-27 Andreas Krebbel <krebbel@linux.ibm.com>
6001 * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
6002 constant definitions.
6003 ("tx_assist"): Replace magic number with PPA_TX_ABORT.
6004 ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
6005 ("speculation_barrier"): New expander definition.
6007 2018-09-26 Indu Bhagat <indu.bhagat@oracle.com>
6009 PR gcov-profile/86957
6010 * common.opt: New warning option -Wmissing-profile.
6011 * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
6012 * doc/invoke.texi: Document -Wmissing-profile.
6014 2018-09-26 Jim Wilson <jimw@sifive.com>
6016 * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
6017 (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
6019 2018-09-26 Martin Sebor <msebor@redhat.com>
6021 * tree.c (zerop): Change return type to bool.
6022 (integer_zerop, integer_onep, integer_each_onep): Same.
6023 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
6024 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
6025 (real_onep, real_minus_onep, chain_index): Same.
6026 (print_type_hash_statistics, type_list_equal): Same.
6027 * tree.h (zerop): Same.
6028 (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
6029 (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
6030 (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
6031 (real_onep, real_minus_onep, chain_index): Same.
6032 (print_type_hash_statistics, type_list_equal): Same.
6034 2018-09-26 Jim Wilson <jimw@sifive.com>
6036 * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
6038 2018-09-26 Jakub Jelinek <jakub@redhat.com>
6041 * config/i386/i386.c: Include debug.h and dwarf2out.h.
6042 (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
6045 2018-09-25 Andrew Stubbs <ams@codesourcery.com>
6047 * builtins.c (get_builtin_sync_mem): Force address mode conversion.
6049 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
6051 * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
6052 and FP_SECOND_SSE_REGS.
6053 (REG_CLASS_NAMES): Ditto.
6054 (REG_CLASS_CONTENTS): Ditto.
6055 * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
6056 FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
6057 (ix86_preferred_output_reload_class): Ditto.
6058 * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
6059 clobber constraint to "=&f".
6060 (fix_truncdi_i387): Ditto.
6061 (lrintxfdi2): Ditto.
6062 (fistdi2_<rounding>): Ditto.
6063 (fpremxf4_i387): Change "=u" constraint to "=f".
6064 (fprem1xf4_i387): Ditto.
6066 (fptanxf4_i387): Ditto.
6067 (fxtractxf3_i387): Ditto.
6068 (fscalexf4_i387): Ditto.
6069 (atan2xf3): Change "u" constraint to "f".
6070 (fyl2xxf3_i387): Ditto.
6071 (fyl2xp1xf3_i387): Ditto.
6073 2018-09-26 Uros Bizjak <ubizjak@gmail.com>
6076 * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
6077 for removed I387_MASK_PM entity.
6080 2018-09-26 Jeff Law <law@redhat.com>
6082 2018-09-26 Alexey Neyman <stilor@att.net>
6084 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
6085 headers are no longer pulled in by <isl/val.h>.
6087 2018-09-26 Richard Biener <rguenther@suse.de>
6090 * dwarf2out.c (gen_lexical_block_die): Do not equate inline
6091 or concrete instance DIE to the tree. Create abstract origin
6092 attributes also for concrete instances.
6094 2018-09-26 Alexey Neyman <stilor@att.net>
6096 * graphite.h: Include <isl/id.h> and <isl/space.h>; these
6097 headers are no longer pulled in by <isl/val.h>.
6099 2018-09-26 Matthew Malcomson <matthew.malcomson@arm.com>
6101 * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
6102 Use new helper functions.
6103 * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
6104 Use new helper functions.
6105 * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
6106 aarch_mm_needs_release): New declarations.
6107 * config/arm/aarch-common.c (aarch_mm_needs_acquire,
6108 aarch_mm_needs_release): New.
6110 2018-09-26 Eric Botcazou <ebotcazou@adacore.com>
6112 * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
6113 (arm32_output_mi_thunk): Deal with long calls.
6115 2018-09-26 Richard Biener <rguenther@suse.de>
6119 * tree-inline.c (expand_call_inline): When the location
6120 of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
6121 or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
6122 the inserted BLOCK to make inlined_function_outer_scope_p
6124 * dwarf2out.c (add_call_src_coords_attributes): Do not add
6125 coords for reserved locations.
6127 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
6129 * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
6130 (*call_indirect_nonlocal_sysv<mode>): Ditto.
6131 (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
6132 (*sibcall_nonlocal_sysv<mode>): Ditto.
6133 (*sibcall_value_nonlocal_sysv<mode>): Ditto.
6134 (<bd>_<mode>): Ditto.
6135 (<bd>tf_<mode>): Ditto.
6137 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
6139 * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
6140 control string as a list of templates instead of as C code.
6141 (*altivec_movti): Ditto.
6142 * config/rs6000/darwin.md (movdf_low_di): Ditto.
6144 2018-09-25 Jim Wilson <jimw@sifive.com>
6146 * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
6147 when target symbol is weak.
6149 2018-09-25 Bernd Edlinger <bernd.edlinger@hotmail.de>
6152 * builtins.c (unterminated_array): Simplify.
6153 * expr.c (string_constant): Handle SSA_NAME. Add more exceptions
6154 where pointer arithmetic is safe.
6156 2018-09-25 Segher Boessenkool <segher@kernel.crashing.org>
6159 * config/rs6000/altivec.md (altivec_vspltb): Use
6160 const_0_to_15_operand instead of u5bit_cint_operand.
6161 (*altivec_vspltb_internal): Ditto.
6162 (altivec_vspltb_direct): Ditto.
6163 (altivec_vsplth): Use const_0_to_7_operand instead of
6165 (*altivec_vsplth_internal): Ditto.
6166 (altivec_vsplth_direct): Ditto.
6167 (altivec_vspltw): Use const_0_to_3_operand instead of
6169 (*altivec_vspltw_internal): Ditto.
6170 (altivec_vspltw_direct): Ditto.
6171 (altivec_vspltsf): Ditto.
6172 (*altivec_vspltsf_internal): Ditto.
6173 * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
6174 various splats with the proper size immediate. Reorder the various
6175 cases by ascending size of immediate, and put all such together.
6177 2018-09-25 Richard Biener <rguenther@suse.de>
6180 * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
6182 (maybe_create_die_with_external_ref): Do not create
6183 DW_TAG_imported_unit here.
6184 (add_abstract_origin_attribute): Handle external BLOCK refs.
6185 (dwarf2out_abstract_function): Simplify LTO case.
6186 (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
6187 rather than using maybe_create_die_with_external_ref.
6189 2018-09-25 Uros Bizjak <ubizjak@gmail.com>
6192 * config/i386/i386.md (frndintxf2_mask_pm): Remove.
6193 (frndintxf2_mask_pm_i387): Ditto.
6194 (nearbyintxf2): Rewrite expander pattern to match rintxf2.
6195 Enable for !flag_trapping_math.
6196 (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
6197 Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
6198 Change operand 1 predicate to nonimmediate_operand.
6199 (attr "i387_cw"): Remove mask_pm.
6200 * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
6201 (enum ix86_entity): Remove I387_MASK_PM.
6202 * config/i386/i386.c (ix86_i387_mode_needed): Do not
6203 handle I387_MASK_PM.
6204 (ix86_mode_needed): Ditto.
6205 (ix86_mode_after): Ditto.
6206 (ix86_mode_entry): Ditto.
6207 (ix86_mode_exit): Ditto.
6208 (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
6210 2018-09-25 Jakub Jelinek <jakub@redhat.com>
6212 * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
6213 to_update_switch_stmts to vNULL instead of calling create on them
6216 2018-09-25 Richard Biener <rguenther@suse.de>
6218 PR tree-optimization/87402
6219 * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
6220 (visit_phi): Re-instantiate handling of supposed to be VARYING
6221 but non-VARYING backedge value.
6223 2018-09-25 Richard Biener <rguenther@suse.de>
6226 * dwarf2out.c (struct sym_off_pair): New.
6227 (external_die_map): New global.
6228 (lookup_decl_die): When in LTO create DIEs lazily from the
6230 (lookup_block_die): New function, create DIEs lazily in LTO.
6231 (equate_block_to_die): New function.
6232 (dwarf2out_die_ref_for_decl): During WPA get the association
6233 from the external DIE map.
6234 (dwarf2out_register_external_die): Record mapping into the
6236 (maybe_create_die_with_external_ref): New function split out from
6237 DIE generation part of old dwarf2out_register_external_die.
6238 (add_abstract_origin_attribute): Do not return the DIE. When
6239 in LTO reference externals directly.
6240 (dwarf2out_abstract_function): When in LTO ignore calls for
6241 decls with external DIEs (already present abstract instances).
6242 (gen_call_site_die): Adjust.
6243 (add_high_low_attributes): Likewise.
6244 (gen_lexical_block_die): Likewise.
6245 (gen_inlined_subroutine_die): Likewie.
6246 (gen_block_die): Likewise.
6247 (dwarf2out_inline_entry): Likewise.
6248 (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
6251 2018-09-25 Martin Liska <mliska@suse.cz>
6253 * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
6254 integers and not by a float value.
6256 2018-09-25 Martin Liska <mliska@suse.cz>
6259 * dbgcnt.c (dbg_cnt_process_single_pair): Return false
6261 * dumpfile.c (dump_enable_all): Remove extra parenthesis.
6262 * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
6263 * godump.c (go_format_type): Remove extra parenthesis.
6265 2018-09-25 Martin Liska <mliska@suse.cz>
6267 * alias.c (set_dest_equal_p): Remove unused function.
6268 * config/i386/i386.c (def_builtin_pure2): Likewise.
6269 * diagnostic-show-locus.c (class layout): Remove
6271 (layout::layout): Likewise here.
6272 * dump-context.h (class temp_dump_context): Likewise.
6273 * dwarf2out.c (add_AT_fde_ref): Remove unused function.
6274 (add_AT_loclistsptr): Likewise.
6275 (add_AT_offset): Likewise.
6276 (get_AT_hi_pc): Likewise.
6277 (is_comdat_die): Likewise.
6278 (type_is_enum): Likewise.
6279 (ceiling): Likewise.
6280 (add_AT_vms_delta): Likewise.
6281 (is_class_die): Likewise.
6282 * edit-context.c (class line_event): Remove unused field.
6283 * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
6285 * ipa-cp.c (ipa_get_vr_lat): Likewise.
6286 * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
6287 (ok_for_base_p_nonstrict): Likewise.
6288 * tree-chrec.c (is_not_constant_evolution): Likewise.
6289 (chrec_fold_poly_cst): Likewise.
6290 * tree-if-conv.c (has_pred_critical_p): Likewise.
6291 * tree-ssa-coalesce.c (print_exprs): Likewise.
6292 * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
6293 * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
6294 * tree-vrp.c (value_ranges_intersect_p): Likewise.
6295 (value_range_nonnegative_p): Likewise.
6297 2018-09-25 Martin Liska <mliska@suse.cz>
6299 * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
6300 Do not handle "GNU Pascal".
6301 * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
6303 * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
6304 from documentation. Likewise.
6305 * dbxout.c (dbxout_range_type): Likewise.
6306 * doc/cpp.texi: Likewise.
6307 * doc/extend.texi: Likewise.
6308 * doc/frontends.texi: Likewise.
6309 * doc/invoke.texi: Remove Pascal entry.
6310 * tree.def (CLEANUP_POINT_EXPR): Likewise.
6311 * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
6313 2018-09-25 Martin Liska <mliska@suse.cz>
6316 * doc/invoke.texi: Document all parameters and remove default
6319 2018-09-25 Ilya Leoshkevich <iii@linux.ibm.com>
6322 * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
6323 contains HOST_WIDE_INTs when computing its size.
6325 2018-09-24 Jim Wilson <jimw@sifive.com>
6328 * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
6330 (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
6332 2018-09-24 Andrew Pinski <apinski@marvell.com>
6334 * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
6335 access prev before checking it for NULLness in the
6336 AARCH64_FUSE_CMP_BRANCH case.
6338 2018-09-24 H.J. Lu <hongjiu.lu@intel.com>
6341 * config/i386/i386.c (rest_of_insert_endbranch): Set
6342 endbr_queued_at_entrance to true and don't insert ENDBR if
6343 x86_function_profiler will be called.
6344 (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
6346 * config/i386/i386.h (machine_function): Add
6347 endbr_queued_at_entrance.
6349 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
6351 * genattrtab.c (mk_attr_alt): Use alternative_mask.
6352 (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
6354 (check_attr_test): Use alternative_mask.
6355 (get_attr_value): Likewise.
6356 (compute_alternative_mask): Use alternative_mask and XWINT.
6357 (make_alternative_compare): Use alternative_mask.
6358 (attr_alt_subset_p): Use XWINT.
6359 (attr_alt_subset_of_compl_p): Likewise.
6360 (attr_alt_intersection): Use alternative_mask and XWINT.
6361 (attr_alt_union): Likewise.
6362 (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
6363 (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
6364 (simplify_test_exp): Use alternative_mask and XWINT.
6365 (write_test_expr): Use alternative_mask and XWINT, adjust bit
6366 number calculation to support 64 bits. Generate code that
6367 checks 64-bit masks.
6368 (main): Use alternative_mask.
6369 * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
6371 2018-09-24 Ilya Leoshkevich <iii@linux.ibm.com>
6374 * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
6375 RETURN+USE when returning via %r14.
6377 2018-09-24 Martin Liska <mliska@suse.cz>
6379 * gcov.c (output_lines): Print colorization legend
6380 for both flag_use_colors and flag_use_hotness_colors.
6383 2018-09-24 Martin Liska <mliska@suse.cz>
6385 * coverage.c (get_coverage_counts): Use warning_at
6386 with current_function_decl location. Use %qD in warning
6389 2018-09-24 Martin Liska <mliska@suse.cz>
6391 * memory-block.h (memory_block_pool::release): Annotate with
6392 valgrind that the memory is not accessible.
6394 2018-09-24 Martin Liska <mliska@suse.cz>
6397 * asan.c: Make asan_handled_variables extern.
6399 * cfgexpand.c (expand_stack_vars): Make sure
6400 a representative is unpoison if another
6401 variable in the partition is handled by
6402 use-after-scope sanitization.
6404 2018-09-24 Richard Biener <rguenther@suse.de>
6406 PR tree-optimization/63155
6407 * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
6408 the worklist when the edge of the respective argument isn't
6411 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
6413 * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
6414 ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
6415 (MASK_CLASS_P): Update for rename.
6416 (MAYBE_MASK_CLASS_P): Ditto.
6417 (REG_CLASS_NAMES): Update.
6418 (REG_CLASS_CONTENT): Update.
6419 * config/i386/i386.c (regclass_map): Update for MASK_REG
6420 and ALL_MASK_REGS rename.
6421 * config/i386/constraints.md (Yk): Update for rename.
6424 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
6426 * config/i386/i386.h (enum reg_class): Remove
6427 EVEX_SSE_REGS and MOD4_SSE_REGS.
6428 (REG_CLASS_NAMES): Update.
6429 (REG_CLASS_CONTENT): Update.
6430 * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
6431 registers as ALL_SSE_REGS.
6432 (ix86_additional_allocno_class_p): Remove.
6433 (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
6434 (ix86_register_priority): Lower priority of EVEX SSE registers.
6435 Use IN_RANGE macro where appropriate.
6436 (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
6437 AVX-5124VNNIW checks.
6438 (ix86_modes_tieable_p): Tie 512-bit SSE modes.
6439 * config/i386/sse.md (avx5124fmaddps_4fmaddps)
6440 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
6441 (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
6442 (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
6443 (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
6444 (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
6445 (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
6446 (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
6447 (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
6448 (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
6449 * config/i386/constraints.md (Yh): Remove.
6451 2018-09-23 Uros Bizjak <ubizjak@gmail.com>
6453 * config/i386/i386.c (regclass_map): Declare integer REX registers
6456 2018-09-23 Gerald Pfeifer <gerald@pfeifer.com>
6458 * doc/service.texi (Service): Switch the fsf.org link to https.
6460 2018-09-22 Chung-Ju Wu <jasonwucj@gmail.com>
6463 * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6464 Define to speculation_safe_value_not_needed.
6466 2018-09-21 Florian Weimer <fweimer@redhat.com>
6469 * doc/extend.texi (Common Function Attributes): Mention that
6470 noreturn suppresses tail call optimization.
6472 2018-09-21 Jeff Law <law@redhat.com>
6474 * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
6475 vr_values::cleanup_edges_and_switches.
6476 * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
6478 (identify_jump_threads): Remove EDGE_IGNORE handling.
6479 (execute_vrp): Move handling of to_remove_edges and
6480 to_update_switch_stmts into vr_values class member functions.
6481 * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
6482 (to_update_switch_stmts): Likewise.
6483 * vr-values.c: Include cfghooks.h.
6484 (vr_values::vr_values): Initialize to_remove_edges and
6485 to_update_switch_stmts.
6486 (vr_values::~vr_values): Verify to_remove_edges and
6487 to_update_switch_stmts are empty.
6488 (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
6489 (vr_values::cleanup_edges_and_switches): New member function.
6490 * vr-values.h (vr_values): Add cleanup_edges_and_switches member
6491 function. Add new data members.
6493 2018-09-21 David Malcolm <dmalcolm@redhat.com>
6495 PR tree-optimization/87309
6496 * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
6497 calls with pflags and alt_flags.
6498 (selftest::test_capture_of_dump_calls): Add test of interaction of
6499 MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
6501 2018-09-21 Olivier Hainque <hainque@adacore.com>
6503 * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
6505 2018-09-21 Olivier Hainque <hainque@adacore.com>
6507 * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
6509 2018-09-21 Olivier Hainque <hainque@adacore.com>
6511 * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
6512 Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
6513 (VXWORKS_LIBS_RTP): Minor reordering.
6515 2018-09-21 Olivier Hainque <hainque@adacore.com>
6517 * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
6518 (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
6520 2018-09-21 Olivier Hainque <hainque@adacore.com>
6522 * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
6523 (PTRDIFF_TYPE): Likewise.
6525 2018-09-21 Olivier Hainque <hainque@adacore.com>
6527 * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
6528 triplet, similar to support for VxWorks7.
6529 * config/vxworks-dummy.h: Provide a default definition
6530 of TARGET_VXWORKS64 to 0.
6532 2018-09-21 Olivier Hainque <hainque@adacore.com>
6534 * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
6535 * config/vxworks-dummy.h: here.
6537 2018-09-21 Olivier Hainque <hainque@adacore.com>
6539 * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
6541 2018-09-21 Shaokun Zhang <zhangshaokun@hisilicon.com>
6542 Bo Zhou <zbo.zhou@hisilicon.com>
6544 * config/aarch64/aarch64-cores.def (tsv110): New CPU.
6545 * config/aarch64/aarch64-tune.md: Regenerated.
6546 * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
6547 * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
6548 * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
6550 2018-09-21 Andrew Stubbs <ams@codesourcery.com>
6551 Julian Brown <julian@codesourcery.com>
6553 * builtins.c (get_builtin_sync_mem): Handle address spaces.
6555 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
6557 * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
6558 if the call takes a static chain.
6560 2018-09-21 Martin Liska <mliska@suse.cz>
6562 * auto-profile.c (autofdo_source_profile::read): Do not
6564 (read_profile): Do not add working sets.
6565 (read_autofdo_file): Remove sum_all.
6566 (afdo_callsite_hot_enough_for_early_inline): Remove const
6568 * coverage.c (struct counts_entry): Remove gcov_summary.
6569 (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
6570 do not support GCOV_TAG_PROGRAM_SUMMARY.
6571 (get_coverage_counts): Remove summary and expected
6573 * coverage.h (get_coverage_counts): Likewise.
6574 * doc/gcov-dump.texi: Remove -w option.
6575 * gcov-dump.c (dump_working_sets): Remove.
6576 (main): Do not support '-w' option.
6577 (print_usage): Likewise.
6578 (tag_summary): Likewise.
6579 * gcov-io.c (gcov_write_summary): Do not dump
6581 (gcov_read_summary): Likewise.
6582 (gcov_histo_index): Remove.
6583 (gcov_histogram_merge): Likewise.
6584 (compute_working_sets): Likewise.
6585 * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
6587 (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
6588 (GCOV_TAG_SUMMARY_LENGTH): Adjust.
6589 (GCOV_HISTOGRAM_SIZE): Remove.
6590 (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
6591 (struct gcov_summary): Simplify rapidly just
6592 to runs and sum_max fields.
6593 (gcov_histo_index): Remove.
6594 (NUM_GCOV_WORKING_SETS): Likewise.
6595 (compute_working_sets): Likewise.
6596 * gcov-tool.c (print_overlap_usage_message): Remove
6597 trailing empty line.
6598 * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
6599 (output_lines): Remove program related line.
6600 * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
6601 * lto-cgraph.c (output_profile_summary): Do not stream GCOV
6603 (input_profile_summary): Do not read it.
6604 (merge_profile_summaries): And do not merge it.
6605 (input_symtab): Do not call removed function.
6606 * modulo-sched.c (sms_schedule): Do not print sum_max.
6607 * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
6608 removed when histogram method was invented.
6609 (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
6611 * postreload-gcse.c (eliminate_partially_redundant_load): Fix
6613 * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
6614 and dump selected value.
6615 * profile.c (add_working_set): Remove.
6616 (get_working_sets): Likewise.
6617 (find_working_set): Likewise.
6618 (get_exec_counts): Do not work with working sets.
6619 (read_profile_edge_counts): Do not inform as sum_max is removed.
6620 (compute_branch_probabilities): Likewise.
6621 (compute_value_histograms): Remove argument for call of
6622 get_coverage_counts.
6623 * profile.h: Do not make gcov_summary const.
6625 2018-09-21 Monk Chiang <sh.chiang04@gmail.com>
6627 * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
6629 2018-09-21 Eric Botcazou <ebotcazou@adacore.com>
6631 PR tree-optimization/86990
6632 * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
6633 Check that the entire merged store group is made of constants only for
6636 2018-09-20 Allan Sandfeld Jensen <allan.jensen@qt.io>
6638 * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
6639 (VTABLE_VERIFICATION_SPEC): Likewise.
6640 (SANITIZER_EARLY_SPEC): Likewise.
6641 (SANITIZER_SPEC): Likewise.
6642 * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
6643 * doc/invoke.texi (Link Options): Document -r.
6645 2018-09-20 Richard Biener <rguenther@suse.de>
6648 * gimplify.c (gimplify_expr): Retain alignment of
6649 addressable lvalue in dereference.
6651 2018-09-20 Alexandre Oliva <aoliva@redhat.com>
6654 * configure.ac: Check for .loc is_stmt support.
6655 * configure, config.in: Rebuilt.
6656 * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
6659 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
6661 * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
6662 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
6665 2018-09-20 Segher Boessenkool <segher@kernel.crashing.org>
6667 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6669 * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
6672 2018-09-20 Marek Polacek <polacek@redhat.com>
6674 * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
6676 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
6678 PR tree-optimization/87288
6679 * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
6680 into account when determining PEELING_FOR_NITERS.
6682 2018-09-20 Richard Sandiford <richard.sandiford@arm.com>
6684 PR tree-optimization/86877
6685 * tree-vect-loop.c (vect_analyze_loop_2): Call
6686 vect_verify_datarefs_alignment.
6688 2018-09-19 Marek Polacek <polacek@redhat.com>
6690 * doc/invoke.texi: Document -Wclass-conversion.
6692 2018-09-19 John David Anglin <danglin@gcc.gnu.org>
6694 * config/pa/pa.c (pa_adjust_priority): Delete.
6695 (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
6697 * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
6698 (atomic_storehi): Likewise.
6699 (atomic_storesi): Likewise.
6700 (atomic_loaddi): Restore compare and swap exchange loop code.
6702 2018-09-19 Segher Boessenkool <segher@kernel.crashing.org>
6704 PR rtl-optimization/86902
6705 * combine.c (try_combine): When changing the CC mode used, don't change
6706 an unrelated mode in other_insn to that new CC mode.
6708 2018-09-19 David Malcolm <dmalcolm@redhat.com>
6710 * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
6711 with %T in place of calls to dump_generic_expr.
6712 (prune_runtime_alias_test_list): Likewise.
6713 (create_runtime_alias_checks): Likewise.
6714 * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
6715 (vect_analyze_data_ref_dependence): Likewise.
6716 (vect_slp_analyze_data_ref_dependence): Likewise.
6717 (vect_record_base_alignment): Likewise. Use %G in place of call
6718 to dump_gimple_stmt.
6719 (vect_compute_data_ref_alignment): Likewise.
6720 (verify_data_ref_alignment): Likewise.
6721 (vect_find_same_alignment_drs): Likewise.
6722 (vect_analyze_group_access_1): Likewise.
6723 (vect_analyze_data_ref_accesses): Likewise.
6724 (dependence_distance_ge_vf): Likewise.
6725 (dump_lower_bound): Likewise.
6726 (vect_prune_runtime_alias_test_list): Likewise.
6727 (vect_find_stmt_data_reference): Likewise.
6728 (vect_analyze_data_refs): Likewise.
6729 (vect_create_addr_base_for_vector_ref): Likewise.
6730 (vect_create_data_ref_ptr): Likewise.
6731 * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
6732 (vect_can_advance_ivs_p): Likewise.
6733 (vect_update_ivs_after_vectorizer): Likewise.
6734 (vect_gen_prolog_loop_niters): Likewise.
6735 (vect_prepare_for_masked_peels): Likewise.
6736 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6737 (vect_determine_vectorization_factor): Likewise.
6738 (vect_is_simple_iv_evolution): Likewise.
6739 (vect_analyze_scalar_cycles_1): Likewise.
6740 (vect_analyze_loop_operations): Likewise.
6741 (report_vect_op): Likewise.
6742 (vect_is_slp_reduction): Likewise.
6743 (check_reduction_path): Likewise.
6744 (vect_is_simple_reduction): Likewise.
6745 (vect_create_epilog_for_reduction): Likewise.
6746 (vect_finalize_reduction:): Likewise.
6747 (vectorizable_induction): Likewise.
6748 (vect_transform_loop_stmt): Likewise.
6749 (vect_transform_loop): Likewise.
6750 (optimize_mask_stores): Likewise.
6751 * tree-vect-patterns.c (vect_pattern_detected): Likewise.
6752 (vect_split_statement): Likewise.
6753 (vect_recog_over_widening_pattern): Likewise.
6754 (vect_recog_average_pattern): Likewise.
6755 (vect_determine_min_output_precision_1): Likewise.
6756 (vect_determine_precisions_from_range): Likewise.
6757 (vect_determine_precisions_from_users): Likewise.
6758 (vect_mark_pattern_stmts): Likewise.
6759 (vect_pattern_recog_1): Likewise.
6760 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6761 (vect_record_max_nunits): Likewise.
6762 (vect_build_slp_tree_1): Likewise.
6763 (vect_build_slp_tree_2): Likewise.
6764 (vect_print_slp_tree): Likewise.
6765 (vect_analyze_slp_instance): Likewise.
6766 (vect_detect_hybrid_slp_stmts): Likewise.
6767 (vect_detect_hybrid_slp_1): Likewise.
6768 (vect_slp_analyze_operations): Likewise.
6769 (vect_slp_analyze_bb_1): Likewise.
6770 (vect_transform_slp_perm_load): Likewise.
6771 (vect_schedule_slp_instance): Likewise.
6772 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6773 (vect_mark_stmts_to_be_vectorized): Likewise.
6774 (vect_init_vector_1): Likewise.
6775 (vect_get_vec_def_for_operand): Likewise.
6776 (vect_finish_stmt_generation_1): Likewise.
6777 (vect_check_load_store_mask): Likewise.
6778 (vectorizable_call): Likewise.
6779 (vectorizable_conversion): Likewise.
6780 (vectorizable_operation): Likewise.
6781 (vectorizable_load): Likewise.
6782 (vect_analyze_stmt): Likewise.
6783 (vect_is_simple_use): Likewise.
6784 (vect_get_vector_types_for_stmt): Likewise.
6785 (vect_get_mask_type_for_stmt): Likewise.
6786 * tree-vectorizer.c (increase_alignment): Likewise.
6788 2018-09-19 Andrew Stubbs <ams@codesourcery.com>
6790 * doc/rtl.texi: Adjust vec_select description.
6791 * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
6792 non-constant selectors.
6794 2018-09-19 Matthew Malcomson <matthew.malcomson@arm.com>
6796 * config/aarch64/aarch64-protos.h
6797 (aarch64_offset_9bit_signed_unscaled_p): New declaration.
6798 * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
6799 (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
6800 * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
6801 (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
6802 (AARCH64_FL_PROFILE): Move index so flags are ordered.
6803 (AARCH64_ISA_RCPC8_4): New flag.
6804 * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
6805 to aarch64_offset_9bit_signed_unscaled_p.
6806 * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
6808 * config/aarch64/constraints.md (Ust): New constraint.
6809 * config/aarch64/predicates.md.
6810 (aarch64_9bit_offset_memory_operand): New predicate.
6811 (aarch64_rcpc_memory_operand): New predicate.
6813 2018-09-19 Eric Botcazou <ebotcazou@adacore.com>
6815 PR rtl-optimization/87361
6816 * rtlanal.c (nonzero_bits1): Revert accidental change.
6818 2018-09-19 Richard Biener <rguenther@suse.de>
6820 PR tree-optimization/87349
6821 PR tree-optimization/87342
6822 * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
6824 2018-09-18 Marek Polacek <polacek@redhat.com>
6826 P1064R0 - Allowing Virtual Function Calls in Constant Expressions
6827 * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
6829 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
6831 * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
6833 2018-09-18 Segher Boessenkool <segher@kernel.crashing.org>
6835 PR rtl-optimization/86882
6836 * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
6838 2018-09-18 Uros Bizjak <ubizjak@gmail.com>
6840 * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
6841 *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
6843 2018-09-18 Jonathan Wakely <jwakely@redhat.com>
6846 * doc/invoke.texi (Link Options): Fix formatting and grammar.
6848 2018-09-18 Richard Biener <rguenther@suse.de>
6851 * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
6852 (compute_samebase_partition_bases): Likewise.
6853 (coalesce_ssa_name): Always use compute_optimized_partition_bases.
6854 (gimple_can_coalesce_p): Simplify.
6856 2018-09-18 Hans-Peter Nilsson <hp@bitrange.com>
6858 Handle a library implementation of ffs calling __builtin_ffs.
6859 * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
6860 (mmix_init_libfuncs): New function: make __builtin_ffs expand
6863 2018-09-17 David Malcolm <dmalcolm@redhat.com>
6865 * diagnostic-show-locus.c (class layout_range): Add field
6867 (layout_range::layout_range): Add "original_idx" param and use it
6868 to initialize new field.
6869 (make_range): Use 0 for original_idx.
6870 (layout::layout): Pass in index to calls to
6871 maybe_add_location_range.
6872 (layout::maybe_add_location_range): Add param "original_idx" and
6873 pass it on to layout_range.
6874 (layout::print_any_labels): Pass on range->m_original_idx to
6876 (gcc_rich_location::add_location_if_nearby): Use 0 for
6878 * gcc-rich-location.h (text_range_label::get_text): Update for new
6880 (range_label_for_type_mismatch::get_text): Likewise.
6882 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
6884 * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
6886 2018-09-17 David Malcolm <dmalcolm@redhat.com>
6888 * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
6889 format_string_diagnostic_t.
6890 (fmtwarn_n): Likewise.
6891 * substring-locations.c
6892 (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
6893 (format_warning_n_va): Convert to...
6894 (format_string_diagnostic_t::emit_warning_n_va): ...this.
6895 (format_warning_va): Convert to...
6896 (format_string_diagnostic_t::emit_warning_va): ...this.
6897 (format_warning_at_substring): Convert to...
6898 (format_string_diagnostic_t::emit_warning): ...this.
6899 (format_warning_at_substring_n): Convert to...
6900 (format_string_diagnostic_t::emit_warning_n): ...this.
6901 * substring-locations.h (class format_string_diagnostic_t): New
6903 (format_warning_va): Convert to
6904 format_string_diagnostic_t::emit_warning_va.
6905 (format_warning_n_va): Convert to
6906 format_string_diagnostic_t::emit_warning_n_va.
6907 (format_warning_at_substring): Convert to
6908 format_string_diagnostic_t::emit_warning.
6909 (format_warning_at_substring_n): Convert to
6910 format_string_diagnostic_t::emit_warning_n.
6912 2018-09-17 Cesar Philippidis <cesar@codesourcery.com>
6913 Bernd Schmidt <bernds_cb1@t-online.de>
6915 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
6918 2018-09-17 Uros Bizjak <ubizjak@gmail.com>
6920 * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
6921 operand 0 predicate to nonimmediate operand.
6922 (rint<mode>2_frndint): Remove insn pattern.
6923 (rint<mode>2): Change operand 1 predicate to general_operand.
6924 Extend operand 1 to XFmode and generate rintxf2 insn.
6925 (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
6926 Do not use X87MODEF mode macro.
6927 (frndintxf2_<rounding>_i387): Rename from
6928 frndint<mode>2_<rounding>_i387. Do not use X87MODEF mode macro.
6929 (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
6930 to XFmode and generate significandxf3 insn.
6932 2018-09-17 Richard Biener <rguenther@suse.de>
6934 PR tree-optimization/87328
6935 * tree-ssa-sccvn.c (process_bb): Remove assertion about not
6936 visiting unexecutable backedges when not iterating.
6937 (do_rpo_vn): Mark all edges not executable even when not
6940 2018-09-17 Martin Jambor <mjambor@suse.cz>
6943 * doc/invoke.texi (Warning Options): Likewise.
6945 2018-09-17 Richard Biener <rguenther@suse.de>
6947 PR tree-optimization/87301
6948 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
6949 clean EH info from leftover copy assignments.
6951 2018-09-17 Martin Liska <mliska@suse.cz>
6953 PR gcov-profile/85871
6954 * gcov.c (output_intermediate_file): Fix out of bounds
6957 2018-09-17 Vineet Gupta <vgupta@synopsys.com>
6959 * config/arc/arc.c: Object attributes for core4 not reflected
6961 * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
6964 2018-09-17 Alexey Brodkin <abrodkin@synopsys.com>
6966 * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
6968 2018-09-17 Martin Liska <mliska@suse.cz>
6970 * doc/gcov.texi: Document new option --use-hotness-colors.
6971 * gcov.c (struct source_info): Declare new field.
6972 (source_info::source_info): Set default for maximum_count.
6973 (print_usage): Add new -q option.
6974 (process_args): Process it.
6975 (accumulate_line_info): Save src->maximum_count.
6976 (output_line_beginning): Make color line number if
6977 flag_use_hotness_colors is set.
6978 (output_line_details): Pass default argument value.
6979 (output_lines): Pass src->maximum_count.
6981 2018-09-17 Martin Liska <mliska@suse.cz>
6983 * common/config/i386/i386-common.c (ix86_get_valid_option_values):
6984 Use processor_names table.
6985 * config/i386/i386.c (ix86_default_align): Use
6986 processor_cost_table for alignment values.
6987 (ix86_option_override_internal): Use processor_names.
6988 (ix86_function_specific_print): Likewise.
6989 * config/i386/i386.h (struct processor_costs):
6990 Add alignment values.
6991 (struct ptt): Remove and replace with const char *.
6992 * config/i386/x86-tune-costs.h (struct processor_costs):
6993 Declare default alignments for all costs.
6995 2018-09-17 Aldy Hernandez <aldyh@redhat.com>
6997 * tree-vrp.c (extract_range_from_unary_expr): Do not special case
6998 symbolics or VR_VARYING ranges for ABS_EXPR.
6999 * wide-int-range.cc (wide_int_range_abs): Return positive numbers
7000 when range will wrap.
7002 2018-09-15 Eric Botcazou <ebotcazou@adacore.com>
7005 * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
7006 before and after a JUMP_TABLE_DATA.
7008 2018-09-14 John David Anglin <danglin@gcc.gnu.org>
7011 * dojump.c (do_compare_and_jump): Canonicalize function pointers
7012 when one operand is a function pointer. Use POINTER_TYPE_P and
7013 FUNC_OR_METHOD_TYPE_P.
7014 * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
7015 * fold-const.c (build_range_check): Likewise.
7016 * match.pd (simple_comparison): Likewise.
7018 2018-09-14 David Malcolm <dmalcolm@redhat.com>
7021 * spellcheck.c (get_edit_distance_cutoff): New function.
7022 (selftest::test_edit_distance_unit_test_oneway): Rename to...
7023 (selftest::test_get_edit_distance_one_way): ...this.
7024 (selftest::test_get_edit_distance_unit): Rename to...
7025 (selftest::test_get_edit_distance_both_ways): ...this.
7026 (selftest::test_edit_distances): Move tests to this new function,
7027 and test some more pairs of strings. Update for above renaming.
7028 (selftest::get_old_cutoff): New function.
7029 (selftest::test_get_edit_distance_cutoff): New function.
7030 (selftest::assert_suggested_for): New function.
7031 (ASSERT_SUGGESTED_FOR): New macro.
7032 (selftest::assert_not_suggested_for): New function.
7033 (ASSERT_NOT_SUGGESTED_FOR): New macro.
7034 (selftest::test_suggestions): New function.
7035 (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
7036 tests to selftest::test_edit_distances and call it. Add calls to
7037 selftest::test_get_edit_distance_cutoff and
7038 selftest::test_suggestions.
7039 * spellcheck.h (get_edit_distance_cutoff): New function declaration.
7040 (best_match::consider): Replace hard-coded cutoff calculation with
7042 (best_match::get_cutoff): New declaration.
7043 (best_match::get_best_meaningful_candidate): Likewise.
7045 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
7047 * builtins.c (fold_builtin_strlen): Remove TODO comment.
7049 2018-09-14 Bernd Edlinger <bernd.edlinger@hotmail.de>
7052 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
7054 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
7055 terminated string literal.
7057 2018-09-14 Martin Sebor <msebor@redhat.com>
7059 * builtins.c (unterminated_array): Handle ARRAY_REF.
7060 (expand_builtin_stpcpy_1): Detect unterminated char arrays.
7061 * builtins.h (unterminated_array): Declare extern.
7062 * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
7064 (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
7067 2018-09-14 Martin Sebor <msebor@redhat.com>
7068 Jeff Law <law@redhat.com>
7070 * builtins.c (unterminated_array): New.
7071 (expand_builtin_strcpy): Adjust.
7072 (expand_builtin_strcpy_args): Detect unterminated arrays.
7073 * gimple-fold.c (get_maxval_strlen): Add argument. Detect
7074 unterminated arrays.
7075 * gimple-fold.h (get_maxval_strlen): Add argument.
7076 (gimple_fold_builtin_strcpy): Detec unterminated arrays.
7078 * gimple-fold.c (get_range_strlen): Add argument.
7079 (get_maxval_strlen): Adjust.
7080 * gimple-fold.h (get_range_strlen): Add argument.
7082 2018-09-14 Wei Xiao <wei3.xiao@intel.com>
7084 * config/i386/movdirintrin.h: Fix copyright year.
7086 2018-09-14 Uros Bizjak <ubizjak@gmail.com>
7088 * reg-stack.c: Include regs.h.
7089 (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
7090 (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
7091 FIRST_STACK_REG, not DFmode.
7092 (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
7093 FIRST_STACK_REG, not XFmode. Explicitly construct swap RTX.
7094 (change stack): Default register mode to the reg_raw_mode of
7095 FIRST_STACK_REG, not DFmode.
7096 * config/i386/i386.md (*swap<mode>): Remove insn pattern.
7097 (*swapxf): Rename from swapxf.
7099 2018-09-14 Carl Love <cel@us.ibm.com>
7101 * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
7102 * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
7104 2018-09-14 Segher Boessenkool <segher@kernel.crashing.org>
7107 * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
7110 2018-09-14 Sam Tebbs <sam.tebbs@arm.com>
7113 * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
7115 2018-09-14 Jason Merrill <jason@redhat.com>
7117 Fix --enable-gather-detailed-mem-stats.
7118 * hash-table.c (hash_table_usage): Change from variable to function.
7119 * hash-table.h: Adjust.
7120 * Makefile.in: Add missing dependencies on hash-table.h.
7122 2018-09-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7124 PR tree-optimization/87259
7126 (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
7127 execute_cse_reciprocals_1 has tried transforming.
7129 2018-09-14 Aldy Hernandez <aldyh@redhat.com>
7131 * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
7132 VR_VARYING for PLUS/MINUS_EXPR.
7134 2018-09-14 Ilya Leoshkevich <iii@linux.ibm.com>
7136 * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
7139 2018-09-14 Richard Biener <rguenther@suse.de>
7142 * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
7143 bits for the merged partition.
7145 2018-09-13 Martin Sebor <msebor@redhat.com>
7146 Bernd Edlinger <bernd.edlinger@hotmail.de>
7148 * builtins.h (c_srlen): Add argument.
7149 * builtins.c (warn_string_no_nul): New function.
7150 (c_strlen): Add argument and use it. Update recursive calls.
7151 Pass DECL argument to string_constant to get info on non
7152 terminated strings. Update *NONSTR as needed.
7153 (fold_builtin_strlen): Add argument to calls to c_strlen.
7154 Warn for unterminated arrays.
7155 (warn_string_no_null): Add prototype.
7156 * expr.c (string_constant): Update arguments. Update recursive
7157 calls appropriately. Detect missing NUL terminator and outermost
7158 declaration its missing in.
7159 Improve checks for arrays with nonzero lower bound or elements
7160 that are not a single byte. Simplify offset computation.
7161 Simplify checks for non-NUL terminated strings.
7162 * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
7163 * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
7165 2018-09-13 Bernd Edlinger <bernd.edlinger@hotmail.de>
7167 * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
7169 * fold-const.c (c_getstr): Fix function comment. Remove unused third
7170 argument. Fix range checks.
7171 * fold-const.h (c_getstr): Adjust protoype.
7172 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
7173 string is constant but contains no NUL byte.
7175 * expr.c (string_constant): Adjust function comment.
7176 Remove bogus check for zero termination.
7178 * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
7180 * varasm.c (compare_constant): Compare type size of STRING_CSTs.
7181 (get_constant_size): Don't make STRING_CSTs larger than they are.
7182 (check_string_literal): New check function for STRING_CSTs.
7183 (output_constant): Use it.
7185 2018-09-13 Eric Botcazou <ebotcazou@adacore.com>
7188 * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
7190 2018-09-13 Richard Biener <rguenther@suse.de>
7192 PR tree-optimization/87263
7193 * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
7194 (struct unwind_state): Add max_rpo field.
7195 (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
7196 Compute max_rpo, the max RPO number a block can be backwards reached
7197 from. Re-write non-iterating mode to a RPO ordered worklist approach,
7198 separating it from the iterating mode.
7200 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
7202 * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
7203 (rfs_decision): New scheduling decision.
7205 2018-09-13 Richard Biener <rguenther@suse.de>
7208 * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
7209 (vn_nary_op_insert_pieces_predicated): Do not write useless
7210 valid_dominated_by_p entry outside of the allocated storage.
7212 2018-09-13 Omar Sandoval <osandov@osandov.com>
7213 Tom de Vries <tdevries@suse.de>
7216 * dwarf2out.c (is_c): New function.
7217 (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
7219 2018-09-13 Sam Tebbs <sam.tebbs@arm.com>
7222 * config/aarch64/aarch64.md (*aarch64_bfxil):
7224 * config/aarch64/constraints.md (Ulc): Define.
7225 * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
7227 * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
7230 2018-09-13 Vlad Lazar <vlad.lazar@arm.com>
7232 * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
7233 * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
7234 aarch64_layout_frame call.
7235 (aarch64_expand_epilogue): Likewise.
7236 (aarch64_initial_elimination_offset): Likewise.
7237 (aarch64_get_separate_components): Likewise.
7238 (aarch64_use_return_insn_p): Likewise.
7239 (aarch64_layout_frame): Remove unneeded check.
7241 2018-09-13 Jakub Jelinek <jakub@redhat.com>
7243 * configure.ac: Only append
7244 " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
7245 gcc_config_arguments if it was never reconfigured or last reconfigure
7246 was with different arguments.
7247 * configure: Regenerated.
7249 2018-09-13 Jakub Jelinek <jakub@redhat.com>
7250 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7253 * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
7254 (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
7256 2018-09-13 Jakub Jelinek <jakub@redhat.com>
7258 PR tree-optimization/87287
7259 * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
7260 X % C == 0 to X % (unsigned) C == 0 optimization to ...
7261 * match.pd (X % C == 0): ... here. New optimization.
7263 2018-09-12 Jakub Jelinek <jakub@redhat.com>
7266 * expr.h (maybe_optimize_mod_cmp): Declare.
7267 * expr.c (mod_inv): New function.
7268 (maybe_optimize_mod_cmp): New function.
7269 (do_store_flag): Use it.
7270 * cfgexpand.c (expand_gimple_cond): Likewise.
7272 2018-09-09 Cesar Philippidis <cesar@codesourcery.com>
7273 Julian Brown <julian@codesourcery.com>
7276 * gimplify.c (gimplify_scan_omp_clauses): Set
7277 target_firstprivatize_array_bases in OpenACC parallel and kernels
7278 region contexts. Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
7279 OpenACC data regions.
7281 2018-09-12 Uros Bizjak <ubizjak@gmail.com>
7283 * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
7284 (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
7286 2018-09-12 Richard Biener <rguenther@suse.de>
7288 PR tree-optimization/87280
7289 * tree-ssa-sccvn.c (process_bb): Handle the case of executable
7290 edge but unreachable target.
7291 (do_rpo_vn): For conservatively handling a PHI only mark
7292 the backedge executable but not the block reachable.
7294 2018-09-12 Richard Biener <rguenther@suse.de>
7296 PR tree-optimization/87266
7297 * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
7300 2018-09-12 Andreas Krebbel <krebbel@linux.ibm.com>
7302 * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
7304 ("trunc<BFP:mode><DFP_ALL:mode>2")
7305 ("trunc<DFP_ALL:mode><BFP:mode>2")
7306 ("extend<BFP:mode><DFP_ALL:mode>2")
7307 ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
7308 according to the target operand type.
7310 2018-09-12 Jakub Jelinek <jakub@redhat.com>
7311 Andreas Krebbel <krebbel@linux.ibm.com>
7313 PR tree-optimization/86844
7314 * gimple-ssa-store-merging.c
7315 (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
7316 there are any overlapping stores in between them, make sure they are
7317 also coalesced or we give up completely.
7319 2018-09-12 Jakub Jelinek <jakub@redhat.com>
7322 * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
7323 BIT_AND_EXPR's second operand is a power of two. Formatting fix.
7325 2018-09-12 Tom de Vries <tdevries@suse.de>
7327 * common.opt (gdescribe-dies): Add option.
7328 * dwarf2out.c (add_name_and_src_coords_attributes): Add description
7329 attribute for artifical and nameless decls.
7330 (dwarf2out_register_external_die): Add description attribute to
7331 external reference die.
7332 (add_desc_attribute): New functions.
7333 (gen_subprogram_die): Add description attribute to
7334 DW_TAG_call_site_parameter.
7335 * tree-pretty-print.c (print_generic_expr_to_str): New function.
7336 * tree-pretty-print.h (print_generic_expr_to_str): Declare.
7337 * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
7339 (@item -gdescribe-dies): Add.
7341 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
7343 * tree-vrp.c (vrp_shift_undefined_p): Remove.
7344 (extract_range_from_binary_expr_1: Call
7345 wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
7346 * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
7349 2018-09-12 Aldy Hernandez <aldyh@redhat.com>
7351 * gimple-ssa-warn-alloca.c
7352 (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
7353 field for ALLOCA_BOUND_*_LARGE.
7355 2018-09-11 Nathan Sidwell <nathan@acm.org>
7357 * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
7359 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
7361 * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
7362 for clobbers. Remove obsolete comment.
7364 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
7366 * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
7367 mpxchk, mpxld and mpxst types.
7368 (define_attr length_immediate): Remove all processing of mpx types.
7369 (define_attr prefix_0f): Ditto.
7370 (define_attr memory): Ditto.
7372 2018-09-11 Uros Bizjak <ubizjak@gmail.com>
7374 * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
7375 (log<mode>2): Change operand 1 predicate to general_operand.
7376 Extend operand 1 to XFmode and generate logxf3 insn.
7377 (log10<mode>2): Change operand 1 predicate to general_operand.
7378 Extend operand 1 to XFmode and generate log10xf3 insn.
7379 (log2<mode>2): Change operand 1 predicate to general_operand.
7380 Extend operand 1 to XFmode and generate log2xf3 insn.
7381 (fyl2xp1_extend<mode>xf3_i387): Remove.
7382 (log1p<mode>2): Change operand 1 predicate to general_operand.
7383 Extend operand 1 to XFmode and generate log1pxf3 insn.
7384 (fxtract_extend<mode>xf3_i387): Remove.
7385 (logb<mode>2): Change operand 1 predicate to general_operand.
7386 Extend operand 1 to XFmode and generate logbxf3 insn.
7387 (ilogb<mode>2): Change operand 1 predicate to general_operand.
7388 Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
7389 (significand<mode>2): Change operand 1 predicate to general_operand.
7390 Extend operand 1 to XFmode and generate significandxf3 insn.
7392 2018-09-11 Nathan Sidwell <nathan@acm.org>
7394 * gcc.c (perror_with_name, pfatal_with_name): Delete.
7395 (load_specs): Use fatal_error.
7396 (DELETE_IF_ORDINARY, process_command): Use error.
7397 (execute, run_attempt): Use fatal_error.
7399 2018-09-11 Andrew Stubbs <ams@codesourcery.com>
7401 * diagnostic-core.h (sorry_at): New prototype.
7402 * diagnostic.c (sorry_at): New function.
7404 2018-09-11 Aldy Hernandez <aldyh@redhat.com>
7406 * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
7407 by zero as VR_UNDEFINED.
7409 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
7411 * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
7412 (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
7413 (<sincos>mode2): New expander.
7414 (sincos_extend<mode>xf3_i387): Remove insn pattern.
7415 (sincos -> sin, cos splitters): Remove splitter patterns.
7416 (sincos<mode>3): Change operand 2 predicate to general_operand.
7417 Extend operand 2 to XFmode and generate sincosxf3 insn.
7418 (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
7419 Change operand 3 predicate to const1_operand.
7420 (fptan_extend<mode>xf4_i387): Remove insn pattern.
7421 (tanxf2): Update operands in the call to fptanxf4_i387.
7422 (tan<mode>2): Change operand 1 predicate to general_operand.
7423 Extend operand 1 to XFmode and generate tanxf3 insn.
7424 (atan2xf3): Rename from *fpatanxf3_i387.
7425 (fpatan_extend<mode>xf3_i387): Remove insn pattern.
7426 (atan2xf3): Remove expander.
7427 (atan2<mode<3): Change operand 1 and 2 predicates to general_operand.
7428 Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
7429 (atan<mode>2): Change operand 1 predicate to general_operand.
7430 Extend operand 1 to XFmode and generate atanxf3 insn.
7432 2018-09-10 Uros Bizjak <ubizjak@gmail.com>
7434 * config/i386/i386.md (x87/SSE constant load splitter): Use
7435 memory_operand instead of nonimmediate_operand for input operand
7438 2018-09-09 Uros Bizjak <ubizjak@gmail.com>
7440 * config/i386/i386.md (float partial SSE register stall splitter): Move
7441 splitter near its instruction pattern.
7442 (float_extend partial SSE register stall splitter): Ditto.
7443 (float_truncate partial SSE register stall splitter): Ditto.
7445 2018-09-09 Hans-Peter Nilsson <hp@bitrange.com>
7448 * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7449 to speculation_safe_value_not_needed.
7452 * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
7453 non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
7454 (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
7455 leaf_function_p, instead use has_hard_reg_initial_val.
7457 2018-09-09 Nathan Sidwell <nathan@acm.org>
7459 * gcc.h (pfatal_with_name): Don't declare here.
7460 * gcc.c (pfatal_with_name): Make static.
7462 2018-09-09 Xianmiao Qu <xianmiao_qu@c-sky.com>
7464 * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
7467 2018-09-08 John David Anglin <danglin@gcc.gnu.org>
7469 PR rtl-optimization/85458
7470 * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
7471 priority hook to reduce the priority of EXPR.
7473 2018-09-07 Uros Bizjak <ubizjak@gmail.com>
7475 * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
7476 DImode for x87 on 32bit targets. Conditionally disable x87 modes
7477 with X87_ENABLE_FLOAT. Remove preparation code.
7478 (*float<SWI48:mode><MODEF:mode>2): Rename from
7479 *float<SWI48:mode><MODEF:mode>2_mixed. Handle x87, SSE and mixed
7480 math using "enabled" attribute.
7481 (*floatdi<MODEF:mode>2_i387): Rename from
7482 *float<SWI48x:mode><MODEF:mode>2_i387. Handle only DImode and
7483 enable for 32bit targets only.
7484 (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
7486 (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
7487 as operand 1 predicate. Rewrite as define_insn_and_split.
7488 (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
7490 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
7492 * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
7493 to fallthru to FLOAT case.
7495 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
7498 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
7499 around folding of vec_sl to handle out of range shift values.
7501 2018-09-06 Uros Bizjak <ubizjak@gmail.com>
7503 * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
7504 Update callers to gen_fix_trunc<mode>_i387_fisttp
7505 (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
7506 nonimmediate_operand.
7507 (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
7508 and corresponding splitters.
7509 (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
7510 (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
7511 (fix_truncdi_i387_with_temp): Remove insn pattern
7512 and corresponding splitters.
7513 (fix_trunc<mode>_i387): Change operand 0 predicate to
7514 nonimmediate_operand.
7515 (fix_trunc<mode>_i387_with_temp): Remove insn pattern
7516 and corresponding splitters.
7517 (*fistdi2_1): Remove.
7519 (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
7520 (lrintxfdi2): Remove expander. Reimplement as define_insn.
7521 (*fist<mode>2_1): Remove.
7522 (fist<mode>2): Ditto.
7523 (fist<mode>2_with_temp): Remove insn pattern and corresponding
7525 (lrintxf<mode>2): Remove expander. Reimplement as define_insn.
7526 (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
7527 (fistdi2_<rounding>): Change operand 0 predicate to
7528 nonimmediate_operand.
7529 (fistdi2_<rounding>_with_temp): Remove insn pattern
7530 and corresponding splitters.
7531 (fist<mode>2_<rounding>): Change operand 0 predicate to
7532 nonimmediate_operand.
7533 (fist<mode>2_<rounding>_with_temp): Remove insn pattern
7534 and corresponding splitters.
7536 (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
7538 2018-09-06 Bernd Edlinger <bernd.edlinger@hotmail.de>
7540 * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
7543 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
7545 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7546 early gimple folding of vec_splat().
7547 * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
7548 * gimple-fold.h: Add an extern define for tree_vec_extract().
7550 2018-09-06 Will Schmidt <will_schmidt@vnet.ibm.com>
7552 * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
7553 wrappers around TREE_TYPE comparisons.
7555 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
7558 * config/s390/predicates.md: Add nonsym_memory_operand.
7559 * config/s390/s390.c (s390_legitimize_cs_operand): If operand
7560 contains a SYMBOL_REF, load it into an intermediate pseudo.
7561 (s390_emit_compare_and_swap): Legitimize operand.
7562 * config/s390/s390.md: Use the new nonsym_memory_operand
7563 with UNSPECV_CAS patterns.
7565 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
7568 * config/s390/s390-passes.def: New file.
7569 * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
7571 (make_pass_s390_early_mach): Add declaration.
7572 * config/s390/s390.c (make_pass_s390_early_mach):
7573 (s390_option_override): Remove dynamic registration.
7574 * config/s390/t-s390: Add s390-passes.def.
7576 2018-09-06 Ilya Leoshkevich <iii@linux.ibm.com>
7578 * config/s390/s390.c (s390_decompose_constant_pool_ref):
7579 Remove UNSPEC_LTREL_BASE check.
7580 (annotate_constant_pool_refs): Likewise.
7581 (find_constant_pool_ref): Likewise.
7582 (find_ltrel_base): Removed.
7583 (replace_ltrel_base): Removed.
7584 (s390_mainpool_finish): Remove replace_ltrel_base call.
7585 (s390_chunkify_start): Remove pending LTREL_BASE logic.
7586 (s390_chunkify_finish): Remove replace_ltrel_base call.
7587 * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
7589 2018-09-06 Hans-Peter Nilsson <hp@axis.com>
7592 * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7593 to speculation_safe_value_not_needed.
7595 2018-09-05 Cesar Philippidis <cesar@codesourcery.com>
7596 Bernd Schmidt <bernds_cb1@t-online.de>
7598 * config/nvptx/nvptx-opts.h: New file.
7599 * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
7600 * config/nvptx/nvptx.h: Include "nvptx-opts.h".
7602 (TARGET_SM35): New macro.
7603 * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
7605 * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
7607 (misa=): New option.
7608 * doc/invoke.texi (Nvidia PTX Options): Document -misa.
7610 2018-09-05 Uros Bizjak <ubizjak@gmail.com>
7612 * config/i386/i386.md (truncdfsf2): Remove expander.
7613 (truncdfsf2_with_temp): Ditto.
7614 (truncxf<mode>2): Ditto.
7615 (*truncdfsf_fast_mixed): Remove insn pattern.
7616 (*truncdfsf_fast_i387): Ditto.
7617 (*truncdfsf_mixed): Ditto.
7618 (*truncdfsf_i387): Ditto.
7619 (*truncdfsf2_i387_1): Ditto.
7620 (*truncxfsf2_mixed): Ditto.
7621 (*truncxfdf2_mixed): Ditto.
7622 (*truncxf<mode>2_i387_noop): Ditto. Update callers
7623 to call gen_truncxf<mode>2 instead.
7624 (*truncxf<mode>2_i387): Remove.
7625 (reg->reg splitters): Remove splitter pattern.
7626 (reg->mem splitters): Ditto.
7628 (truncdfsf2): New insn pattern.
7629 (truncxf<mode>2): Ditto.
7631 2018-09-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7633 * tree-ssa-math-opts.c (is_mult_by): New function.
7634 (is_square_of): Use the above.
7635 (optimize_recip_sqrt): New function.
7636 (pass_cse_reciprocals::execute): Use the above.
7638 2018-09-05 Richard Biener <rguenther@suse.de>
7641 * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
7642 to zero-init the emplaced vec.
7644 2018-09-05 Martin Liska <mliska@suse.cz>
7646 PR tree-optimization/87205
7647 * tree-switch-conversion.c (pass_lower_switch::execute):
7648 Group cases for switch statements.
7650 2018-09-05 Richard Biener <rguenther@suse.de>
7652 PR tree-optimization/87217
7653 * tree-ssa-sccvn.c (vuse_valueize): New.
7654 (vn_reference_lookup_pieces): Use it.
7655 (vn_reference_lookup): Likewise.
7657 2018-09-05 Nathan Sidwell <nathan@acm.org>
7660 * stor-layout.c (place_field): Scan forwards to check last
7661 bitfield when ms_bitfield_placement is in effect.
7663 2018-09-05 Richard Biener <rguenther@suse.de>
7666 * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
7669 2018-09-05 Siddhesh Poyarekar <siddhesh@sourceware.org>
7670 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
7672 * config/aarch64/falkor-tag-collision-avoidance.c: New file.
7673 * config.gcc (extra_objs): Build it.
7674 * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
7676 * config/aarch64/aarch64-passes.def
7677 (pass_tag_collision_avoidance): New pass.
7678 * config/aarch64/aarch64.c (qdf24xx_tunings): Add
7679 AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
7680 (aarch64_classify_address): Remove static qualifier.
7681 (aarch64_address_info, aarch64_address_type): Move to...
7682 * config/aarch64/aarch64-protos.h: ... here.
7683 (make_pass_tag_collision_avoidance): New function.
7684 * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
7687 2018-09-05 Martin Liska <mliska@suse.cz>
7689 * doc/gcov.texi: Update documentation of humar
7691 * gcov.c (format_count): Print one decimal place, it provides
7692 more fine number of situations like '1G' vs. '1.4G'.
7694 2018-09-05 Martin Liska <mliska@suse.cz>
7697 * config/rs6000/rs6000.opt: Mark the option as Deprecated.
7698 * optc-gen.awk: Allow 'Var' for Deprecated options in order
7699 to generate a MASK value.
7701 2018-09-04 H.J. Lu <hongjiu.lu@intel.com>
7704 * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
7705 if frame pointer isn't used.
7706 (compute_frame_pointer_to_fb_displacement): Likewise.
7708 2018-09-04 Jakub Jelinek <jakub@redhat.com>
7711 * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
7712 OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
7713 OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
7714 (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
7715 and OPTION_MASK_ISA_XSAVEC_UNSET.
7717 2018-09-04 Max Filippov <jcmvbkbc@gmail.com>
7719 * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
7720 XOR operations in NAND case.
7722 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
7724 * wide-int-range.cc (wide_int_range_convert): New.
7725 * wide-int-range.h (wide_int_range_convert): New.
7726 * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
7727 code into wide_int_range_convert.
7728 (extract_range_into_wide_ints): Do not munge anti range constants
7729 into the entire domain. Just return the range back.
7731 2018-09-04 Martin Liska <mliska@suse.cz>
7733 * genmatch.c (output_line_directive): Add new argument
7735 (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
7737 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
7739 * doc/invoke.texi (Option Summary): Add whitespace.
7741 * doc/invoke.texi (Option Summary): Add -Waligned-new.
7743 2018-09-04 Richard Biener <rguenther@suse.de>
7745 PR tree-optimization/87211
7746 * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
7747 backedge value we're supposed to treat as VARYING also number
7748 the PHI to VARYING in case it got a different value-number already.
7750 2018-09-04 Aldy Hernandez <aldyh@redhat.com>
7752 * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
7753 (extract_range_from_binary_expr_1): Do not call
7754 vrp_can_optimize_bit_op.
7755 * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
7757 (wide_int_range_get_mask_and_bounds): New.
7758 (wide_int_range_optimize_bit_op): New.
7759 (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
7760 (wide_int_range_bit_and): Same.
7761 * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
7762 (wide_int_range_optimize_bit_op): New.
7763 (wide_int_range_get_mask_and_bounds): New.
7765 2018-09-04 Richard Biener <rguenther@suse.de>
7767 PR tree-optimization/87176
7768 * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
7769 variable. When value-numbering a virtual PHI node make sure
7770 to not value-number to the backedge value.
7772 2018-09-04 Jonathan Wakely <jwakely@redhat.com>
7774 * doc/extend.texi (Long Long, Hex Floats): Document support for
7775 long long and hex floats in more recent versions of ISO C++.
7777 2018-09-03 Richard Biener <rguenther@suse.de>
7779 PR tree-optimization/87177
7780 * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
7783 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
7785 * bb-reorder.c (edge_order): Convert to C-qsort-style
7786 tri-state comparator.
7787 (reorder_basic_blocks_simple): Change std::stable_sort to
7790 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
7792 * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
7793 tri-state comparator.
7794 (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
7796 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
7798 * sort.cc (struct sort_ctx): New field 'nlim'. Use it...
7799 (mergesort): ... here as maximum count for using netsort.
7800 (gcc_qsort): Set nlim to 3 if stable sort is requested.
7801 (gcc_stablesort): New.
7802 * system.h (gcc_stablesort): Declare.
7804 2018-09-03 Alexander Monakov <amonakov@ispras.ru>
7806 * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
7807 * system.h (qsort): Always redirect to gcc_qsort. Update comment.
7808 * vec.c (qsort_chk): Do not call gcc_qsort. Update comment.
7810 2018-09-03 Segher Boessenkool <segher@kernel.crashing.org>
7812 * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
7813 lxsdx and stxsdx alternatives.
7814 (*mov<mode>_hardfloat64): Ditto.
7815 * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
7817 2018-09-03 Richard Biener <rguenther@suse.de>
7819 PR tree-optimization/87200
7820 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
7823 2018-09-03 Martin Liska <mliska@suse.cz>
7825 PR tree-optimization/87201
7826 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7827 Fix parenthesis in an expression.
7829 2018-09-03 Richard Biener <rguenther@suse.de>
7831 PR tree-optimization/87197
7832 * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
7833 visited. CSE the VN_INFO hashtable lookup.
7835 PR tree-optimization/87169
7836 * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
7837 iterating make sure there's no extra backedges from irreducible
7838 regions feeding the header. Mark the destination block
7841 2018-09-03 Martin Liska <mliska@suse.cz>
7844 * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
7845 * common/common-targhooks.c (default_get_valid_option_values):
7847 * common/common-targhooks.h (default_get_valid_option_values):
7849 * common/config/i386/i386-common.c: Move processor_target_table
7851 (ix86_get_valid_option_values): New function.
7852 (TARGET_GET_VALID_OPTION_VALUES): New macro.
7853 * config/i386/i386.c (struct ptt): Move to i386-common.c.
7854 (PTA_*): Move all defined masks into i386-common.c.
7855 (ix86_function_specific_restore): Use new processor_cost_table.
7856 * config/i386/i386.h (struct ptt): Moved from i386.c.
7857 (struct pta): Likewise.
7858 * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
7859 * doc/tm.texi.in: Likewise.
7860 * opt-suggestions.c (option_proposer::suggest_option):
7861 Pass prefix to build_option_suggestions.
7862 (option_proposer::get_completions): Likewise.
7863 (option_proposer::build_option_suggestions): Use the new target
7865 * opts.c (struct option_help_tuple): New struct.
7866 (print_filtered_help): Use the new target hook.
7868 2018-09-03 Martin Liska <mliska@suse.cz>
7871 * predict.c (set_even_probabilities): Add likely_edges
7872 argument and handle cases where we have precisely one
7874 (combine_predictions_for_bb): Catch also likely_edges.
7875 (tree_predict_by_opcode): Handle gswitch statements.
7876 * tree-cfg.h (find_case_label_for_value): New declaration.
7877 (find_taken_edge_switch_expr): Likewise.
7878 * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7879 Find pivot in decision tree based on probabily, not by number of
7882 2018-09-02 Gerald Pfeifer <gerald@pfeifer.com>
7884 * doc/standards.texi (Standards): Update Objective-C reference.
7886 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
7888 * doc/install.texi (Prerequisites): Update link for MPC.
7890 2018-09-01 Michael Matz <matz@suse.de>
7892 PR tree-optimization/87074
7893 * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
7894 PHIs for outer-loop uses.
7896 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
7898 * doc/generic.texi (OpenMP): Adjust link to openmp.org.
7899 * doc/invoke.texi (C Dialect Options): Ditto.
7901 2018-09-01 Gerald Pfeifer <gerald@pfeifer.com>
7903 * doc/install.texi (Prerequisites): Adjust link mpfr.org.
7905 2018-08-31 Richard Biener <rguenther@suse.de>
7907 PR tree-optimization/87168
7908 * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
7909 (rpo_elim::eliminate_avail): When OP was not visited it must
7912 2018-08-31 David Malcolm <dmalcolm@redhat.com>
7914 * tree-vrp.c (copy_value_range): Convert param "from" from
7915 "value_range *" to "const value_range *".
7916 (range_is_null): Likewise for param "vr".
7917 (range_int_cst_p): Likewise.
7918 (range_int_cst_singleton_p): Likewise.
7919 (symbolic_range_p): Likewise.
7920 (value_ranges_intersect_p): Likewise for both params.
7921 (value_range_nonnegative_p): Likewise for param "vr".
7922 (value_range_constant_singleton): Likewise.
7923 (vrp_set_zero_nonzero_bits): Likewise for param "ar".
7924 (extract_range_into_wide_ints): Likewise for param "vr".
7925 (extract_range_from_multiplicative_op): Likewise for params "vr0"
7927 (vrp_can_optimize_bit_op): Likewise.
7928 (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
7930 (extract_range_from_unary_expr): Likewise.
7931 (debug_value_range): Likewise for param "vr".
7932 (value_range::dump): Add "const" qualifier.
7933 (vrp_prop::check_array_ref): Convert local "vr" from
7934 "value_range *" to "const value_range *".
7935 (vrp_prop::check_mem_ref): Likewise.
7936 (vrp_prop::visit_stmt): Likewise for local "old_vr".
7937 (vrp_intersect_ranges_1): Likewise for param "vr_1".
7938 (vrp_intersect_ranges): Likewise.
7939 (simplify_stmt_for_jump_threading): Likewise for local "vr".
7940 (vrp_prop::vrp_finalize): Likewise.
7941 * tree-vrp.h (value_range::dump): Add "const" qualifier.
7942 (vrp_intersect_ranges): Add "const" qualifier to params as above.
7943 (extract_range_from_unary_expr): Likewise.
7944 (value_range_constant_singleton): Likewise.
7945 (symbolic_range_p): Likewise.
7946 (copy_value_range): Likewise.
7947 (extract_range_from_binary_expr_1): Likewise.
7948 (range_int_cst_p): Likewise.
7949 (vrp_set_zero_nonzero_bits): Likewise.
7950 (range_int_cst_singleton_p): Likewise.
7952 2018-08-31 Vlad Lazar <vlad.lazar@arm.com>
7954 * config/aarch64/arm_neon.h (vabsd_s64): New.
7955 (vnegd_s64): Likewise.
7957 2018-08-31 Martin Jambor <mjambor@suse.cz>
7959 * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
7961 2018-08-31 Martin Liska <mliska@suse.cz>
7963 * ipa-icf.c (sem_item::add_type): Use
7964 sem_item::m_type_hash_cache.
7965 * ipa-icf.h: Move the cache from sem_item_optimizer
7968 2018-08-31 Nathan Sidwell <nathan@acm.org>
7970 * doc/extend.texi (Backwards Compatibility): Remove implicit
7971 extern C leeway of () being (...).
7973 2018-08-31 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7975 * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
7977 2018-08-31 Segher Boessenkool <segher@kernel.crashing.org>
7981 * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
7983 2018-08-31 Jakub Jelinek <jakub@redhat.com>
7986 * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
7987 gen_int_mode. Formatting fixes.
7989 2018-08-30 Sandra Loosemore <sandra@codesourcery.com>
7991 * target.def (custom_function_descriptors): Improve documentation.
7992 * doc/tm.texi.in (Trampolines): Expand discussion of function
7993 descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
7994 beginning of the section.
7995 * doc/tm.texi: Regenerated.
7997 2018-08-30 Jose E. Marchesi <jose.marchesi@oracle.com>
7999 * cfg.h (class auto_edge_flag): Spell out the template-id of the
8000 base class in the initializer list. This is a workaround for
8001 building with older GCC.
8002 (class auto_bb_flag): Likewise.
8004 2018-08-30 Aaron Sawdey <acsawdey@linux.ibm.com>
8006 * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
8007 (altivec_vcmpequ<VI_char>_p): Remove star.
8008 * config/rs6000/rs6000-string.c (do_load_for_compare): Support
8010 (expand_strncmp_vec_sequence): New function.
8011 (emit_final_str_compare_vec): New function.
8012 (expand_strn_compare): Add support for vector strncmp.
8013 * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
8014 length specification to bytes.
8015 * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
8016 (vcmpnezb_p): New pattern.
8017 * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
8018 for option -mstring-compare-inline-limit.
8020 2018-08-30 Thiago Macieira <thiago.macieira@intel.com>
8022 * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
8023 (PTA_SKYLAKE): Add PTA_AES.
8024 (PTA_GOLDMONT): Likewise.
8026 2018-08-29 Jan Hubicka <jh@suse.cz>
8029 * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
8030 * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
8032 2018-08-29 Jan Hubicka <jh@suse.cz>
8034 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
8036 (hash_tree): Do not visit TYPE_STUB_DECL.
8037 * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
8038 stream TYPE_STUB_DECL.
8039 * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
8040 * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
8041 after free_lang_data.
8042 (type_in_anonymous_namespace_p): Likewise.
8044 2018-08-29 Jan Hubicka <jh@suse.cz>
8046 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
8047 comment that it has to be even number.
8048 (class sreal): Change m_sig type to int32_t.
8049 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
8050 int64_t for temporary calculations.
8051 (sreal_verify_basics): Drop one bit from minimum and maximum.
8053 2018-08-30 Richard Biener <rguenther@suse.de>
8055 PR tree-optimization/87147
8056 * tree-ssa-sccvn.c (SSA_VISITED): New function.
8057 (visit_phi): When the degenerate result is from the backedge and
8058 we didn't visit its definition yet drop to VARYING.
8059 (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
8061 2018-08-29 Jan Hubicka <jh@suse.cz>
8063 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
8065 (hash_tree): Likewise.
8067 2018-08-29 Jan Hubicka <jh@suse.cz>
8069 * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
8070 and TYPE_NEXT_REF_TO.
8072 2018-08-29 Jan Hubicka <jh@suse.cz>
8074 * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
8075 comment that it has to be even number.
8076 (class sreal): Change m_sig type to int32_t.
8077 * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
8078 int64_t for temporary calculations.
8079 (sreal_verify_basics): Drop one bit from minimum and maximum.
8081 2018-08-30 Tamar Christina <tamar.christina@arm.com>
8083 * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
8085 2018-08-30 Vlad Lazar <vlad.lazar@arm.com>
8088 * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
8089 if to_add is negative.
8091 2018-08-29 Bernd Edlinger <bernd.edlinger@hotmail.de>
8094 * builtins.c (c_strlen): Improve range checks.
8096 2018-08-29 Martin Sebor <msebor@redhat.com>
8097 Jeff Law <law@redhat.com>
8099 PR tree-optimization/86714
8100 PR tree-optimization/86711
8101 * builtins.c (c_strlen): Add arguments to call to string_constant.
8102 * expr.c (string_constant): Add argument. Detect missing nul
8103 terminator and outermost declaration it's missing in.
8104 * expr.h (string_constant): Add argument.
8105 * fold-const.c (read_from_constant_string): Add arguments to call to
8107 (c_getstr): Likewise.
8108 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
8110 * tree-ssa-strlen.c (get_stridx): Likewise.
8112 2018-08-29 Jan Hubicka <jh@suse.cz>
8114 * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
8115 Do not stream DECL_VINDEX.
8116 * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
8117 * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
8118 (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
8120 2018-08-29 Richard Biener <rguenther@suse.de>
8122 * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
8123 virtual operands that are not default defs to honor region
8125 (rpo_vn_valueize): Remove ineffective code here.
8127 2018-08-29 Richard Biener <rguenther@suse.de>
8129 PR tree-optimization/87132
8130 * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
8131 when skipping defs reachable over backedges.
8133 2018-08-29 Richard Biener <rguenther@suse.de>
8135 * tree-core.h: Document use of deprecated_flag in SSA_NAME.
8136 * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
8137 * tree-into-ssa.c (pass_build_ssa::execute): Initialize
8138 function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
8139 * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
8140 (vn_reference_lookup_3): Remove use of const_parms.
8141 (free_rpo_vn): Do not free const_parms.
8142 (do_rpo_vn): Do not call init_const_parms.
8143 * tree-ssa-alias.c (refs_may_alias_p_1): Honor
8144 SSA_NAME_POINTS_TO_READONLY_MEMORY.
8145 (call_may_clobber_ref_p_1): Likewise.
8147 2018-08-29 Alexander Monakov <amonakov@ispras.ru>
8150 * invoke.texi (Optimization Options): List -ftree-scev-cprop.
8152 (-ftree-scev-cprop): Document.
8154 2018-08-29 Jan Hubicka <jh@suse.cz>
8156 * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
8158 (sreal constructor): Update.
8159 * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
8160 sreal:operator/): Update.
8162 2018-08-29 Martin Liska <mliska@suse.cz>
8164 * tree-switch-conversion.c (switch_conversion::expand):
8165 Strenghten assumption about gswitch statements.
8167 2018-08-29 Richard Biener <rguenther@suse.de>
8169 PR tree-optimization/87117
8170 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
8171 re-value-number released SSA VDEFs.
8173 2018-08-29 Richard Biener <rguenther@suse.de>
8175 PR tree-optimization/87126
8176 * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
8178 2018-08-28 Jim Wilson <jimw@sifive.com>
8180 * config/riscv/pic.md: Rewrite.
8181 * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
8183 * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
8184 (SOFTF, default_load, softload, softstore): New.
8186 2018-08-28 Jeff Law <law@redhat.com>
8188 * fold-const.c (fold_binary_loc): Remove recently added assert.
8190 2018-08-28 Joern Rennecke <joern.rennecke@riscy-ip.com>
8192 * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
8193 to OP parmeter of generated function.
8195 2018-08-28 MCC CS <deswurstes@users.noreply.github.com>
8197 PR tree-optimization/87009
8198 * match.pd: Add boolean optimizations.
8200 2018-08-28 Martin Sebor <msebor@redhat.com>
8203 * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
8204 * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
8205 (pass_walloca::gate): Use it.
8206 (alloca_call_type): Same.
8207 (pass_walloca::execute): Same.
8208 * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
8210 2018-08-28 David Malcolm <dmalcolm@redhat.com>
8212 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
8213 GCC_VERSION for usage of "__gcc_dump_printf__" format from
8216 2018-08-28 Richard Biener <rguenther@suse.de>
8218 PR tree-optimization/87124
8219 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
8220 constants before looking up avail.
8222 2018-08-28 Jakub Jelinek <jakub@redhat.com>
8225 * calls.c (maybe_warn_nonstring_arg): Punt early if
8226 warn_stringop_overflow is zero. Don't call get_range_strlen
8227 on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
8228 Swap comparison operands to have constants on rhs. Only use
8229 lenrng[1] if non-NULL and INTEGER_CST. Don't uselessly
8230 increment lenrng[0].
8232 2018-08-28 Richard Sandiford <richard.sandiford@arm.com>
8234 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
8235 use of tree_to_shwi. Remove duplicated test for the size being
8236 a whole number of bytes.
8238 2018-08-28 Richard Biener <rguenther@suse.de>
8240 PR tree-optimization/87117
8241 * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
8242 Handle removed stmt without LHS (GIMPLE_NOP).
8244 2018-08-28 Richard Biener <rguenther@suse.de>
8246 PR tree-optimization/87117
8247 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
8248 void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
8250 2018-08-28 Richard Biener <rguenther@suse.de>
8252 PR tree-optimization/87117
8253 * tree-ssa-pre.c (compute_avail): Do not make expressions
8254 with predicated values available.
8255 (get_expr_value_id): Assert we do not run into predicated value
8258 2018-08-28 Richard Biener <rguenther@suse.de>
8260 PR tree-optimization/87117
8261 * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
8262 get virtual operands.
8263 (get_expr_operands): Handle STRING_CST like other decls.
8265 2018-08-28 Martin Liska <mliska@suse.cz>
8267 * tree.h: Update documentation of fndecl_built_in_p
8271 2018-08-27 Jeff Law <law@redhat.com>
8272 PR tree-optimization/87110
8273 * tree-ssa-dse.c (compute_trims): Handle non-constant
8276 2018-08-27 Martin Sebor <msebor@redhat.com>
8278 PR tree-optimization/86914
8279 * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
8281 2018-08-27 Martin Sebor <msebor@redhat.com>
8283 PR tree-optimization/87112
8284 * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
8285 the type of the bound argument.
8287 2018-08-27 Jeff Law <law@redhat.com>
8289 * tree-ssa-dse.c (compute_trims): Handle case where the reference's
8290 type does not have a TYPE_SIZE_UNIT.
8292 2018-08-27 Steve Ellcey <sellcey@cavium.com>
8294 * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
8295 with include of backend.h.
8297 2018-08-27 Richard Biener <rguenther@suse.de>
8299 PR tree-optimization/86927
8300 * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
8301 use const cond reduction code.
8303 2018-08-27 Alexander Monakov <amonakov@ispras.ru>
8305 PR tree-optimization/85758
8306 * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
8308 2018-08-27 David Malcolm <dmalcolm@redhat.com>
8311 * diagnostic-show-locus.c (class layout_range): Update for
8312 conversion of show_caret_p to a tri-state.
8313 (layout_range::layout_range): Likewise.
8314 (make_range): Likewise.
8315 (layout::maybe_add_location_range): Likewise.
8316 (layout::should_print_annotation_line_p): Don't show annotation
8317 lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
8318 (layout::get_state_at_point): Update for conversion of
8319 show_caret_p to a tri-state. Bail out early for
8320 SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
8321 underlining or source colorization.
8322 (gcc_rich_location::add_location_if_nearby): Update for conversion
8323 of show_caret_p to a tri-state.
8324 (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
8325 (selftest::test_one_liner_fixit_replace_equal_secondary_range):
8327 (selftest::test_one_liner_labels): Likewise.
8328 * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
8329 conversion of show_caret_p to a tri-state.
8330 * pretty-print.c (text_info::set_location): Likewise.
8331 * pretty-print.h (text_info::set_location): Likewise.
8332 * substring-locations.c (format_warning_n_va): Likewise.
8333 * tree-diagnostic.c (default_tree_printer): Likewise.
8334 * tree-pretty-print.c (newline_and_indent): Likewise.
8336 2018-08-27 David Malcolm <dmalcolm@redhat.com>
8339 * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
8340 line above for line-insertion fix-it hints.
8341 (selftest::test_fixit_insert_containing_newline): Update the
8342 expected results, and add a test with line-numbering enabled.
8344 2018-08-27 Martin Liska <mliska@suse.cz>
8347 * sanopt.c (sanitize_rewrite_addressable_params): Ignore
8348 params with DECL_HAS_VALUE_EXPR_P.
8350 2018-08-27 Martin Liska <mliska@suse.cz>
8352 * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
8353 selected expansion strategy.
8355 2018-08-27 Martin Liska <mliska@suse.cz>
8357 * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
8358 * builtins.c (is_builtin_fn): Likewise.
8359 * attribs.c (diag_attr_exclusions): Use new function
8360 fndecl_built_in_p and remove check for FUNCTION_DECL if
8362 (builtin_mathfn_code): Likewise.
8363 (fold_builtin_expect): Likewise.
8364 (fold_call_expr): Likewise.
8365 (fold_builtin_call_array): Likewise.
8366 (fold_call_stmt): Likewise.
8367 (set_builtin_user_assembler_name): Likewise.
8368 (is_simple_builtin): Likewise.
8369 * calls.c (gimple_alloca_call_p): Likewise.
8370 (maybe_warn_nonstring_arg): Likewise.
8371 * cfgexpand.c (expand_call_stmt): Likewise.
8372 * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
8373 (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
8374 (cgraph_node::verify_node): Likewise.
8375 * cgraphclones.c (build_function_decl_skip_args): Likewise.
8376 (cgraph_node::create_clone): Likewise.
8377 * config/arm/arm.c (arm_insert_attributes): Likewise.
8378 * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
8379 * dse.c (scan_insn): Likewise.
8380 * expr.c (expand_expr_real_1): Likewise.
8381 * fold-const.c (operand_equal_p): Likewise.
8382 (fold_binary_loc): Likewise.
8383 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
8384 * gimple-low.c (lower_stmt): Likewise.
8385 * gimple-pretty-print.c (dump_gimple_call): Likewise.
8386 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8388 * gimple.c (gimple_build_call_from_tree): Likewise.
8389 (gimple_call_builtin_p): Likewise.
8390 (gimple_call_combined_fn): Likewise.
8391 * gimplify.c (gimplify_call_expr): Likewise.
8392 (gimple_boolify): Likewise.
8393 (gimplify_modify_expr): Likewise.
8394 (gimplify_addr_expr): Likewise.
8395 * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
8396 * ipa-cp.c (determine_versionability): Likewise.
8397 * ipa-fnsummary.c (compute_fn_summary): Likewise.
8398 * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
8399 * ipa-split.c (visit_bb): Likewise.
8400 (split_function): Likewise.
8401 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
8402 * lto-cgraph.c (input_node): Likewise.
8403 * lto-streamer-out.c (write_symbol): Likewise.
8404 * omp-low.c (setjmp_or_longjmp_p): Likewise.
8405 (lower_omp_1): Likewise.
8406 * predict.c (strip_predict_hints): Likewise.
8407 * print-tree.c (print_node): Likewise.
8408 * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
8409 * trans-mem.c (is_tm_irrevocable): Likewise.
8410 (is_tm_load): Likewise.
8411 (is_tm_simple_load): Likewise.
8412 (is_tm_store): Likewise.
8413 (is_tm_simple_store): Likewise.
8414 (is_tm_abort): Likewise.
8415 (tm_region_init_1): Likewise.
8416 * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
8417 * tree-cfg.c (verify_gimple_call): Likewise.
8418 (move_stmt_r): Likewise.
8419 (stmt_can_terminate_bb_p): Likewise.
8420 * tree-eh.c (lower_eh_constructs_2): Likewise.
8421 * tree-if-conv.c (if_convertible_stmt_p): Likewise.
8422 * tree-inline.c (remap_gimple_stmt): Likewise.
8423 (copy_bb): Likewise.
8424 (estimate_num_insns): Likewise.
8425 (fold_marked_statements): Likewise.
8426 * tree-sra.c (scan_function): Likewise.
8427 * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
8428 (optimize_stack_restore): Likewise.
8429 (pass_fold_builtins::execute): Likewise.
8430 * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
8431 (mark_all_reaching_defs_necessary_1): Likewise.
8432 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
8433 * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
8434 (pass_forwprop::execute): Likewise.
8435 * tree-ssa-loop-im.c (stmt_cost): Likewise.
8436 * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
8437 * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
8438 * tree-ssa-strlen.c (get_string_length): Likewise.
8439 * tree-ssa-structalias.c (handle_lhs_call): Likewise.
8440 (find_func_aliases_for_call): Likewise.
8441 * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
8442 * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
8443 * tree-tailcall.c (find_tail_calls): Likewise.
8444 * tree.c (need_assembler_name_p): Likewise.
8445 (free_lang_data_in_decl): Likewise.
8446 (get_call_combined_fn): Likewise.
8447 * ubsan.c (is_ubsan_builtin_p): Likewise.
8448 * varasm.c (incorporeal_function_p): Likewise.
8449 * tree.h (DECL_BUILT_IN): Remove and replace with
8451 (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
8452 (fndecl_built_in_p): New.
8454 2018-08-27 Martin Liska <mliska@suse.cz>
8456 PR tree-optimization/86847
8457 * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
8458 Dump also subtree probability.
8459 (switch_decision_tree::do_jump_if_equal): New function.
8460 (switch_decision_tree::emit_case_nodes): Handle special
8461 situations in balanced tree that can be emitted much simpler.
8462 Fix calculation of probabilities that happen in tree expansion.
8463 * tree-switch-conversion.h (struct cluster): Add
8465 (struct simple_cluster): Likewise.
8466 (struct case_tree_node): Add new function has_child.
8467 (do_jump_if_equal): New.
8469 2018-08-27 Martin Liska <mliska@suse.cz>
8471 * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
8472 Add new argument to bit_test_cluster constructor.
8473 (bit_test_cluster::emit): Set bits really number of values
8475 (bit_test_cluster::hoist_edge_and_branch_if_true): Add
8476 probability argument.
8477 * tree-switch-conversion.h (struct bit_test_cluster):
8478 Add m_handles_entire_switch.
8480 2018-08-27 Martin Liska <mliska@suse.cz>
8482 PR tree-optimization/86702
8483 * tree-switch-conversion.c (jump_table_cluster::emit):
8484 Make probabilities even for values in jump table
8485 according to number of cases handled.
8486 (switch_decision_tree::compute_cases_per_edge): Pass
8487 argument to reset_out_edges_aux function.
8488 (switch_decision_tree::analyze_switch_statement): Likewise.
8489 * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
8492 2018-08-27 Martin Liska <mliska@suse.cz>
8494 * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
8495 cfun argument explicitly.
8496 * gimple-pretty-print.c (dump_gimple_switch): Likewise.
8497 * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
8498 function gimple_switch_default_bb.
8499 (convert_switch_statements):
8501 * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
8502 * stmt.c (label_to_block_fn): Use label_to_block and pass
8503 cfun argument explicitly and use gimple_switch_label_bb.
8504 (expand_case): Likewise.
8505 * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
8506 cfun argument explicitly. Likewise.
8507 (make_edges_bb): Likewise.
8508 (make_cond_expr_edges): Likewise.
8509 (get_cases_for_edge): Likewise.
8510 (make_gimple_switch_edges): Likewise.
8511 (label_to_block_fn): Likewise.
8512 (label_to_block): Likewise.
8513 (make_goto_expr_edges): Likewise.
8514 (make_gimple_asm_edges): Likewise.
8515 (main_block_label): Likewise.
8516 (group_case_labels_stmt): Likewise.
8517 (find_taken_edge_computed_goto): Likewise.
8518 (find_taken_edge_switch_expr): Likewise.
8519 (gimple_verify_flow_info): Likewise.
8520 (gimple_redirect_edge_and_branch): Likewise.
8521 (gimple_switch_label_bb): New function.
8522 (gimple_switch_default_bb): Likewise.
8523 (gimple_switch_edge): Likewise.
8524 (gimple_switch_default_edge): Likewise.
8525 * tree-cfg.h (label_to_block_fn): Remove and replace ...
8526 (label_to_block): ... with this.
8527 (gimple_switch_label_bb): New.
8528 (gimple_switch_default_bb): Likewise.
8529 (gimple_switch_edge): Likewise.
8530 (gimple_switch_default_edge): Likewise.
8531 * tree-cfgcleanup.c (convert_single_case_switch): Use
8532 new gimple functions and pass new argument to label_to_block.
8533 (cleanup_control_flow_bb):
8534 * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
8535 cfun argument explicitly.
8536 (make_eh_edges): Likewise.
8537 (redirect_eh_dispatch_edge): Likewise.
8538 (lower_resx): Likewise.
8539 (lower_eh_dispatch): Likewise.
8540 (maybe_remove_unreachable_handlers): Likewise.
8541 (unsplit_eh): Likewise.
8542 (cleanup_empty_eh): Likewise.
8543 (verify_eh_edges): Likewise.
8544 (verify_eh_dispatch_edge): Likewise.
8545 * tree-ssa-dom.c (record_edge_info): Likewise.
8546 * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
8547 * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
8548 (thread_through_normal_block): Likewise.
8549 * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
8550 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
8551 * tree-switch-conversion.c (switch_conversion::collect): Use new
8553 (switch_conversion::check_final_bb): Likewise.
8554 (switch_conversion::gather_default_values): Pass new argument
8556 (switch_conversion::build_constructors): Likewise.
8557 (switch_decision_tree::compute_cases_per_edge): Use new
8558 gimple_switch_edge function.
8559 (switch_decision_tree::analyze_switch_statement): Pass new argument
8561 (switch_decision_tree::try_switch_expansion): Use
8562 gimple_switch_default_edge.
8563 * tree-vrp.c (find_switch_asserts): Pass new argument
8565 * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
8566 (vr_values::simplify_switch_using_ranges): Likewise.
8568 2018-08-27 Richard Biener <rguenther@suse.de>
8570 * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
8571 * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
8573 * tree-ssa-sccvn.h (struct vn_pval): New structure.
8574 (struct vn_nary_op_s): Add unwind_to member. Add
8575 predicated_values flag and put result into a union together
8576 with a linked list of vn_pval.
8577 (struct vn_ssa_aux): Add name member to make maintaining
8578 a map of SSA name to vn_ssa_aux possible. Remove no longer
8579 needed info, dfsnum, low, visited, on_sccstack, use_processed
8580 and range_info_anti_range_p members.
8581 (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
8582 (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
8584 (vn_valueize): New global.
8585 (vn_context_bb): Likewise.
8586 (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
8587 VN_INFO_PTR_INFO): Remove.
8588 * tree-ssa-sccvn.c: ... (rewrite)
8589 (pass_fre::execute): For -O2+ initialize loops and run
8590 RPO VN in optimistic mode (iterating). For -O1 and -Og
8591 run RPO VN in non-optimistic mode.
8592 * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
8593 (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
8594 * doc/invoke.texi (sccvn-max-scc-size): Remove.
8595 (rpo-vn-max-loop-depth): Document.
8596 * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
8597 when valuezing the VUSE signals we walked out of the region.
8598 * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
8599 (phi_translate): Set VN context block to use for availability
8601 (compute_avail): Likewise.
8602 (pre_valueize): New function.
8603 (pass_pre::execute): Adjust to the RPO VN API.
8605 * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
8606 (propagate_constants_for_unrolling): Remove.
8607 (tree_unroll_loops_completely): Perform value-numbering
8608 on the unrolled bodies loop parent.
8610 2018-08-27 Richard Biener <rguenther@suse.de>
8612 * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
8613 for partial antic compute.
8615 2018-08-27 Jakub Jelinek <jakub@redhat.com>
8617 PR rtl-optimization/87065
8618 * combine.c (simplify_if_then_else): Formatting fix.
8619 (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
8621 (known_cond): Don't return const_true_rtx for vector modes. Use
8622 CONST0_RTX instead of const0_rtx. Formatting fixes.
8624 2018-08-27 Martin Liska <mliska@suse.cz>
8626 PR gcov-profile/87069
8627 * gcov.c (process_file): Record files already processed
8628 and warn about a file being processed multiple times.
8630 2018-08-27 Martin Liska <mliska@suse.cz>
8633 * config/aarch64/aarch64.c (aarch64_override_options_internal):
8634 Set default values for x_aarch64_*_string strings.
8635 * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
8636 prefix. For -mabi do not print '=ABI' in help and use
8637 <option_value> format for -msve-vector-bits and -moverride
8640 2018-08-26 Jeff Law <law@redhat.com>
8642 * config/mips/frame-header-opt.c: Include "backend.h" rather than
8645 2018-08-26 Marek Polacek <polacek@redhat.com>
8647 PR c++/87029, Implement -Wredundant-move.
8648 * doc/invoke.texi: Document -Wredundant-move.
8650 2018-08-25 Martin Sebor <msebor@redhat.com>
8652 PR tree-optimization/87059
8653 * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
8654 to the same type as the other.
8655 * fold-const.c (fold_binary_loc): Assert expectation.
8657 2018-08-25 Iain Sandoe <iain@sandoe.co.uk>
8659 * config/darwin.c (machopic_legitimize_pic_address): Clean up
8660 extraneous parentheses, dead code section and formatting.
8662 2018-08-24 David Malcolm <dmalcolm@redhat.com>
8665 * diagnostic-show-locus.c (layout::layout): Ensure the margin is
8666 wide enough for jumps in the line-numbering to be visible.
8667 (layout::print_gap_in_line_numbering): New member function.
8668 (layout::calculate_line_spans): When using line numbering, merge
8669 line spans that are only 1 line apart.
8670 (diagnostic_show_locus): When printing line numbers, show gaps in
8671 line numbering directly, rather than printing headers.
8672 (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
8673 line-numbering with multiple line spans.
8674 (selftest::test_fixit_insert_containing_newline_2): Add test of
8675 line-numbering, in which the spans are close enough to be merged.
8677 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
8679 * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
8680 to range_includes_zero_p. Do not special case VR_ANTI_RANGE.
8681 * tree-vrp.c (range_is_nonnull): Remove.
8682 (range_includes_zero_p): Accept value_range instead of min/max.
8683 (extract_range_from_binary_expr_1): Do not early bail on
8685 Use range_includes_zero_p instead of range_is_nonnull.
8686 (extract_range_from_unary_expr): Use range_includes_zero_p instead
8687 of range_is_nonnull.
8688 (vrp_meet_1): Pass value_range to range_includes_zero_p. Do not
8689 special case VR_ANTI_RANGE.
8690 (vrp_finalize): Same.
8691 * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
8693 (range_is_nonnull): Remove.
8694 * vr-values.c (vrp_stmt_computes_nonzero): Use
8695 range_includes_zero_p instead of range_is_nonnull.
8696 (extract_range_basic): Pass value_range to range_includes_zero_p
8697 instead of range_is_nonnull.
8699 2018-08-24 Uros Bizjak <ubizjak@gmail.com>
8701 * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
8702 * emit-rtl.h (rtl_data): Remove return_bnd.
8703 * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
8704 * function.c (diddle_return_value): Do not handle crtl->return_bnd.
8705 * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
8706 (POINTER_BOUNDS_MODE): Remove definition.
8707 (make_pointer_bounds_mode): Remove.
8708 (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
8709 * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
8710 (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
8711 * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
8712 * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
8713 * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
8714 * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
8716 * config/i386/i386-modes.def (BND32, BND64): Remove.
8717 * config/i386/i386.c (dbx_register_map): Remove bound registers.
8718 (dbx64_register_map): Ditto.
8719 (svr4_dbx_register_map): Ditto.
8720 (indirect_thunk_bnd_needed): Remove.
8721 (indirect_thunks_bnd_used): Ditto.
8722 (indirect_return_bnd_needed): Ditto.
8723 (indirect_return_via_cx_bnd): Ditto.
8724 (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
8725 (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
8726 (output_indirect_thunk): Ditto. Remove need_prefix argument.
8727 (output_indirect_thunk_function): Remove handling of
8728 indirect_return_bnd_needed, indirect_return_via_cx_bnd,
8729 indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
8730 (ix86_save_reg): Remove handling of crtl->return_bnd.
8731 (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
8732 (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
8733 and UNSPEC_BNDLX_ADDR.
8734 (ix86_output_indirect_branch_via_reg): Remove handling of
8735 indirect_thunk_prefix_bnd.
8736 (ix86_output_indirect_branch_via_push): Ditto.
8737 (ix86_output_function_return): Ditto.
8738 (ix86_output_indirect_function_return): Ditto.
8739 (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
8740 * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
8741 (CALL_USED_REGISTERS): Ditto.
8742 (REG_ALLOC_ORDER): Update for removal of bound registers.
8743 (HI_REGISTER_NAMES): Ditto.
8744 * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
8745 (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
8746 (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
8747 (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
8748 (FIRST_PSEUDO_REG): Update.
8749 (BND): Remove mode iterator.
8750 * config/i386/predicates.md (bnd_mem_operator): Remove.
8752 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
8754 * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
8757 2018-08-24 Richard Sandiford <richard.sandiford@arm.com>
8759 * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
8760 the case in which the permute needs only a single element and
8761 repeats for every vector of the result. Extend that case to
8762 handle variable-length vectors.
8763 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
8765 2018-08-24 H.J. Lu <hongjiu.lu@intel.com>
8768 * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
8769 on DW_MACINFO_start_file for -gsplit-dwarf -g3.
8771 2018-08-24 Richard Biener <rguenther@suse.de>
8773 * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
8774 bb_flags_allocated members.
8775 (auto_flag): New RAII class for allocating flags.
8776 (auto_edge_flag): New RAII class for allocating edge flags.
8777 (auto_bb_flag): New RAII class for allocating bb flags.
8778 * cfgloop.c (verify_loop_structure): Allocate temporary edge
8780 * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
8781 in favor of temporarily allocated BB flag.
8782 * hsa-brig.c: Re-order includes.
8783 * hsa-dump.c: Likewise.
8784 * hsa-regalloc.c: Likewise.
8785 * print-rtl.c: Likewise.
8786 * profile-count.c: Likewise.
8788 2018-08-24 Segher Boessenkool <segher@kernel.crashing.org>
8791 * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
8794 2018-08-24 Aldy Hernandez <aldyh@redhat.com>
8797 * wide-int-range.cc (wide_int_range_div): Do not ignore result
8798 from wide_int_range_multiplicative_op.
8800 2018-08-23 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
8802 * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
8805 2018-08-23 Giuliano Belinassi <giuliano.belinassi@usp.br>
8807 * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
8810 2018-08-23 Alexander Monakov <amonakov@ispras.ru>
8812 * tree-scalar-evolution.c (final_value_replacement_loop): Dump
8813 full GENERIC expression used for replacement.
8815 2018-08-23 Aldy Hernandez <aldyh@redhat.com>
8817 * tree-vrp.c (abs_extent_range): Remove.
8818 (extract_range_into_wide_ints): Pass wide ints by reference.
8819 (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
8820 Pass wide ints by reference in all calls to
8821 extract_range_into_wide_ints.
8822 * wide-int-range.cc (wide_int_range_div): New.
8823 * wide-int-range.h (wide_int_range_div): New.
8824 (wide_int_range_includes_zero_p): New.
8825 (wide_int_range_zero_p): New.
8827 2018-08-23 Matthew Malcomson <matthew.malcomson@arm.com>
8829 * config/aarch64/aarch64.md (arches): New enum.
8830 (arch): New enum attr.
8831 (arch_enabled): New attr.
8832 (enabled): Now uses arch_enabled only.
8833 (simd, sve, fp16): Removed attribute.
8834 (fp): Attr now defined in terms of 'arch'.
8835 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
8836 *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
8837 <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
8838 <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
8839 attributes into 'arch'.
8840 (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
8841 subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
8842 *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
8843 *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
8844 *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
8845 'simd' attribute into 'arch'.
8846 (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
8847 store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
8848 Convert use of 'fp' attribute to 'arch'.
8849 * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
8850 move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
8852 (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
8853 (different modes) Merge 'fp' and 'simd' into 'arch'.
8854 (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
8857 2018-08-23 Segher Boessenkool <segher@kernel.crashing.org>
8859 PR rtl-optimization/87026
8860 * expmed.c (canonicalize_comparison): If we can no longer create
8861 pseudoregisters, don't.
8863 2018-08-23 Richard Earnshaw <rearnsha@arm.com>
8866 * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
8868 * config/arm/arm.c (speculation_barrier_libfunc): New static
8870 (arm_init_libfuncs): Initialize it.
8871 (arm_emit_speculation_barrier): New function.
8872 * config/arm/arm.md (speculation_barrier): Call
8873 arm_emit_speculation_barrier for architectures that do not have
8875 (speculation_barrier_insn): Only match on Armv7 or later.
8877 2018-08-23 Richard Biener <rguenther@suse.de>
8880 * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
8883 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
8885 * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
8886 of single-vector TBLs.
8887 (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
8890 2018-08-23 Richard Sandiford <richard.sandiford@arm.com>
8893 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
8894 aarch64_evpc_tbl guard.
8896 2018-08-22 Bernd Edlinger <bernd.edlinger@hotmail.de>
8898 * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
8901 2018-08-22 Martin Sebor <msebor@redhat.com>
8904 * tree-pretty-print.c (pretty_print_string): Add argument.
8905 (dump_generic_node): Call to pretty_print_string with string size.
8907 2018-08-22 Segher Boessenkool <segher@kernel.crashing.org>
8909 PR rtl-optimization/86771
8910 * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
8911 of two SETs into those two SETs, one to be placed at i2, if that SETs
8912 destination is modified between i2 and i3.
8914 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
8916 PR tree-optimization/86725
8917 * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
8919 (vect_analyze_scalar_cycles_1): Check it.
8921 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
8923 PR tree-optimization/86725
8924 * tree-vect-loop.c (vect_is_simple_reduction): When treating
8925 an outer loop phi as a double reduction, make sure that the
8926 single user of the phi result is an inner loop phi.
8928 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
8930 * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
8931 grouped stores with gaps to a strided group.
8933 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
8935 * tree-vect-stmts.c (get_group_load_store_type)
8936 (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
8937 first statement in a group.
8939 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
8941 * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
8942 the sequence used in gcc/gcc.c.
8944 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
8947 * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
8950 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
8952 * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
8953 Darwin10-specific unwinder-shim.
8954 * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
8955 * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
8956 New to cater for Darwin10 Rosetta.
8958 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
8960 * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
8961 specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
8963 2018-08-22 Iain Sandoe <iain@sandoe.co.uk>
8968 * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
8969 (dwarf2out_switch_text_section): Generate a local label for the
8970 second function sub-section and apply it as the second FDE start
8972 * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
8973 second sub-section start.
8975 2018-08-22 Richard Biener <rguenther@suse.de>
8977 PR tree-optimization/86988
8978 * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
8980 2018-08-22 Richard Biener <rguenther@suse.de>
8982 PR tree-optimization/86945
8983 * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
8985 2018-08-22 Alexandre Oliva <oliva@adacore.com>
8987 * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
8988 a comment about how uses of r2 for .sdata2 come about.
8990 2018-08-22 Alexandre Oliva <aoliva@redhat.com>
8992 * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
8994 2018-08-21 Marek Polacek <polacek@redhat.com>
8996 PR c++/86981, Implement -Wpessimizing-move.
8997 * doc/invoke.texi: Document -Wpessimizing-move.
8999 2018-08-21 Jan Hubicka <jh@suse.cz>
9001 * tree.c (find_decls_types_r): Do not check for redundant typedefs.
9002 * tree.h (is_redundant_typedef): Remove.
9003 * dwarf2out.c (is_redundant_typedef): Turn into static function.
9005 2018-08-21 Jan Hubicka <jh@suse.cz>
9007 * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
9010 2018-08-21 Tamar Christina <tamar.christina@arm.com>
9012 * expmed.c (extract_low_bits): Reject invalid subregs early.
9014 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
9017 * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
9020 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
9022 * config/vxworks.h: Guard vxworks_asm_out_constructor and
9023 vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
9024 * config/vxworks.c: Likewise.
9026 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
9028 * config/vxworks.c: Set targetm.have_ctors_dtors
9029 if HAVE_INITFINI_ARRAY_SUPPORT.
9030 * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
9031 if HAVE_INITFINI_ARRAY_SUPPORT.
9033 2018-08-21 Rasmus Villemoes <rv@rasmusvillemoes.dk>
9035 * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
9036 default search path for VxWorks < 7.
9038 2018-08-21 Bernd Edlinger <bernd.edlinger@hotmail.de>
9040 * gimple-ssa-sprintf.c (decl_constant_value): Remove.
9041 (get_format_string): Refer to c_getstr.
9043 2018-08-21 Tom de Vries <tdevries@suse.de>
9045 * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
9046 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
9047 (debuginfo_early_stop): Declare.
9048 * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
9049 (debuginfo_start, debuginfo_stop, debuginfo_early_start)
9050 (debuginfo_early_stop): New function.
9051 (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
9052 and debuginfo_early_stop.
9053 * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
9054 * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
9055 (general_init): Call debuginfo_early_init.
9056 (finalize): Call debuginfo_fini.
9057 (do_compile): Call debuginfo_init.
9058 * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
9060 (@item -fdump-debug, @item -fdump-earlydebug): Add.
9062 2018-08-21 Tom de Vries <tdevries@suse.de>
9064 * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
9065 flag_dump_noaddr and flag_dump_unnumbered.
9067 2018-08-21 Aldy Hernandez <aldyh@redhat.com>
9069 * wide-int-range.cc (wide_int_range_abs): New.
9070 (wide_int_range_order_set): Rename from wide_int_range_min_max.
9071 * wide-int-range.h (wide_int_range_abs): New.
9072 (wide_int_range_min_max): New.
9073 * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
9074 case to call wide_int_range_abs.
9075 Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
9076 (extract_range_from_abs_expr): Delete.
9078 2018-08-20 Michael Meissner <meissner@linux.ibm.com>
9081 * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
9082 from 'Y' to 'YZ' to enable the LWAX instruction to be generated
9085 2018-08-20 Nathan Sidwell <nathan@acm.org>
9086 Jeff Law <law@redhat.com>
9088 * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
9089 * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
9091 2018-08-20 David Malcolm <dmalcolm@redhat.com>
9094 * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
9095 (decl_attributes): Likewise.
9096 * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
9098 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
9099 * diagnostic-core.h (class auto_diagnostic_group): New class.
9100 * diagnostic.c (diagnostic_initialize): Initialize the new fields.
9101 (diagnostic_report_diagnostic): Handle the first diagnostics within
9103 (emit_diagnostic): Add auto_diagnostic_group instance.
9105 (inform_n): Likewise.
9106 (warning): Likewise.
9107 (warning_at): Likewise.
9108 (warning_n): Likewise.
9109 (pedwarn): Likewise.
9110 (permerror): Likewise.
9112 (error_n): Likewise.
9113 (error_at): Likewise.
9115 (fatal_error): Likewise.
9116 (internal_error): Likewise.
9117 (internal_error_no_backtrace): Likewise.
9118 (auto_diagnostic_group::auto_diagnostic_group): New ctor.
9119 (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
9120 * diagnostic.h (struct diagnostic_context): Add fields
9121 "diagnostic_group_nesting_depth",
9122 "diagnostic_group_emission_count", "begin_group_cb",
9124 * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
9125 Add auto_diagnostic_group instance(s).
9126 (find_explicit_erroneous_behavior): Likewise.
9127 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
9128 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
9129 * gimplify.c (warn_implicit_fallthrough_r): Likewise.
9130 (gimplify_va_arg_expr): Likewise.
9131 * hsa-gen.c (HSA_SORRY_ATV): Likewise.
9132 (HSA_SORRY_AT): Likewise.
9133 * ipa-devirt.c (compare_virtual_tables): Likewise.
9134 (warn_odr): Likewise.
9135 * multiple_target.c (expand_target_clones): Likewise.
9136 * opts-common.c (cmdline_handle_error): Likewise.
9137 * reginfo.c (globalize_reg): Likewise.
9138 * substring-locations.c (format_warning_n_va): Likewise.
9139 * tree-inline.c (expand_call_inline): Likewise.
9140 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
9141 * tree-ssa-loop-niter.c
9142 (do_warn_aggressive_loop_optimizations): Likewise.
9143 * tree-ssa-uninit.c (warn_uninit): Likewise.
9144 * tree.c (warn_deprecated_use): Likewise.
9146 2018-08-20 H.J. Lu <hongjiu.lu@intel.com>
9149 * config/i386/i386.md (eh_return): Always update EH return
9150 address in word_mode.
9152 2018-08-20 Chung-Lin Tang <cltang@codesourcery.com>
9154 * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
9155 TARGET_SPLIT_COMPLEX_ARG is defined.
9157 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
9159 * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
9161 2018-08-20 Bernd Edlinger <bernd.edlinger@hotmail.de>
9164 * expr.c (expand_assignment): Assert that bitpos is positive.
9165 (store_field): Likewise
9166 (expand_expr_real_1): Make sure that bitpos is positive.
9167 * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
9170 2018-08-20 Nathan Sidwell <nathan@acm.org>
9172 * Makefile.in (CPP_ID_DATA_H): Delete.
9173 (CPP_INTERNAL_H): Don't add it.
9174 (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
9175 * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
9177 2018-08-20 Richard Biener <rguenther@suse.de>
9179 PR tree-optimization/78655
9180 * tree-vrp.c (extract_range_from_binary_expr_1): Make
9181 pointer + offset nonnull if either operand is nonnull work.
9183 2018-08-20 Tom de Vries <tdevries@suse.de>
9185 * dwarf2out.c (add_scalar_info): Don't add reference to existing die
9186 unless the referenced die describes the added property using
9187 DW_AT_location or DW_AT_const_value. Fall back to exprloc case.
9188 Otherwise, add a DW_AT_location to the referenced die.
9190 2018-08-19 Uros Bizjak <ubizjak@gmail.com>
9193 * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
9194 register_operand when calling ix86_set_reg_reg_cost.
9195 [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
9196 Set *total to 0 for operands that satisfy x86_64_immediate_operand
9197 predicate and to 1 otherwise.
9199 2018-08-18 Iain Sandoe <iain@sandoe.co.uk>
9201 * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
9202 emit a diagnostic that it is not supported and reset the option.
9203 * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
9204 supported and consume the option. (ASM_FINAL_SPEC): New.
9206 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
9208 * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
9211 2018-08-17 Sandra Loosemore <sandra@codesourcery.com>
9213 C-SKY port: Documentation
9215 * doc/extend.texi (C-SKY Function Attributes): New section.
9216 * doc/invoke.texi (Option Summary): Add C-SKY options.
9217 (C-SKY Options): New section.
9218 * doc/md.texi (Machine Constraints): Document C-SKY constraints.
9220 2018-08-17 Jojo <jijie_rong@c-sky.com>
9221 Huibin Wang <huibin_wang@c-sky.com>
9222 Sandra Loosemore <sandra@codesourcery.com>
9223 Chung-Lin Tang <cltang@codesourcery.com>
9225 C-SKY port: Backend implementation
9227 * config/csky/*: New.
9228 * common/config/csky/*: New.
9230 2018-08-17 Jojo <jijie_rong@c-sky.com>
9231 Huibin Wang <huibin_wang@c-sky.com>
9232 Sandra Loosemore <sandra@codesourcery.com>
9233 Chung-Lin Tang <cltang@codesourcery.com>
9234 Andrew Jenner <andrew@codesourcery.com>
9236 C-SKY port: Configury
9238 * config.gcc (csky-*-*): New.
9239 * configure.ac: Add csky to targets for dwarf2 debug_line support.
9240 * configure: Regenerated.
9242 2018-08-17 David Malcolm <dmalcolm@redhat.com>
9244 * dump-context.h: Include "dumpfile.h".
9245 (dump_context::dump_printf_va): Convert final param from va_list
9246 to va_list *. Convert from ATTRIBUTE_PRINTF to
9247 ATTRIBUTE_GCC_DUMP_PRINTF.
9248 (dump_context::dump_printf_loc_va): Likewise.
9249 * dumpfile.c: Include "stringpool.h".
9250 (make_item_for_dump_printf_va): Delete.
9251 (make_item_for_dump_printf): Delete.
9252 (class dump_pretty_printer): New class.
9253 (dump_pretty_printer::dump_pretty_printer): New ctor.
9254 (dump_pretty_printer::emit_items): New member function.
9255 (dump_pretty_printer::emit_any_pending_textual_chunks): New member
9257 (dump_pretty_printer::emit_item): New member function.
9258 (dump_pretty_printer::stash_item): New member function.
9259 (dump_pretty_printer::format_decoder_cb): New member function.
9260 (dump_pretty_printer::decode_format): New member function.
9261 (dump_context::dump_printf_va): Reimplement in terms of
9262 dump_pretty_printer.
9263 (dump_context::dump_printf_loc_va): Convert final param from va_list
9265 (dump_context::begin_scope): Reimplement call to
9266 make_item_for_dump_printf.
9267 (dump_printf): Update for change to dump_printf_va.
9268 (dump_printf_loc): Likewise.
9269 (selftest::test_capture_of_dump_calls): Convert "stmt" from
9270 greturn * to gimple *. Add a test_decl. Add tests of dump_printf
9271 with %T, %E, and %G.
9272 * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
9273 (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
9274 ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
9275 (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
9276 ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
9277 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
9278 use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
9279 within a dump_printf_loc call to "%wu".
9280 (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
9281 converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd". Add a
9282 missing space after "=".
9283 * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
9284 call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
9285 * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
9286 convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
9287 * tree-vectorizer.c (try_vectorize_loop_1): Likewise. Remove
9288 duplicate "vectorized" from message.
9290 2018-08-17 Szabolcs Nagy <szabolcs.nagy@arm.com>
9292 * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
9293 polyNxK_t element's TYPE_STRING_FLAG.
9295 2018-08-17 Segher Boessenkool <segher@kernel.crashing.org>
9297 * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
9298 (they were unnamed before). Fix comments.
9300 2018-08-17 Nathan Sidwell <nathan@acm.org>
9302 * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
9304 2018-08-17 Richard Biener <rguenther@suse.de>
9306 PR tree-optimization/86841
9307 * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
9309 2018-08-17 Martin Liska <mliska@suse.cz>
9311 * common.opt: Remove Warn, Init and Report for options with
9312 Ignore/Deprecated flag. Warning is done automatically for
9314 * config/i386/i386.opt: Likewise.
9315 * config/ia64/ia64.opt: Likewise.
9316 * config/rs6000/rs6000.opt: Likewise.
9317 * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
9318 Remove usage of flag_check_pointer_bounds.
9319 * lto-wrapper.c (merge_and_complain): Do not handle
9320 OPT_fcheck_pointer_bounds.
9321 (append_compiler_options): Likewise.
9322 * opt-functions.awk: Do not handle Deprecated.
9323 * optc-gen.awk: Check that Var, Report and Init are not
9324 used for an option with Ignore/Deprecated flag.
9325 * opts-common.c (decode_cmdline_option): Do not report
9327 (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
9329 * opts.h (struct cl_option): Remove cl_deprecated flag.
9330 (CL_ERR_DEPRECATED): Remove error enum value.
9332 2018-08-17 Richard Biener <rguenther@suse.de>
9335 * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
9336 across a va-arg-pack using call adjust its return value accordingly.
9338 2018-08-16 Martin Sebor <msebor@redhat.com>
9340 PR tree-optimization/86853
9341 * gimple-ssa-sprintf.c (struct format_result): Rename member.
9342 (struct fmtresult): Add member and initialize it in ctors.
9343 (format_character): Handle %C. Extend range to NUL. Set MAYFAIL.
9344 (format_string): Handle %S the same as %ls. Set MAYFAIL.
9345 (format_directive): Set POSUNDER4K when MAYFAIL is set.
9346 (parse_directive): Handle %C same as %c.
9347 (sprintf_dom_walker::compute_format_length): Adjust.
9348 (is_call_safe): Adjust.
9350 2018-08-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
9352 * builtins.c (c_strlen): Add new parameter eltsize. Use it
9353 for determining how to count the elements.
9354 * builtins.h (c_strlen): Adjust prototype.
9355 * expr.c (string_constant): Add new parameter mem_size.
9356 Set *mem_size appropriately.
9357 * expr.h (string_constant): Adjust protoype.
9358 * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
9359 * gimple-fold.h (get_range_strlen): Adjust prototype.
9360 * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
9361 (format_string): Call get_string_length with eltsize.
9363 2018-08-16 David Malcolm <dmalcolm@redhat.com>
9365 * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
9366 to emit the span, rather than setting it as the prefix.
9368 2018-08-16 David Malcolm <dmalcolm@redhat.com>
9370 * diagnostic-show-locus.c (layout::start_annotation_line): Add
9371 "margin_char" parameter, defaulting to space. Use it in place
9372 of pp_space for the initial part of the margin.
9373 (layout::print_leading_fixits): Use '+' when filling the margin
9374 of line-insertion fix-it hints.
9376 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
9378 * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
9381 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
9383 * config/rs6000/altivec.md: Don't set length attribute to the default
9385 * config/rs6000/darwin.md: Ditto.
9386 * config/rs6000/dfp.md: Ditto.
9387 * config/rs6000/htm.md: Ditto.
9388 * config/rs6000/rs6000.md: Ditto.
9389 * config/rs6000/sync.md: Ditto.
9390 * config/rs6000/vsx.md: Ditto.
9392 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
9394 * config/rs6000/altivec.md: Don't set length attribute to the default
9395 value, for branch instructions.
9396 * config/rs6000/darwin.md: Ditto.
9397 * config/rs6000/rs6000.md: Ditto.
9399 2018-08-16 Segher Boessenkool <segher@kernel.crashing.org>
9401 * config/rs6000/rs6000.md (length): Always define as const_int 4.
9402 (unnamed conditional branch define_insn): Set length to 4 or 8
9403 depending on offset.
9404 (<bd>_<mode>): Similar, for alternative 0.
9405 (<bd>tf_<mode>): Ditto.
9407 2018-08-16 Tamar Christina <tamar.christina@arm.com>
9409 * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
9411 2018-08-16 Matthew Malcomson <matthew.malcomson@arm.com>
9413 * doc/rtl.texi: Replace old RTX class names with new names.
9416 2018-08-16 Vlad Lazar <vlad.lazar@arm.com>
9418 * expmed.h (canonicalize_comparison): New declaration.
9419 * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
9420 * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
9421 * optabs.c (prepare_cmp_insn): Likewise.
9422 * rtl.h (unsigned_condition_p): New function which checks if a
9423 comparison operator is unsigned.
9425 2018-08-16 Nathan Sidwell <nathan@acm.org>
9427 * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
9428 * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
9430 2018-08-16 Tamar Christina <tamar.christina@arm.com>
9433 * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
9434 * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
9435 (mov<mov>): ..this and enable unconditionally.
9437 2018-08-16 Tamar Christina <tamar.christina@arm.com>
9439 * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
9441 2018-08-16 Sam Tebbs <sam.tebbs@arm.com>
9443 * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
9444 (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
9445 "Common" with "Target".
9447 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
9449 * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
9450 * doc/invoke.texi (mmitigate-rop): Remove.
9451 * config/i386/i386.c: Do not include "regrename.h".
9452 (ix86_rop_should_change_byte_p, reg_encoded_number)
9453 (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
9455 (ix86_reorg): Remove call to ix86_mitigate_rop.
9456 * config/i386/i386.md (attr "modrm_class"): Remove.
9457 (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
9458 (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
9459 (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
9461 2018-08-15 Will Schmidt <will_schmidt@vnet.ibm.com>
9463 * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
9464 allow folding of mergeh() and mergel() for the float and double types.
9465 (fold_mergehl_helper): Rework to handle building a permute tree
9468 2018-08-15 Uros Bizjak <ubizjak@gmail.com>
9470 * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
9473 2018-08-15 David Malcolm <dmalcolm@redhat.com>
9475 * common.opt (fdiagnostics-show-labels): New option.
9476 * diagnostic-show-locus.c (class layout_range): Add field
9478 (class layout): Add field "m_show_labels_p".
9479 (layout_range::layout_range): Add param "label" and use it to
9481 (make_range): Pass in NULL for new "label" param of layout_range's
9483 (layout::layout): Initialize m_show_labels_p.
9484 (layout::maybe_add_location_range): Pass in loc_range->m_label
9485 when constructing layout_range instances.
9486 (struct line_label): New struct.
9487 (layout::print_any_labels): New member function.
9488 (layout::print_line): Call it if label-printing is enabled.
9489 (selftest::test_one_liner_labels): New test.
9490 (selftest::test_diagnostic_show_locus_one_liner): Call it.
9491 * diagnostic.c (diagnostic_initialize): Initialize
9492 context->show_labels_p.
9493 * diagnostic.h (struct diagnostic_context): Add field
9495 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9496 -fno-diagnostics-show-labels.
9497 * dwarf2out.c (gen_producer_string): Add
9498 OPT_fdiagnostics_show_labels to the ignored options.
9499 * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
9501 (gcc_rich_location::maybe_add_expr): Likewise.
9502 * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
9503 label" param, defaulting to NULL.
9504 (gcc_rich_location::add_expr): Add "label" param.
9505 (gcc_rich_location::maybe_add_expr): Likewise.
9506 (class text_range_label): New class.
9507 (class range_label_for_type_mismatch): New class.
9508 * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
9509 of format_warning_va.
9510 (fmtwarn_n): Likewise for new params of format_warning_n_va.
9511 * lto-wrapper.c (merge_and_complain): Add
9512 OPT_fdiagnostics_show_labels to the "pick one setting" options.
9513 (append_compiler_options): Likewise to the dropped options.
9514 (append_diag_options): Likewise to the passed-on options.
9515 * opts.c (common_handle_option): Handle the new option.
9516 * selftest-diagnostic.c
9517 (test_diagnostic_context::test_diagnostic_context): Enable
9519 * substring-locations.c: Include "gcc-rich-location.h".
9520 (format_warning_n_va): Add "fmt_label" and "param_label" params
9521 and use them as appropriate.
9522 (format_warning_va): Add "fmt_label" and "param_label" params,
9523 passing them on to format_warning_n_va.
9524 (format_warning_at_substring): Likewise.
9525 (format_warning_at_substring_n): Likewise.
9526 * substring-locations.h (format_warning_va): Add "fmt_label" and
9527 "param_label" params.
9528 (format_warning_n_va): Likewise.
9529 (format_warning_at_substring): Likewise.
9530 (format_warning_at_substring_n): Likewise.
9531 * toplev.c (general_init): Initialize global_dc->show_labels_p.
9533 2018-08-15 Qing Zhao <qing.zhao@oracle.com>
9536 * builtins.c (expand_builtin_memcmp): Do not expand the call
9537 when overflow is detected.
9539 2018-08-15 Martin Sebor <msebor@redhat.com>
9541 PR tree-optimization/71625
9542 * config/aarch64/aarch64-builtins.c
9543 (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
9545 2018-08-15 Ilya Leoshkevich <iii@linux.ibm.com>
9547 * config/s390/s390.c (s390_reorg): Remove loop.
9549 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
9552 (darwin_function_switched_text_sections): Delete.
9553 * gcc/config/darwin.h
9554 (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
9556 2018-08-15 Iain Sandoe <iain@sandoe.co.uk>
9559 * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
9560 DEBUG_RNGLISTS_SECTION) new macros. (DEBUG_PUBNAMES_SECTION,
9561 DEBUG_PUBTYPES_SECTION) update to include GNU variant.
9563 2018-08-15 Martin Liska <mliska@suse.cz>
9565 PR tree-optimization/86925
9566 * predict.c (expr_expected_value_1): When taking
9567 later predictor, assign also probability.
9568 Use fold_build2_initializer_loc in order to fold
9569 the expression in -frounding-math.
9571 2018-08-14 Allan Sandfeld Jensen <allan.jensen@qt.io>
9573 * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
9575 (expand_vec_perm_1): Try the new method.
9577 2018-08-14 Ilya Leoshkevich <iii@linux.ibm.com>
9580 * lra-lives.c (remove_some_program_points_and_update_live_ranges):
9581 Check whether lra_live_max_point is 0 before dividing.
9583 2018-08-14 Martin Sebor <msebor@redhat.com>
9585 PR tree-optimization/86650
9586 * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
9587 (vrp_prop::check_mem_ref): Same.
9589 2018-08-13 Liu Hao <lh_mouse@126.com>
9591 * pretty-print.c (eat_esc_sequence): Swap the foreground and
9592 background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
9593 and clear it thereafter, as it only works for DBCS.
9595 2018-08-13 Liu Hao <lh_mouse@126.com>
9597 * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
9598 handle returned by _get_osf_handle().
9600 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
9602 * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
9603 for folding vec_perm.
9605 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
9607 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
9608 Add support for gimple-folding of vec_pack() and vec_unpack()
9611 2018-08-13 Will Schmidt <will_schmidt@vnet.ibm.com>
9613 * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
9614 vec_xst variants to the list.
9615 (rs6000_gimple_fold_builtin): Add support for folding unaligned
9616 vector loads and stores.
9618 2018-08-13 David Edelsohn <dje.gcc@gmail.com>
9620 * config.gcc (rs6000-ibm-aix4.x): Delete.
9621 (rs6000-ibm-aix5.1): Delete.
9622 (rs6000-ibm-aix5.2): Delete.
9623 (rs6000-ibm-aix5.3): Delete.
9624 * config/rs6000/aix43.h: Delete.
9625 * config/rs6000/aix51.h: Delete.
9626 * config/rs6000/aix52.h: Delete.
9627 * config/rs6000/t-aix43: Delete.
9629 2018-08-13 Ilya Leoshkevich <iii@linux.ibm.com>
9631 * config/s390/s390.c (s390_decompose_constant_pool_ref):
9633 (s390_decompose_address): Factor out constant pool ref
9636 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
9638 * config/nds32/nds32-predicates.c
9639 (nds32_can_use_bclr_p): Change return type as bool.
9640 (nds32_can_use_bset_p): Ditto.
9641 (nds32_can_use_btgl_p): Ditto.
9642 (nds32_can_use_bitci_p): Ditto.
9643 * config/nds32/nds32-protos.h
9644 (nds32_can_use_bclr_p): Change declaration.
9645 (nds32_can_use_bset_p): Ditto.
9646 (nds32_can_use_btgl_p): Ditto.
9647 (nds32_can_use_bitci_p): Ditto.
9649 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
9651 * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
9652 Support -msched-prolog-epilog option.
9653 * config/nds32/nds32.opt (msched-prolog-epilog): New option.
9655 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
9657 * common/config/nds32/nds32-common.c
9658 (nds32_option_optimization_table): Enalbe -malways-align.
9660 2018-08-12 Chung-Ju Wu <jasonwucj@gmail.com>
9662 * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
9664 * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
9665 OPT_misr_secure_ case.
9666 * config/nds32/nds32-isr.c: Implementation of backward compatibility.
9667 * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
9668 * config/nds32/nds32.c (nds32_attribute_table): Add critical and
9670 * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
9671 (nds32_isr_info): New field security_level.
9672 (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
9673 * config/nds32/nds32.md (return_internal): Consider critical attribute.
9674 * config/nds32/nds32.opt (misr-secure): New option.
9675 * config/nds32/nds32_init.inc: New file.
9676 * config/nds32/nds32_isr.h: New file.
9678 2018-08-11 John David Anglin <danglin@gcc.gnu.org>
9680 * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
9681 Update comment for atomic instructions.
9682 (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
9683 atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
9685 (atomic_loaddi): Revise fence expansion to only emit fence prior to
9686 load for __ATOMIC_SEQ_CST model.
9687 (atomic_loaddi_1): Remove float register target.
9688 (atomic_storedi): Handle CONST_INT values.
9689 (atomic_storedi_1): Remove float register source. Add special case
9691 (memory_barrier): New expander and insn.
9693 2018-08-11 Jakub Jelinek <jakub@redhat.com>
9695 PR tree-optimization/86835
9696 * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
9697 new_stmt after def_gsi, make sure to insert new_square_stmt after
9698 that stmt, not 2 stmts before it.
9700 2018-08-10 Alexander Monakov <amonakov@ispras.ru>
9703 * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
9706 2018-08-10 Martin Liska <mliska@suse.cz>
9709 * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
9711 * builtins.c (expand_builtin_expect_with_probability):
9713 (expand_builtin_expect_with_probability): New function.
9714 (build_builtin_expect_predicate): Add new argumnet probability
9715 for BUILT_IN_EXPECT_WITH_PROBABILITY.
9716 (fold_builtin_expect):
9719 * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
9720 * builtins.h (fold_builtin_expect): Set new argument.
9721 * doc/extend.texi: Document __builtin_expect_with_probability.
9722 * doc/invoke.texi: Likewise.
9723 * gimple-fold.c (gimple_fold_call): Pass new argument.
9724 * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
9725 also BUILT_IN_EXPECT_WITH_PROBABILITY.
9726 * predict.c (get_predictor_value): New function.
9727 (expr_expected_value): Add new argument probability. Assume
9728 that predictor and probability are always non-null.
9729 (expr_expected_value_1): Likewise. For __builtin_expect and
9730 __builtin_expect_with_probability set probability. Handle
9731 combination in binary expressions.
9732 (tree_predict_by_opcode): Simplify code by simply calling
9733 get_predictor_value.
9734 (pass_strip_predict_hints::execute): Add handling of
9735 BUILT_IN_EXPECT_WITH_PROBABILITY.
9736 * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
9738 * tree.h (DECL_BUILT_IN_P): New function.
9740 2018-08-10 Martin Liska <mliska@suse.cz>
9742 PR tree-optimization/85799
9743 * passes.def: Add argument for pass_strip_predict_hints.
9744 * predict.c (class pass_strip_predict_hints): Add new argument
9746 (strip_predictor_early): New function.
9747 (pass_strip_predict_hints::execute): Call the function to
9749 (strip_predict_hints): New function.
9750 * predict.def: Fix comment.
9752 2018-08-10 Thomas Preud'homme <thomas.preudhomme@linaro.org>
9754 * Makefile.in: Clarify which tm.texi to copy over to assert the
9755 right to grant a GFDL license for all.
9757 2018-08-09 Jeff Law <law@redhat.com>
9759 * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
9762 2018-08-09 Andreas Schwab <schwab@linux-m68k.org>
9764 * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
9767 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
9769 * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
9770 reductions for variable-length vectors.
9772 2018-08-09 David Malcolm <dmalcolm@redhat.com>
9775 * common.opt (fdiagnostics-show-line-numbers): New option.
9776 * diagnostic-show-locus.c (class layout): Add fields
9777 "m_show_line_numbers_p" and "m_linenum_width";
9778 (num_digits): New function.
9779 (test_num_digits): New function.
9780 (layout::layout): Initialize new fields. Update m_x_offset
9781 logic to handle any left margin.
9782 (layout::print_source_line): Print line number when requested.
9783 (layout::start_annotation_line): New member function.
9784 (layout::print_annotation_line): Call it.
9785 (layout::print_leading_fixits): Likewise.
9786 (layout::print_trailing_fixits): Likewise. Update calls to
9787 move_to_column for new parameter.
9788 (layout::get_x_bound_for_row): Add "add_left_margin" param and use
9789 it to potentially call start_annotation_line.
9790 (layout::show_ruler): Call start_annotation_line.
9791 (selftest::test_line_numbers_multiline_range): New selftest.
9792 (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
9793 and selftest::test_line_numbers_multiline_range.
9794 * diagnostic.c (diagnostic_initialize): Initialize
9795 show_line_numbers_p.
9796 * diagnostic.h (struct diagnostic_context): Add field
9797 "show_line_numbers_p".
9798 * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9799 -fno-diagnostics-show-line-numbers.
9800 * dwarf2out.c (gen_producer_string): Add
9801 OPT_fdiagnostics_show_line_numbers to the ignored options.
9802 * lto-wrapper.c (merge_and_complain): Likewise to the "pick
9803 one setting" options.
9804 (append_compiler_options): Likewise to the dropped options.
9805 (append_diag_options): Likewise to the passed-on options.
9806 * opts.c (common_handle_option): Handle the new option.
9807 * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
9809 2018-08-09 Kelvin Nilsen <kelvin@gcc.gnu.org>
9811 * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
9812 ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub. Add
9813 third argument of type "const signed char" to descriptions of
9814 __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
9815 __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
9816 __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
9817 __builtin_bcdsub_ov functions.
9819 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
9821 PR tree-optimization/86858
9822 * tree-vect-loop.c (vect_is_simple_reduction): Restore
9823 flow_bb_inside_loop_p calls.
9825 2018-08-09 Richard Sandiford <richard.sandiford@arm.com>
9827 PR tree-optimization/86871
9828 * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
9829 instead of gimple_assign_lhs.
9831 2018-08-09 Richard Earnshaw <rearnsha@arm.com>
9834 * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
9835 register constraint to operand 0.
9836 (add<mode>3_carryinC): Likewise.
9837 (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
9839 2018-08-09 Martin Liska <mliska@suse.cz>
9842 * common.opt: Remove extra line.
9844 2018-08-09 Martin Liska <mliska@suse.cz>
9846 * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
9847 at the end of a line, make first letter capital and end up
9848 a sentence with a dot.
9849 (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
9850 (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
9851 (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
9852 (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
9853 (PARAM_MAX_ISL_OPERATIONS): Likewise.
9854 (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
9855 (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
9856 (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
9857 (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
9858 (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
9859 (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
9860 (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
9861 (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
9862 (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
9863 (PARAM_TREE_REASSOC_WIDTH): Likewise.
9864 (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
9865 (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
9866 (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
9868 2018-08-09 Andreas Krebbel <krebbel@linux.ibm.com>
9871 * config/s390/s390.c (s390_option_override_internal): Reduce the
9872 stack-clash-protection-probe-interval param if it would be too big
9875 2018-08-08 Andreas Schwab <schwab@linux-m68k.org>
9878 * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
9879 * config/m68k/m68k.c (handle_move_double): Don't call
9880 m68k_final_prescan_insn.
9881 (m68k_adjust_decorated_operand): Renamed from
9882 m68k_final_prescan_insn, remove first and third operand and
9884 (print_operand): Call it.
9885 (print_operand_address): Call it.
9887 2018-08-08 Nathan Sidwell <nathan@acm.org>
9889 * diagnostic.c (diagnostic_report_current_module): Use
9890 linemap_included_from & linemap_included_from_linemap.
9892 2018-08-08 Hongbo Zhang <hongbo.zhang@linaro.org>
9894 * config/aarch64/aarch64-cores.def: Add phecda core.
9895 * config/aarch64/aarch64-tune.md: Regenerate.
9896 * doc/invoke.texi: Add phecda core.
9898 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
9901 * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
9903 * config/s390/s390.md ("movti"): Add more alternatives for
9904 constant to GPR copies.
9906 2018-08-08 Andreas Krebbel <krebbel@linux.ibm.com>
9908 * config/s390/s390.c: Fix whitespace damage throughout the file.
9909 * config/s390/s390.h: Likewise.
9910 * config/s390/tpf.h: Likewise.
9912 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
9914 * config/s390/s390.c (s390_loadrelative_operand_p):
9915 Remove TARGET_CPU_ZARCH usages.
9916 (s390_rtx_costs): Likewise.
9917 (s390_legitimate_constant_p): Likewise.
9918 (s390_cannot_force_const_mem): Likewise.
9919 (legitimate_reload_constant_p): Likewise.
9920 (s390_preferred_reload_class): Likewise.
9921 (legitimize_pic_address): Likewise.
9922 (legitimize_tls_address): Likewise.
9923 (s390_split_branches): Removed.
9924 (s390_add_execute): Removed.
9925 (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
9926 (s390_mainpool_start): Likewise.
9927 (s390_mainpool_finish): Likewise.
9928 (s390_mainpool_cancel): Removed.
9929 (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
9930 (s390_chunkify_cancel): Likewise.
9931 (s390_return_addr_rtx): Likewise.
9932 (s390_register_info): Remove split_branches_pending_p uages.
9933 (s390_optimize_register_info): Likewise.
9934 (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
9935 split_branches_pending_p usages.
9936 (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
9937 (s390_load_got): Likewise.
9938 (s390_expand_split_stack_prologue): Likewise.
9939 (output_asm_nops): Likewise.
9940 (s390_function_profiler): Likewise.
9941 (s390_emit_call): Likewise.
9942 (s390_conditional_register_usage): Likewise.
9943 (s390_optimize_prologue): Likewise.
9944 (s390_reorg): Remove TARGET_CPU_ZARCH and
9945 split_branches_pending_p usages.
9946 (s390_option_override_internal): Remove TARGET_CPU_ZARCH
9948 (s390_output_indirect_thunk_function): Likewise.
9949 * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
9950 (TARGET_CPU_ZARCH_P): Removed.
9951 (struct machine_function): Remove split_branches_pending_p.
9952 * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
9954 2018-08-08 Ilya Leoshkevich <iii@linux.ibm.com>
9956 * common/config/s390/s390-common.c (processor_flags_table):
9958 * config.gcc: Remove with_arch/with_tune support.
9959 * config/s390/2064.md: Remove cpu attribute comparisons.
9960 * config/s390/driver-native.c (s390_host_detect_local_cpu):
9962 * config/s390/linux.h (ASM_SPEC):
9963 Remove -march support.
9964 * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
9965 Use a table to get an arch level.
9966 * config/s390/s390-opts.h (enum processor_type):
9968 * config/s390/s390.c
9969 (processor_table): Remove entries, add arch_level values.
9970 (s390_issue_rate): Remove cases.
9971 (s390_option_override): Adjust
9972 s390_option_override_internal() call.
9973 (s390_option_override_internal): Remove deprecation warning.
9974 (s390_valid_target_attribute_tree): Adjust
9975 s390_option_override_internal() call.
9976 * config/s390/s390.h (struct s390_processor):
9977 Share with s390-c.c, add arch_level field.
9978 * config/s390/s390.md:
9979 Remove occurrences in cpu attribute.
9980 * config/s390/s390.opt: Remove -march/-mtune support.
9981 * config/s390/tpf.h (ASM_SPEC): Remove -march support.
9982 * doc/invoke.texi: Remove deprecation warning.
9984 2018-08-08 Luis Machado <luis.machado@linaro.org>
9986 * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
9988 (qdf24xx_tunings): Set vector cost structure to
9989 qdf24xx_vector_cost.
9991 * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
9992 <register_sextend>: Set to 3.
9994 2018-08-07 Richard Sandiford <richard.sandiford@arm.com>
9997 * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
9998 * config/aarch64/aarch64-simd.md
9999 (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
10000 (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
10001 (aarch64_frecpx<mode>): ...this new pattern.
10002 * config/aarch64/aarch64-simd-builtins.def: Remove comment
10003 about aarch64_frecp<FRECP:frecp_suffix><mode>.
10005 2018-08-07 Martin Liska <mliska@suse.cz>
10007 PR middle-end/83023
10008 * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
10009 BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
10010 * predict.def (PRED_MALLOC_NONNULL): New predictor.
10011 * doc/extend.texi: Document that malloc attribute adds
10014 2018-08-06 John David Anglin <danglin@gcc.gnu.org>
10017 * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10018 Define to speculation_safe_value_not_needed.
10020 2018-08-06 Jeff Law <law@redhat.com>
10022 * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
10023 the vr_values instance to cprop_into_stmt.
10024 (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
10025 (cprop_operand): Also query EVRP to determine if OP is a constant.
10027 2018-08-06 Nathan Sidwell <nathan@acm.org>
10029 * diagnostic.c (diagnostic_report_current_module): Reroll
10030 included-at loop. Translate text.
10032 2018-08-06 David Malcolm <dmalcolm@redhat.com>
10034 * function-tests.c (selftest::test_expansion_to_rtl): Call
10035 free_after_compilation.
10037 2018-08-06 Alan Hayward <alan.hayward@arm.com>
10039 * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
10041 2018-08-06 Andreas Krebbel <krebbel@linux.ibm.com>
10043 * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
10044 loops with memory block operations from getting unrolled.
10046 2018-08-06 Ulrich Weigand <uweigand@de.ibm.com>
10049 * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10050 Define to speculation_safe_value_not_needed.
10052 2018-08-06 Jeff Law <law@redhat.com>
10054 * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
10057 2018-08-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10060 * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
10061 with all enabled __intN types.
10063 * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
10065 2018-08-06 Alan Hayward <alan.hayward@arm.com>
10067 * alias.c (record_set): Check for clobber high.
10068 * cfgexpand.c (expand_gimple_stmt): Likewise.
10069 * combine-stack-adj.c (single_set_for_csa): Likewise.
10070 * combine.c (find_single_use_1): Likewise.
10071 (set_nonzero_bits_and_sign_copies): Likewise.
10072 (get_combine_src_dest): Likewise.
10073 (is_parallel_of_n_reg_sets): Likewise.
10074 (try_combine): Likewise.
10075 (record_dead_and_set_regs_1): Likewise.
10076 (reg_dead_at_p_1): Likewise.
10077 (reg_dead_at_p): Likewise.
10078 * dce.c (deletable_insn_p): Likewise.
10079 (mark_nonreg_stores_1): Likewise.
10080 (mark_nonreg_stores_2): Likewise.
10081 * df-scan.c (df_find_hard_reg_defs): Likewise.
10082 (df_uses_record): Likewise.
10083 (df_get_call_refs): Likewise.
10084 * dwarf2out.c (mem_loc_descriptor): Likewise.
10085 * haifa-sched.c (haifa_classify_rtx): Likewise.
10086 * ira-build.c (create_insn_allocnos): Likewise.
10087 * ira-costs.c (scan_one_insn): Likewise.
10088 * ira.c (equiv_init_movable_p): Likewise.
10089 (rtx_moveable_p): Likewise.
10090 (interesting_dest_for_shprep): Likewise.
10091 * jump.c (mark_jump_label_1): Likewise.
10092 * postreload-gcse.c (record_opr_changes): Likewise.
10093 * postreload.c (reload_cse_simplify): Likewise.
10094 (struct reg_use): Add source expr.
10095 (reload_combine): Check for clobber high.
10096 (reload_combine_note_use): Likewise.
10097 (reload_cse_move2add): Likewise.
10098 (move2add_note_store): Likewise.
10099 * print-rtl.c (print_pattern): Likewise.
10100 * recog.c (decode_asm_operands): Likewise.
10101 (store_data_bypass_p): Likewise.
10102 (if_test_bypass_p): Likewise.
10103 * regcprop.c (kill_clobbered_value): Likewise.
10104 (kill_set_value): Likewise.
10105 * reginfo.c (reg_scan_mark_refs): Likewise.
10106 * reload1.c (maybe_fix_stack_asms): Likewise.
10107 (eliminate_regs_1): Likewise.
10108 (elimination_effects): Likewise.
10109 (mark_not_eliminable): Likewise.
10110 (scan_paradoxical_subregs): Likewise.
10111 (forget_old_reloads_1): Likewise.
10112 * reorg.c (find_end_label): Likewise.
10113 (try_merge_delay_insns): Likewise.
10114 (redundant_insn): Likewise.
10115 (own_thread_p): Likewise.
10116 (fill_simple_delay_slots): Likewise.
10117 (fill_slots_from_thread): Likewise.
10118 (dbr_schedule): Likewise.
10119 * resource.c (update_live_status): Likewise.
10120 (mark_referenced_resources): Likewise.
10121 (mark_set_resources): Likewise.
10122 * rtl.c (copy_rtx): Likewise.
10123 * rtlanal.c (reg_referenced_p): Likewise.
10124 (single_set_2): Likewise.
10125 (noop_move_p): Likewise.
10126 (note_stores): Likewise.
10127 * sched-deps.c (sched_analyze_reg): Likewise.
10128 (sched_analyze_insn): Likewise.
10130 2018-08-06 Alan Hayward <alan.hayward@arm.com>
10132 * cse.c (invalidate_reg): New function extracted from...
10133 (invalidate): ...here.
10134 (canonicalize_insn): Check for clobber high.
10135 (invalidate_from_clobbers): invalidate clobber highs.
10136 (invalidate_from_sets_and_clobbers): Likewise.
10137 (count_reg_usage): Check for clobber high.
10138 (insn_live_p): Likewise.
10139 * cselib.c (cselib_expand_value_rtx_1):Likewise.
10140 (cselib_invalidate_regno): Check for clobber in setter.
10141 (cselib_invalidate_rtx): Pass through setter.
10142 (cselib_invalidate_rtx_note_stores):
10143 (cselib_process_insn): Check for clobber high.
10144 * cselib.h (cselib_invalidate_rtx): Add operand.
10146 2018-08-06 Alan Hayward <alan.hayward@arm.com>
10148 * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
10149 (mark_not_eliminable): Likewise.
10150 * lra-int.h (struct lra_insn_reg): Add clobber high marker.
10151 * lra-lives.c (process_bb_lives): Check for clobber high.
10152 * lra.c (new_insn_reg): Remember clobber highs.
10153 (collect_non_operand_hard_regs): Check for clobber high.
10154 (lra_set_insn_recog_data): Likewise.
10155 (add_regs_to_insn_regno_info): Likewise.
10156 (lra_update_insn_regno_info): Likewise.
10158 2018-08-06 Alan Hayward <alan.hayward@arm.com>
10160 * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
10161 * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
10163 2018-08-06 Alan Hayward <alan.hayward@arm.com>
10165 * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
10166 (copy_insn_1): Likewise.
10167 (gen_hard_reg_clobber_high): New gen function.
10168 * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
10169 * genemit.c (gen_exp): Likewise.
10170 (gen_emit_seq): Pass through info.
10171 (gen_insn): Check for CLOBBER_HIGH.
10172 (gen_expand): Pass through info.
10173 (gen_split): Likewise.
10174 (output_add_clobbers): Likewise.
10175 * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
10176 (remove_clobbers): Likewise.
10177 * rtl.h (gen_hard_reg_clobber_high): New declaration.
10179 2018-08-06 Alan Hayward <alan.hayward@arm.com>
10181 * doc/rtl.texi (clobber_high): Add.
10182 (parallel): Add in clobber high
10183 * rtl.c (rtl_check_failed_code3): Add function.
10184 * rtl.def (CLOBBER_HIGH): Add expression.
10185 * rtl.h (RTL_CHECKC3): Add macro.
10186 (rtl_check_failed_code3): Add declaration.
10187 (XC3EXP): Add macro.
10189 2018-08-05 H.J. Lu <hongjiu.lu@intel.com>
10192 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
10193 cfun->machine->max_used_stack_alignment if needed.
10195 2018-08-04 Martin Sebor <msebor@redhat.com>
10197 PR tree-optimization/86571
10198 * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
10201 2018-08-03 Sandra Loosemore <sandra@codesourcery.com>
10204 * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10207 2018-08-03 Jeff Law <law@redhat.com>
10210 * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10211 Define to speculation_safe_value_not_needed.
10213 2018-08-03 David Malcolm <dmalcolm@redhat.com>
10215 * doc/gcov.texi (-x): Remove duplicate "to".
10216 * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
10217 (-Wif-not-aligned): Remove duplicate "is".
10218 (-flto): Remove duplicate "the".
10219 (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
10220 duplicate "v5.00.b".
10221 (MSP430 Options): Remove duplicate "and" from the description
10222 of "-mgprel-sec=regexp".
10223 (x86 Options): Remove duplicate copies of "vmldLog102" and
10224 vmlsLog104 from description of "-mveclibabi=type".
10226 2018-08-03 Richard Sandiford <richard.sandiford@arm.com>
10228 * internal-fn.h (first_commutative_argument): Declare.
10229 * internal-fn.c (first_commutative_argument): New function.
10230 * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
10231 restrictions for pattern statements. Use first_commutative_argument
10232 to look for commutative operands in calls to internal functions.
10234 2018-08-03 Aldy Hernandez <aldyh@redhat.com>
10236 * Makefile.in (wide-int-range.o): New.
10237 * tree-vrp.c: Move all the wide_int_* functions to...
10238 * wide-int-range.cc: ...here.
10239 * tree-vrp.h: Move all the wide_int_* prototypes to...
10240 * wide-int-range.h: ...here.
10242 2018-08-03 Tom de Vries <tdevries@suse.de>
10244 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
10246 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
10247 * except.c (output_function_exception_table): Do early exit if
10248 targetm_common.except_unwind_info (&global_options) == UI_NONE.
10250 2018-08-03 Martin Liska <mliska@suse.cz>
10252 * predict.c (dump_prediction): Change to 2 digits
10255 2018-08-03 Siddhesh Poyarekar <siddhesh@sourceware.org>
10257 * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
10259 (falkor_am_1_gtov_gtov): ... a new insn reservation.
10261 2018-07-19 Ilya Leoshkevich <iii@linux.ibm.com>
10263 * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
10264 * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
10265 * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
10266 * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
10267 * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
10269 2018-08-02 David Malcolm <dmalcolm@redhat.com>
10271 * diagnostic-show-locus.c (diagnostic_show_locus): Use
10272 pp_take_prefix when saving the existing prefix.
10273 * diagnostic.c (diagnostic_append_note): Likewise.
10274 * langhooks.c (lhd_print_error_function): Likewise.
10275 * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
10276 param's type. Free the existing prefix.
10277 (pp_take_prefix): New function.
10278 (pretty_printer::pretty_printer): Drop the prefix parameter.
10279 Rename the length parameter to match the comment.
10280 (pretty_printer::~pretty_printer): Free the prefix.
10281 * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
10283 (struct pretty_printer): Drop the "const" from "prefix" field's
10284 type and clarify memory management.
10285 (pp_set_prefix): Drop the "const" from the 2nd param.
10286 (pp_take_prefix): New decl.
10288 2018-08-02 Aaron Sawdey <acsawdey@linux.ibm.com>
10290 * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
10291 for word_mode_ok here instead of passing as argument.
10292 (expand_block_compare): Change select_block_compare_mode() call.
10293 (expand_strncmp_gpr_sequence): New function.
10294 (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
10296 2018-08-02 Jeff Law <law@redhat.com>
10299 * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10300 Define to speculation_safe_value_not_needed.
10303 * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10304 Define to speculation_safe_value_not_needed.
10306 2018-08-02 Tom de Vries <tdevries@suse.de>
10309 * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
10310 function. Return UI_TARGET unconditionally.
10311 (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
10312 * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
10314 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
10316 * genemit.c (print_overload_test): Fix typo.
10318 2018-08-02 Richard Biener <rguenther@suse.de>
10320 PR tree-optimization/86816
10321 * tree-ssa-tail-merge.c (tail_merge_valueize): New function
10322 which checks for value availability before querying it.
10323 (gvn_uses_equal): Use it.
10324 (same_succ_hash): Likewise.
10325 (gimple_equal_p): Likewise.
10327 2018-08-02 Nick Clifton <nickc@redhat.com>
10330 * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10331 Define to speculation_safe_value_not_needed.
10334 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10335 Define to speculation_safe_value_not_needed.
10338 * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10339 Define to speculation_safe_value_not_needed.
10342 * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10343 Define to speculation_safe_value_not_needed.
10346 * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10347 Define to speculation_safe_value_not_needed.
10350 * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10351 Define to speculation_safe_value_not_needed.
10354 * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10355 Define to speculation_safe_value_not_needed.
10358 * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10359 Define to speculation_safe_value_not_needed.
10362 * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10363 speculation_safe_value_not_needed.
10366 * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
10367 to speculation_safe_value_not_needed.
10369 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
10371 * doc/md.texi: Expand the documentation of instruction names
10372 to mention port-local uses. Document '@' in pattern names.
10373 * read-md.h (overloaded_instance, overloaded_name): New structs.
10374 (mapping): Declare.
10375 (md_reader::handle_overloaded_name): New member function.
10376 (md_reader::get_overloads): Likewise.
10377 (md_reader::m_first_overload): New member variable.
10378 (md_reader::m_next_overload_ptr): Likewise.
10379 (md_reader::m_overloads_htab): Likewise.
10380 * read-md.c (md_reader::md_reader): Initialize m_first_overload,
10381 m_next_overload_ptr and m_overloads_htab.
10382 * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
10383 (get_mode_token, get_code_token, get_int_token): New functions.
10384 (map_attr_string): Add an optional argument that passes back
10385 the associated iterator.
10386 (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
10387 (md_reader::handle_overloaded_name, add_overload_instance): New
10389 (apply_iterators): Handle '@' names. Report an error if '@'
10390 is used without iterators.
10391 (initialize_iterators): Initialize the new iterator_group fields.
10392 * genopinit.c (handle_overloaded_code_for)
10393 (handle_overloaded_gen): New functions.
10394 (main): Use them to print declarations of maybe_code_for_* and
10395 maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
10396 * genemit.c (print_overload_arguments, print_overload_test)
10397 (handle_overloaded_code_for, handle_overloaded_gen): New functions.
10398 (main): Use it to print definitions of maybe_code_for_* and
10399 maybe_gen_* functions.
10400 * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
10401 gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
10402 instead of explicit mode checks.
10403 (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
10404 (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
10405 (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
10406 (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
10407 (aarch64_expand_compare_and_swap): Likewise
10408 gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
10409 (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
10410 (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
10411 (aarch64_constant_pool_reload_icode): Delete.
10412 (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
10413 instead of aarch64_constant_pool_reload_icode. Use
10414 code_for_aarch64_reload_mov instead of explicit mode checks.
10415 (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
10416 (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
10417 get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
10418 (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
10419 (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
10420 get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
10421 (aarch64_atomic_load_op_code): Delete.
10422 (aarch64_emit_atomic_load_op): Likewise.
10423 (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
10424 aarch64_atomic_load_op_code. Use gen_aarch64_atomic_load
10425 instead of aarch64_emit_atomic_load_op.
10426 * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
10427 (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
10428 (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
10429 (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
10430 character before the pattern name.
10431 * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
10432 (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
10433 (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
10434 (aarch64_frecps<mode>): Likewise.
10435 * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
10436 (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
10437 (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
10438 (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
10439 (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
10441 2018-08-02 Richard Sandiford <richard.sandiford@arm.com>
10443 * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
10444 Allow HFmode constants if TARGET_FP_F16INST.
10446 2018-08-02 Jackson Woodruff <jackson.woodruff@arm.com>
10449 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
10450 No longer check last store for clobber of address register.
10452 2018-08-02 Martin Liska <mliska@suse.cz>
10454 PR gcov-profile/86817
10455 * gcov.c (process_all_functions): New function.
10457 (process_file): Move functions processing to
10458 process_all_functions.
10460 2018-08-02 David Malcolm <dmalcolm@redhat.com>
10462 * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
10463 "const" to the "gimple *" and "rtx_insn *" parameters.
10464 * dumpfile.h (dump_user_location_t::dump_user_location_t):
10466 (dump_location_t::dump_location_t): Likewise.
10468 2018-08-01 Martin Sebor <msebor@redhat.com>
10470 PR tree-optimization/86650
10471 * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
10472 rather than a "gcall *". Directly pass the data of interest
10473 to percent_K_format, rather than building a temporary CALL_EXPR
10475 * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
10476 (gimple_fold_builtin_strncat): Adjust.
10477 * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
10478 gcall* argument with gimple*.
10479 * gimple-ssa-warn-restrict.c (check_call): Same.
10480 (wrestrict_dom_walker::before_dom_children): Same.
10481 (builtin_access::builtin_access): Same.
10482 (check_bounds_or_overlap): Same
10483 (maybe_diag_overlap): Same.
10484 (maybe_diag_offset_bounds): Same.
10485 * tree-diagnostic.c (default_tree_printer): Move usage of
10486 EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
10488 * tree-pretty-print.c (percent_K_format): Add argument.
10489 * tree-pretty-print.h: Add argument.
10490 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
10491 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
10492 (maybe_diag_stxncpy_trunc): Same.
10493 (handle_builtin_stxncpy): Same.
10494 (handle_builtin_strcat): Same.
10496 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10498 * match.pd: Optimise pointer range checks.
10500 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10502 PR tree-optimization/86758
10503 * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
10504 to remove pattern statements.
10506 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10508 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
10509 result of dfs_enumerate_from when constructing stmt_vec_infos,
10510 instead of additionally calling get_loop_body.
10512 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10514 * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
10516 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
10517 When creating an iv, assert that the step is not known to be zero.
10518 (vect_setup_realignment): Update call accordingly.
10519 * tree-vect-stmts.c (vectorizable_store): Likewise.
10520 (vectorizable_load): Likewise. Handle VMAT_INVARIANT separately.
10522 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10524 * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
10525 * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
10526 (vectorizable_reduction): Likewise.
10527 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10528 (vect_detect_hybrid_slp_stmts): Likewise.
10529 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10531 2018-08-01 Aldy Hernandez <aldyh@redhat.com>
10533 * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
10534 (wide_int_set_zero_nonzero_bits): ...this.
10535 (zero_nonzero_bits_from_vr): Rename to...
10536 (vrp_set_zero_nonzero_bits): ...this.
10537 (extract_range_from_multiplicative_op_1): Abstract wide int
10539 (wide_int_range_multiplicative_op): ...here.
10540 (extract_range_from_binary_expr_1): Extract wide int binary
10541 operations into their own functions.
10542 (wide_int_range_lshift): New.
10543 (wide_int_range_can_optimize_bit_op): New.
10544 (wide_int_range_shift_undefined_p): New.
10545 (wide_int_range_bit_xor): New.
10546 (wide_int_range_bit_ior): New.
10547 (wide_int_range_bit_and): New.
10548 (wide_int_range_trunc_mod): New.
10549 (extract_range_into_wide_ints): New.
10550 (vrp_shift_undefined_p): New.
10551 (extract_range_from_multiplicative_op): New.
10552 (vrp_can_optimize_bit_op): New.
10553 * tree-vrp.h (value_range::dump): New.
10554 (wide_int_range_multiplicative_op): New.
10555 (wide_int_range_lshift):New.
10556 (wide_int_range_shift_undefined_p): New.
10557 (wide_int_range_bit_xor): New.
10558 (wide_int_range_bit_ior): New.
10559 (wide_int_range_bit_and): New.
10560 (wide_int_range_trunc_mod): New.
10561 (zero_nonzero_bits_from_bounds): Rename to...
10562 (wide_int_set_zero_nonzero_bits): ...this.
10563 (zero_nonzero_bits_from_vr): Rename to...
10564 (vrp_set_zero_nonzero_bits): ...this.
10565 (range_easy_mask_min_max): Rename to...
10566 (wide_int_range_can_optimize_bit_op): this.
10567 * vr-values.c (simplify_bit_ops_using_ranges): Rename
10568 zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
10570 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10572 * tree-vectorizer.h (vect_orig_stmt): New function.
10573 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
10574 * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
10575 (vect_create_epilog_for_reduction): Likewise.
10576 (vectorizable_live_operation): Likewise.
10577 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
10578 (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
10579 * tree-vect-stmts.c (vectorizable_call): Likewise.
10580 (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
10582 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10584 * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
10586 * tree-vect-stmts.c (vect_transform_stmt): Likewise.
10587 * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
10588 (vect_transform_loop): Likewise.
10589 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10591 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10593 * tree-vectorizer.h (vect_schedule_slp): Return void.
10594 * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10595 (vect_schedule_slp): Likewise.
10597 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10599 * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
10601 (vect_transform_loop): Update calls accordingly. Schedule SLP
10602 instances before the main loop, if any exist.
10604 2018-08-01 Richard Sandiford <richard.sandiford@arm.com>
10606 PR tree-optimization/86749
10607 * tree-vect-patterns.c (vect_determine_min_output_precision_1):
10608 If the lhs is used in a COND_EXPR, check that it is being used
10609 as the "then" or "else" value.
10611 2018-08-01 Tom de Vries <tdevries@suse.de>
10614 * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10615 speculation_safe_value_not_needed.
10617 2018-08-01 Richard Biener <rguenther@suse.de>
10619 * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
10620 as base and offset.
10622 2018-08-01 Martin Liska <mliska@suse.cz>
10624 * value-prof.c (gimple_divmod_fixed_value_transform): Unify
10625 format how successful transformation is dumped.
10626 (gimple_mod_pow2_value_transform): Likewise.
10627 (gimple_mod_subtract_transform): Likewise.
10628 (gimple_stringops_transform): Likewise.
10630 2018-08-01 Martin Liska <mliska@suse.cz>
10632 PR value-prof/35543
10633 * value-prof.c (interesting_stringop_to_profile_p):
10634 Simplify the code and add BUILT_IN_MEMMOVE.
10635 (gimple_stringops_transform): Likewise.
10637 2018-08-01 Sam Tebbs <sam.tebbs@arm.com>
10639 * config/aarch64/aarch64-simd.md
10640 (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
10641 (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
10642 use GPI iterator instead of SI mode.
10644 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
10646 * config/rs6000/rs6000.md (speculation_barrier): Renamed from
10647 rs6000_speculation_barrier.
10648 * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
10649 new barrier pattern name.
10651 2018-08-01 Richard Earnshaw <rearnsha@arm.com>
10653 * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
10654 (speculation_barrier): New insn.
10656 2018-08-01 Richard Biener <rguenther@suse.de>
10659 * graphite.h: Include isl/id.h and isl/space.h to allow build
10662 2018-08-01 Jan Willem Jagersma <jwjagersma@gmail.com>
10665 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
10666 mode for COFF targets.
10667 * defaults.h (TARGET_COFF): Define.
10668 * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
10669 TARGET_COFF): Define.
10670 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
10671 * config/i386/djgpp.c (saved_debug_info_level): New static variable.
10672 (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
10674 2018-07-31 Alexandre Oliva <oliva@adacore.com>
10676 * gimple-streamer-in.c (input_bb): Restore BB discriminator.
10677 * gimple-streamer-out.c (output_bb): Save it.
10678 * lto-streamer-in.c (input_struct_function_base): Restore
10679 instance discriminator if available. Create map on demand.
10680 * lto-streamer-out.c (output_struct_function_base): Save it if
10682 * final.c (decl_to_instance_map): Document LTO strategy.
10684 2018-07-31 Alexandre Oliva <oliva@adacore.com>
10685 Olivier Hainque <hainque@adacore.com>
10687 * debug.h (decl_to_instance_map_t): New type.
10688 (decl_to_instance_map): Declare.
10689 (maybe_create_decl_to_instance_map): New inline function.
10690 * final.c (bb_discriminator, last_bb_discriminator): New statics,
10691 to track basic block discriminators.
10692 (final_start_function_1): Initialize them.
10693 (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
10695 (decl_to_instance_map): New variable.
10696 (map_decl_to_instance, maybe_set_discriminator): New functions.
10697 (notice_source_line): Set discriminator.
10699 2018-07-31 Ian Lance Taylor <iant@golang.org>
10701 * targhooks.c (default_have_speculation_safe_value): Add
10704 2018-07-31 David Malcolm <dmalcolm@redhat.com>
10706 * dump-context.h: Include "pretty-print.h".
10707 (dump_context::refresh_dumps_are_enabled): New decl.
10708 (dump_context::emit_item): New decl.
10709 (class dump_context): Add fields "m_test_pp" and
10711 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10712 (temp_dump_context::get_dumped_text): New decl.
10713 (class temp_dump_context): Add field "m_pp".
10714 * dumpfile.c (refresh_dumps_are_enabled): Convert to...
10715 (dump_context::refresh_dumps_are_enabled): ...and add a test for
10717 (set_dump_file): Update for above change.
10718 (set_alt_dump_file): Likewise.
10719 (dump_loc): New overload, taking a pretty_printer *.
10720 (dump_context::dump_loc): Call end_any_optinfo. Dump the location
10721 to any test pretty-printer.
10722 (make_item_for_dump_gimple_stmt): New function, adapted from
10723 optinfo::add_gimple_stmt.
10724 (dump_context::dump_gimple_stmt): Call it, and use the result,
10725 eliminating the direct usage of dump_file and alt_dump_file in
10726 favor of indirectly using them via emit_item.
10727 (make_item_for_dump_gimple_expr): New function, adapted from
10728 optinfo::add_gimple_expr.
10729 (dump_context::dump_gimple_expr): Call it, and use the result,
10730 eliminating the direct usage of dump_file and alt_dump_file in
10731 favor of indirectly using them via emit_item.
10732 (make_item_for_dump_generic_expr): New function, adapted from
10734 (dump_context::dump_generic_expr): Call it, and use the result,
10735 eliminating the direct usage of dump_file and alt_dump_file in
10736 favor of indirectly using them via emit_item.
10737 (make_item_for_dump_printf_va): New function, adapted from
10738 optinfo::add_printf_va.
10739 (make_item_for_dump_printf): New function.
10740 (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
10741 and use the result, eliminating the direct usage of dump_file and
10742 alt_dump_file in favor of indirectly using them via emit_item.
10743 (make_item_for_dump_dec): New function.
10744 (dump_context::dump_dec): Call it, and use the result,
10745 eliminating the direct usage of dump_file and alt_dump_file in
10746 favor of indirectly using them via emit_item.
10747 (make_item_for_dump_symtab_node): New function, adapted from
10748 optinfo::add_symtab_node.
10749 (dump_context::dump_symtab_node): Call it, and use the result,
10750 eliminating the direct usage of dump_file and alt_dump_file in
10751 favor of indirectly using them via emit_item.
10752 (dump_context::begin_scope): Reimplement, avoiding direct usage
10753 of dump_file and alt_dump_file in favor of indirectly using them
10755 (dump_context::emit_item): New member function.
10756 (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10757 Set up test pretty-printer on the underlying context. Call
10758 refresh_dumps_are_enabled.
10759 (temp_dump_context::~temp_dump_context): Call
10760 refresh_dumps_are_enabled.
10761 (temp_dump_context::get_dumped_text): New member function.
10762 (selftest::verify_dumped_text): New function.
10763 (ASSERT_DUMPED_TEXT_EQ): New macro.
10764 (selftest::test_capture_of_dump_calls): Run all tests twice, with
10765 and then without optinfo enabled. Add uses of
10766 ASSERT_DUMPED_TEXT_EQ to all tests. Add test of nested scopes.
10767 * dumpfile.h: Update comment for the dump_* API.
10768 * optinfo-emit-json.cc
10769 (selftest::test_building_json_from_dump_calls): Update for new
10770 param for temp_dump_context ctor.
10771 * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
10772 and "m_owned" field.
10773 (optinfo_item::~optinfo_item): Likewise.
10774 (optinfo::add_item): New member function.
10775 (optinfo::emit): Update comment.
10776 (optinfo::add_string): Delete.
10777 (optinfo::add_printf): Delete.
10778 (optinfo::add_printf_va): Delete.
10779 (optinfo::add_gimple_stmt): Delete.
10780 (optinfo::add_gimple_expr): Delete.
10781 (optinfo::add_tree): Delete.
10782 (optinfo::add_symtab_node): Delete.
10783 (optinfo::add_dec): Delete.
10784 * optinfo.h (class dump_context): New forward decl.
10785 (optinfo::add_item): New decl.
10786 (optinfo::add_string): Delete.
10787 (optinfo::add_printf): Delete.
10788 (optinfo::add_printf_va): Delete.
10789 (optinfo::add_gimple_stmt): Delete.
10790 (optinfo::add_gimple_expr): Delete.
10791 (optinfo::add_tree): Delete.
10792 (optinfo::add_symtab_node): Delete.
10793 (optinfo::add_dec): Delete.
10794 (optinfo::add_poly_int): Delete.
10795 (optinfo_item::optinfo_item): Remove "owned" param.
10796 (class optinfo_item): Remove field "m_owned".
10798 2018-07-31 Jozef Lawrynowicz <jozef.l@mittosystems.com>
10800 PR middle-end/86705
10801 * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
10802 MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
10803 requested variable alignment.
10804 (expand_one_ssa_partition): Likewise.
10805 (expand_one_var): Likewise.
10807 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10809 * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
10810 to speculation_safe_value_not_needed.
10812 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10814 * targhooks.h (speculation_safe_value_not_needed): New prototype.
10815 * targhooks.c (speculation_safe_value_not_needed): New function.
10816 * target.def (have_speculation_safe_value): Update documentation.
10817 * doc/tm.texi: Regenerated.
10819 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10821 * config/aarch64/iterators.md (ALLI_TI): New iterator.
10822 * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
10824 (despeculate_copy<ALLI:mode>_insn): New insn.
10825 (despeculate_copyti_insn): New insn.
10826 (despeculate_simple<ALLI:mode>): New insn
10827 (despeculate_simpleti): New insn.
10828 * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
10830 (TARGET_SPECULATION_SAFE_VALUE): Redefine to
10831 aarch64_speculation_safe_value.
10832 (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
10834 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10836 * config/aarch64/aarch64-speculation.cc: New file.
10837 * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
10838 before pass_reorder_blocks.
10839 * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
10841 * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
10842 X14 and X15 when tracking speculation.
10843 * config/aarch64/aarch64.md (register name constants): Add
10844 SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
10845 (unspec): Add UNSPEC_SPECULATION_TRACKER.
10846 (speculation_barrier): New insn attribute.
10847 (cmp<mode>): Allow SP in comparisons.
10848 (speculation_tracker): New insn.
10849 (speculation_barrier): Add speculation_barrier attribute.
10850 * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
10851 * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
10852 * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
10854 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10856 * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
10857 aarch64_track_speculation is true.
10858 (tb<optab><mode>1): Likewise.
10859 * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
10860 generate CB[N]Z when tracking speculation.
10861 (aarch64_split_compare_and_swap): Likewise.
10862 (aarch64_split_atomic_op): Likewise.
10864 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10866 * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
10868 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10870 * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
10871 (speculation_barrier): New insn.
10873 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10875 * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
10876 * config/arm/arm.md (speculation_barrier): New expand.
10877 (speculation_barrier_insn): New pattern.
10879 2018-07-31 Richard Earnshaw <rearnsha@arm.com>
10881 * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
10882 (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
10883 (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
10884 * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
10886 * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
10887 (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
10888 (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
10889 (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
10890 (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
10891 (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
10892 (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
10893 * builtins.c (expand_speculation_safe_value): New function.
10894 (expand_builtin): Call it.
10895 * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
10896 * doc/extend.texi: Document __builtin_speculation_safe_value.
10897 * doc/md.texi: Document "speculation_barrier" pattern.
10898 * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
10899 TARGET_HAVE_SPECULATION_SAFE_VALUE.
10900 * doc/tm.texi: Regenerated.
10901 * target.def (have_speculation_safe_value, speculation_safe_value): New
10903 * targhooks.c (default_have_speculation_safe_value): New function.
10904 (default_speculation_safe_value): New function.
10905 * targhooks.h (default_have_speculation_safe_value): Add prototype.
10906 (default_speculation_safe_value): Add prototype.
10908 2018-07-31 David Malcolm <dmalcolm@redhat.com>
10910 * dump-context.h (dump_context::dump_loc): New decl.
10911 * dumpfile.c (dump_context::dump_loc): New member function.
10912 (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
10913 and dump_gimple_stmt.
10914 (dump_context::dump_gimple_expr_loc): Likewise, using
10916 (dump_context::dump_generic_expr_loc): Likewise, using
10918 (dump_context::dump_printf_loc_va): Likewise, using
10920 (dump_context::begin_scope): Explicitly using the global function
10921 "dump_loc", rather than the member function.
10923 2018-07-31 Martin Sebor <msebor@redhat.com>
10925 PR tree-optimization/86741
10926 * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
10928 2018-07-31 Andreas Krebbel <krebbel@linux.ibm.com>
10930 * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
10931 depend on whether prefetch instructions will be emitted or not.
10932 Use TARGET_SETMEM_PFD for checking whether prefetch instructions
10933 will be emitted or not.
10934 * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
10935 (TARGET_SETMEM_PFD): New macros.
10937 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10939 * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
10940 (NULL_STMT_VEC_INFO): Delete.
10941 (stmt_vec_info::operator*): Likewise.
10942 (stmt_vec_info::operator gimple *): Likewise.
10943 * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
10944 of NULL_STMT_VEC_INFO.
10945 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10946 (vect_reassociating_reduction_p): Likewise.
10947 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
10948 (vectorizable_store): Likewise.
10949 * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
10950 (vec_info::free_stmt_vec_infos): Likewise.
10952 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10954 * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
10955 * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
10957 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10959 * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
10960 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10961 (vec_info::free_stmt_vec_info): New private member functions.
10962 (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
10963 (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
10964 * tree-parloops.c (gather_scalar_reductions): Remove calls to
10965 set_stmt_vec_info_vec and free_stmt_vec_infos.
10966 * tree-vect-loop.c (_loop_vec_info): Remove call to
10967 set_stmt_vec_info_vec.
10968 * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
10969 (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
10970 * tree-vectorizer.c (vec_info::new_stmt_vec_info)
10971 (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10972 (vec_info::free_stmt_vec_info): ...these new functions. Remove
10973 assignments in {vec_info::,}new_stmt_vec_info that are redundant
10974 with the clearing in the xcalloc.
10975 (stmt_vec_info_vec): Delete.
10976 (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
10977 (vectorize_loops): Likewise.
10978 (vec_info::~vec_info): Remove argument from call to
10979 free_stmt_vec_infos.
10980 (vec_info::add_stmt): Remove vinfo argument from call to
10983 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
10985 * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
10986 rather than a gimple stmt.
10987 * tree-vect-stmts.c (free_stmt_vec_info): Likewise. Don't free
10988 information for pattern statements when passed the original
10989 statement; instead wait to be passed the pattern statement itself.
10990 Don't call set_vinfo_for_stmt here.
10991 (free_stmt_vec_infos): Update call to free_stmt_vec_info.
10992 * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
10993 stmt_vec_infos here.
10994 * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
10995 * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
10996 stmt_vec_infos entry.
10998 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11000 * tree-vectorizer.h (vec_info::replace_stmt): Declare.
11001 * tree-vectorizer.c (vec_info::replace_stmt): New function.
11002 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
11003 * tree-vect-stmts.c (vectorizable_call): Likewise.
11004 (vectorizable_simd_clone_call): Likewise.
11006 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11008 * tree-vectorizer.h (vec_info::remove_stmt): Declare.
11009 * tree-vectorizer.c (vec_info::remove_stmt): New function.
11010 * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
11011 * tree-vect-loop.c (vect_transform_loop): Likewise.
11012 * tree-vect-slp.c (vect_schedule_slp): Likewise.
11013 * tree-vect-stmts.c (vect_remove_stores): Likewise.
11015 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11017 * tree-vectorizer.h (vec_info::lookup_dr): New member function.
11018 (vect_dr_stmt): Delete.
11019 * tree-vectorizer.c (vec_info::lookup_dr): New function.
11020 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
11022 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
11023 (vect_analyze_data_ref_dependence, vect_record_base_alignments)
11024 (vect_verify_datarefs_alignment, vect_peeling_supportable)
11025 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
11026 (vect_analyze_data_refs): Likewise.
11027 (vect_slp_analyze_data_ref_dependence): Likewise. Take a vec_info
11029 (vect_find_same_alignment_drs): Likewise.
11030 (vect_slp_analyze_node_dependences): Update calls accordingly.
11031 (vect_analyze_data_refs_alignment): Likewise. Use vec_info::lookup_dr
11032 instead of DR_VECT_AUX.
11033 (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
11034 of a vector data references. Use vec_info::lookup_dr instead of
11036 (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
11037 (vect_enhance_data_refs_alignment): Likewise. Use vec_info::lookup_dr
11038 instead of DR_VECT_AUX.
11040 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11042 * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
11044 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
11046 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11047 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11048 (vect_gen_prolog_loop_niters): Likewise.
11050 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11052 * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
11053 (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
11054 (vect_known_alignment_in_bytes, vect_dr_behavior)
11055 (vect_get_scalar_dr_size): Take references as dr_vec_infos
11056 instead of data_references. Update calls to other routines for
11057 which the same change has been made.
11058 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
11059 dr_vec_infos instead of stmt_vec_infos.
11060 (vect_analyze_data_ref_dependence): Update call accordingly.
11061 (vect_slp_analyze_data_ref_dependence)
11062 (vect_record_base_alignments): Use DR_VECT_AUX.
11063 (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
11064 (vect_update_misalignment_for_peel, verify_data_ref_alignment)
11065 (vector_alignment_reachable_p, vect_get_data_access_cost)
11066 (vect_peeling_supportable, vect_analyze_group_access_1)
11067 (vect_analyze_group_access, vect_analyze_data_ref_access)
11068 (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
11069 (vect_compile_time_alias, vect_small_gap_p)
11070 (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
11071 (vect_supportable_dr_alignment): Take references as dr_vec_infos
11072 instead of data_references. Update calls to other routines for
11073 which the same change has been made.
11074 (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
11075 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
11076 (vect_slp_analyze_and_verify_node_alignment)
11077 (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
11078 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
11079 (vect_setup_realignment): Use dr_vec_infos. Update calls after
11081 (_vect_peel_info::dr): Replace with...
11082 (_vect_peel_info::dr_info): ...this new field.
11083 (vect_peeling_hash_get_most_frequent)
11084 (vect_peeling_hash_choose_best_peeling): Update accordingly.
11085 (vect_peeling_hash_get_lowest_cost):
11086 (vect_enhance_data_refs_alignment): Likewise. Update calls to other
11087 routines for which the same change has been made.
11088 (vect_peeling_hash_insert): Likewise. Take a dr_vec_info instead of a
11090 * tree-vect-loop-manip.c (get_misalign_in_elems)
11091 (vect_gen_prolog_loop_niters): Use dr_vec_infos. Update calls after
11093 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11094 * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
11095 (vect_truncate_gather_scatter_offset, compare_step_with_zero)
11096 (get_group_load_store_type, get_negative_load_store_type)
11097 (vect_get_data_ptr_increment, vectorizable_store)
11098 (vectorizable_load): Likewise.
11099 (ensure_base_align): Take a dr_vec_info instead of a data_reference.
11100 Update calls to other routines for which the same change has been made.
11102 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11104 * tree-vectorizer.h (vec_info::move_dr): New member function.
11105 (dataref_aux): Rename to...
11106 (dr_vec_info): ...this and add "dr" and "stmt" fields.
11107 (_stmt_vec_info::dr_aux): Update accordingly.
11108 (_stmt_vec_info::data_ref_info): Delete.
11109 (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
11110 (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
11111 (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
11112 (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
11114 (STMT_VINFO_DATA_REF): Likewise. Turn into an lvalue.
11115 (STMT_VINFO_DR_INFO): New macro.
11116 (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
11117 (set_dr_misalignment): Update after rename of dataref_aux.
11118 (vect_dr_stmt): Move earlier in file. Return dr_aux.stmt.
11119 * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
11120 initialization of STMT_VINFO_DATA_REF.
11121 * tree-vectorizer.c (vec_info::move_dr): New function.
11122 * tree-vect-patterns.c (vect_recog_bool_pattern)
11123 (vect_recog_mask_conversion_pattern)
11124 (vect_recog_gather_scatter_pattern): Use it.
11125 * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
11126 the "dr" and "stmt" fields of dr_vec_info instead of
11127 STMT_VINFO_DATA_REF.
11129 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11131 * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
11132 (is_pattern_stmt_p): Use it.
11133 * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
11134 on pattern statements.
11136 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11138 * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
11139 original stmt as a stmt_vec_info rather than a gimple stmt.
11140 (vect_pattern_recog_1): Take the statement directly as a
11141 stmt_vec_info, rather than via a gimple_stmt_iterator.
11142 Update call to vect_mark_pattern_stmts.
11143 (vect_pattern_recog): Update calls accordingly.
11145 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11147 * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
11148 (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
11149 a vect_def_type for the first argument.
11150 * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
11151 (vect_get_vec_def_for_stmt_copy): Likewise. Return the original
11152 operand if it isn't defined by a vectorized statement.
11153 (vect_build_gather_load_calls): Remove the mask_dt argument and
11154 update calls to vect_get_vec_def_for_stmt_copy.
11155 (vectorizable_bswap): Likewise the dt argument.
11156 (vectorizable_call): Update calls to vectorizable_bswap and
11157 vect_get_vec_def_for_stmt_copy.
11158 (vectorizable_simd_clone_call, vectorizable_assignment)
11159 (vectorizable_shift, vectorizable_operation, vectorizable_condition)
11160 (vectorizable_comparison): Update calls to
11161 vect_get_vec_def_for_stmt_copy.
11162 (vectorizable_store): Likewise. Remove now-unnecessary calls to
11163 vect_is_simple_use.
11164 (vect_get_loop_based_defs): Remove dt argument and update call
11165 to vect_get_vec_def_for_stmt_copy.
11166 (vectorizable_conversion): Update calls to vect_get_loop_based_defs
11167 and vect_get_vec_def_for_stmt_copy.
11168 (vectorizable_load): Update calls to vect_build_gather_load_calls
11169 and vect_get_vec_def_for_stmt_copy.
11170 * tree-vect-loop.c (vect_create_epilog_for_reduction)
11171 (vectorizable_reduction, vectorizable_live_operation): Update calls
11172 to vect_get_vec_def_for_stmt_copy.
11174 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11176 * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
11177 and gimple stmt arguments with a stmt_vec_info.
11178 (vect_record_base_alignments): Update calls accordingly.
11179 * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
11180 and gimple stmt arguments with a stmt_vec_info.
11181 (vect_build_slp_tree_1): Remove vinfo argument and update call
11182 to vect_record_max_nunits.
11183 (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
11184 and vect_record_max_nunits.
11186 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11188 * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
11189 file and take a stmt_vec_info instead of a gimple stmt.
11190 (supportable_widening_operation, vect_finish_replace_stmt)
11191 (vect_finish_stmt_generation, vect_get_store_rhs)
11192 (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
11193 (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
11194 (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
11195 (vect_get_smallest_scalar_type, vect_check_gather_scatter)
11196 (vect_create_data_ref_ptr, bump_vector_ptr)
11197 (vect_permute_store_chain, vect_setup_realignment)
11198 (vect_transform_grouped_load, vect_record_grouped_load_vectors)
11199 (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
11200 (vectorizable_reduction, vectorizable_induction)
11201 (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
11202 (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
11203 than gimple stmts as arguments.
11204 * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
11205 (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
11206 (can_group_stmts_p, vect_check_gather_scatter)
11207 (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
11208 (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
11209 (vect_permute_load_chain, vect_shift_permute_load_chain)
11210 (vect_transform_grouped_load)
11211 (vect_record_grouped_load_vectors): Likewise.
11212 * tree-vect-loop.c (vect_fixup_reduc_chain)
11213 (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
11214 (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
11215 (vectorizable_reduction, vectorizable_induction)
11216 (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
11217 * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
11218 (vect_get_load_store_mask): Likewise.
11219 * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
11220 (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
11221 * tree-vect-stmts.c (vect_mark_relevant)
11222 (is_simple_and_all_uses_invariant)
11223 (exist_non_indexing_operands_for_use_p, process_use)
11224 (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
11225 (vect_get_vec_def_for_operand, vect_get_vec_defs)
11226 (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
11227 (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
11228 (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
11229 (get_negative_load_store_type, get_load_store_type)
11230 (vect_check_load_store_mask, vect_check_store_rhs)
11231 (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
11232 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11233 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11234 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11235 (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
11236 (vectorizable_load, vectorizable_condition, vectorizable_comparison)
11237 (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
11238 (supportable_widening_operation): Likewise.
11240 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11242 * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
11243 a stmt_vec_info instead of a gcall.
11244 (vect_check_gather_scatter): Update call accordingly.
11245 * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
11247 (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
11248 (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
11249 * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
11250 instead of a gimple stmt.
11251 (vect_transform_loop): Update calls accordingly.
11252 * tree-vect-slp.c (vect_split_slp_store_group): Take and return
11253 stmt_vec_infos instead of gimple stmts.
11254 (vect_analyze_slp_instance): Update use accordingly.
11255 * tree-vect-stmts.c (read_vector_array, write_vector_array)
11256 (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
11257 (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
11258 (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
11259 (vect_gen_widened_results_half, vect_get_loop_based_defs)
11260 (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
11261 Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
11262 down to subroutines.
11264 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11266 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
11267 of the worklist from a vector of gimple stmts to a vector of
11269 * tree-vect-stmts.c (vect_mark_relevant, process_use)
11270 (vect_mark_stmts_to_be_vectorized): Likewise
11272 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11274 * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
11275 statement before passing it to vect_analyze_stmt.
11276 (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
11277 the chain of phi vector definitions. Track the exit phi via its
11279 (vectorizable_reduction): Set cond_stmt_vinfo directly from the
11280 STMT_VINFO_REDUC_DEF.
11281 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
11282 stmt_vec_infos to handle the statement chains.
11283 (vect_get_slp_defs): Record the first statement in the node
11284 using a stmt_vec_info.
11285 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
11286 statements here and pass their stmt_vec_info down to subroutines.
11287 (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
11288 down to vect_finish_stmt_generation.
11289 (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
11290 (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
11291 stmt_vec_infos to subroutines.
11292 (vect_remove_stores): Use stmt_vec_infos to handle the statement
11295 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11297 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
11298 (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
11299 (vect_permute_store_chain, vect_setup_realignment)
11300 (vect_permute_load_chain, vect_shift_permute_load_chain)
11301 (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
11302 stmts internally, and when passing values to other vectorizer routines.
11303 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
11304 * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
11305 (vect_analyze_loop_operations, get_initial_def_for_reduction)
11306 (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
11307 (vectorizable_reduction, vectorizable_induction)
11308 (vectorizable_live_operation, vect_transform_loop_stmt)
11309 (vect_transform_loop): Likewise.
11310 * tree-vect-patterns.c (vect_reassociating_reduction_p)
11311 (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
11312 (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
11313 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11314 (vect_slp_analyze_node_operations_1): Likewise.
11315 * tree-vect-stmts.c (vect_mark_relevant, process_use)
11316 (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
11317 (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
11318 (vect_finish_stmt_generation_1, get_group_load_store_type)
11319 (get_load_store_type, vect_build_gather_load_calls)
11320 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11321 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11322 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11323 (vectorizable_store, vectorizable_load, vectorizable_condition)
11324 (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
11325 (supportable_widening_operation): Likewise.
11326 (vect_get_vector_types_for_stmt): Likewise.
11327 * tree-vectorizer.h (vect_dr_behavior): Likewise.
11329 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11331 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11332 (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
11333 (vect_permute_store_chain, vect_permute_load_chain)
11334 (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
11335 repeated stmt_vec_info lookups.
11336 * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
11337 (vect_update_ivs_after_vectorizer): Likewise.
11338 * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
11339 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11340 * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
11341 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11342 (vect_bb_slp_scalar_cost): Likewise.
11343 * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
11345 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11347 * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
11348 gcall rather than the generic gimple stmt to gimple_call_internal_fn.
11349 (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
11350 to get gassigns and gcalls, rather than operating on generc gimple
11352 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
11353 (vect_mark_stmts_to_be_vectorized, vectorizable_store)
11354 (vectorizable_load, vect_analyze_stmt): Likewise.
11355 * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
11357 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11359 * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
11360 return stmt_vec_infos rather than gimple stmts. Do not accept
11362 (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
11364 * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
11365 Update use of get_later_stmt.
11366 (vect_get_constant_vectors): Update call accordingly.
11367 (vect_schedule_slp_instance): Likewise
11368 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
11369 (vect_slp_analyze_instance_dependence): Likewise.
11370 (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
11372 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11374 * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
11375 (stmt_info_for_cost::stmt_info): ...this new field.
11376 (add_stmt_costs): Update accordingly.
11377 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11378 (vect_get_known_peeling_cost): Likewise.
11379 (vect_estimate_min_profitable_iters): Likewise.
11380 * tree-vect-stmts.c (record_stmt_cost): Likewise.
11382 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11384 * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
11385 from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11386 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
11388 * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
11390 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11392 * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
11393 a gimple stmt to a stmt_vec_info.
11394 * tree-vect-stmts.c (vectorizable_load): Update accordingly.
11396 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11398 * tree-vectorizer.h (vec_info::grouped_stores): Change from
11399 an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11400 (_loop_vec_info::reduction_chains): Likewise.
11401 * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
11403 * tree-vect-slp.c (vect_analyze_slp): Likewise.
11405 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11407 * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
11408 a gimple stmt to a stmt_vec_info.
11409 (_stmt_vec_info::next_element): Likewise.
11410 * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
11411 (vect_slp_analyze_and_verify_node_alignment)
11412 (vect_analyze_group_access_1, vect_analyze_group_access)
11413 (vect_small_gap_p, vect_prune_runtime_alias_test_list)
11414 (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
11415 (vect_supportable_dr_alignment): Update accordingly.
11416 * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
11417 (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
11418 (vect_is_simple_reduction, vectorizable_reduction): Likewise.
11419 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
11420 * tree-vect-slp.c (vect_build_slp_tree_1)
11421 (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
11422 (vect_split_slp_store_group, vect_analyze_slp_instance)
11423 (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
11424 * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
11425 (get_group_load_store_type, get_load_store_type)
11426 (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
11427 (vect_transform_stmt, vect_remove_stores): Likewise.
11429 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11431 * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
11432 than a gimple stmt.
11433 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11434 (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
11435 (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
11436 (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
11437 (vector_alignment_reachable_p, vect_get_data_access_cost)
11438 (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
11439 (vect_peeling_supportable, vect_enhance_data_refs_alignment)
11440 (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
11441 (vect_analyze_group_access_1, vect_analyze_group_access)
11442 (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
11443 (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
11444 (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
11445 result of vect_dr_stmt and use the stmt_vec_info instead of
11446 the associated gimple stmt.
11447 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11448 (vect_gen_prolog_loop_niters): Likewise.
11449 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11451 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11453 * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
11454 to a vec<stmt_vec_info>.
11455 * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
11456 (vect_create_new_slp_node): Take a vec<gimple *> instead of a
11457 vec<stmt_vec_info>.
11458 (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
11459 to a vec<stmt_vec_info>.
11460 (bst_traits::value_type, bst_traits::value_type): Likewise.
11461 (bst_traits::hash): Update accordingly.
11462 (vect_get_and_check_slp_defs): Change the stmts parameter from
11463 a vec<gimple *> to a vec<stmt_vec_info>.
11464 (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
11465 (vect_build_slp_tree): Likewise.
11466 (vect_build_slp_tree_2): Likewise. Update uses of
11467 SLP_TREE_SCALAR_STMTS.
11468 (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
11469 (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
11470 (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
11471 (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
11472 (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
11473 (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
11474 (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
11475 (vect_get_constant_vectors, vect_get_slp_defs)
11476 (vect_transform_slp_perm_load, vect_schedule_slp_instance)
11477 (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
11478 (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
11479 instead of gimple stmts.
11480 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
11481 the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
11482 (vect_slp_analyze_instance_dependence): Update uses of
11483 SLP_TREE_SCALAR_STMTS.
11484 (vect_slp_analyze_and_verify_node_alignment): Likewise.
11485 (vect_slp_analyze_and_verify_instance_alignment): Likewise.
11486 * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
11487 (get_initial_defs_for_reduction): Likewise.
11488 (vect_create_epilog_for_reduction): Likewise.
11489 (vectorize_fold_left_reduction): Likewise.
11490 * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
11491 (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
11492 (can_vectorize_live_stmts): Likewise.
11494 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11496 * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
11497 auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11498 (vect_force_simple_reduction): Take and return stmt_vec_infos rather
11500 * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
11502 (gather_scalar_reductions): Update after above interface changes.
11503 * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
11504 (vect_is_simple_reduction): Take and return stmt_vec_infos rather
11506 (vect_force_simple_reduction): Likewise.
11507 * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
11508 LOOP_VINFO_REDUCTIONS.
11509 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11511 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11513 * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
11514 a gimple stmt to a stmt_vec_info.
11515 * tree-vect-loop.c (vect_active_double_reduction_p)
11516 (vect_force_simple_reduction, vectorizable_reduction): Update
11519 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11521 * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
11522 vec<gimple *> to a vec<stmt_vec_info>.
11523 * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
11524 the reduction_phis argument from a vec<gimple *> to a
11525 vec<stmt_vec_info>.
11526 (vectorizable_reduction): Likewise the phis local variable that
11527 is passed to vect_create_epilog_for_reduction. Update for new type
11528 of SLP_TREE_VEC_STMTS.
11529 (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
11530 (vectorizable_live_operation): Likewise.
11531 * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
11532 (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
11534 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11536 * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
11537 a gimple stmt to a stmt_vec_info.
11538 (vectorizable_condition, vectorizable_live_operation)
11539 (vectorizable_reduction, vectorizable_induction): Pass back the
11540 vectorized statement as a stmt_vec_info.
11541 * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
11542 use of STMT_VINFO_VEC_STMT.
11543 * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
11544 accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
11545 as stmt_vec_infos rather than gimple stmts.
11546 (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
11547 to a stmt_vec_info.
11548 (vectorizable_live_operation): Likewise.
11549 (vectorizable_reduction, vectorizable_induction): Likewise,
11550 updating use of STMT_VINFO_VEC_STMT.
11551 * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
11552 of STMT_VINFO_VEC_STMT.
11553 (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
11554 (vectorizable_simd_clone_call, vectorizable_conversion)
11555 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11556 (vectorizable_store, vectorizable_load, vectorizable_condition)
11557 (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
11558 from a gimple stmt to a stmt_vec_info.
11559 (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT. Pass a
11560 pointer to a stmt_vec_info to the vectorizable_* routines.
11562 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11564 * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
11565 a gimple stmt to a stmt_vec_info.
11566 (is_pattern_stmt_p): Update accordingly.
11567 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
11568 (vect_record_grouped_load_vectors): Likewise.
11569 * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
11570 (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
11571 (vect_model_reduction_cost): Likewise.
11572 (vect_create_epilog_for_reduction): Likewise.
11573 (vectorizable_reduction, vectorizable_induction): Likewise.
11574 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11575 Return the stmt_vec_info for the pattern statement.
11576 (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
11577 (vect_split_statement, vect_mark_pattern_stmts): Likewise.
11578 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11579 (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
11580 * tree-vect-stmts.c (vect_mark_relevant): Likewise.
11581 (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
11582 (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
11583 (free_stmt_vec_info, vect_is_simple_use): Likewise.
11585 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11587 * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
11588 (vect_finish_stmt_generation): Likewise.
11589 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11590 (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
11591 (vect_build_gather_load_calls): Use the return value of the above
11592 functions instead of a separate call to vinfo_for_stmt. Use narrow
11593 scopes for the input gimple stmt and wider scopes for the associated
11594 stmt_vec_info. Use vec_info::lookup_def when setting these
11595 stmt_vec_infos from an SSA_NAME definition.
11596 (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11597 (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11598 (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11599 (vectorizable_store, vectorizable_load, vectorizable_condition)
11600 (vectorizable_comparison): Likewise.
11601 * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
11602 (vectorizable_reduction): Likewise.
11604 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11606 * tree-vectorizer.h (vect_is_simple_use): Add an optional
11607 stmt_vec_info * parameter before the optional gimple **.
11608 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11609 (process_use, vect_get_vec_def_for_operand_1): Update callers.
11610 (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
11611 * tree-vect-loop.c (vectorizable_reduction): Likewise.
11612 (vectorizable_live_operation): Likewise.
11613 * tree-vect-patterns.c (type_conversion_p): Likewise.
11614 (vect_look_through_possible_promotion): Likewise.
11615 (vect_recog_rotate_pattern): Likewise.
11616 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11618 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11620 * tree-vectorizer.h (stmt_vec_info): Temporarily change from
11621 a typedef to a wrapper class.
11622 (NULL_STMT_VEC_INFO): New macro.
11623 (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
11624 (stmt_vec_info::operator*): New function.
11625 (stmt_vec_info::operator gimple *): Likewise.
11626 (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
11627 (add_stmt_costs): Likewise.
11628 * tree-vect-loop-manip.c (iv_phi_p): Likewise.
11629 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11630 (vect_get_known_peeling_cost): Likewise.
11631 (vect_estimate_min_profitable_iters): Likewise.
11632 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11633 * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
11634 * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
11635 (vectorizable_store, free_stmt_vec_infos): Likewise.
11636 (new_stmt_vec_info): Change return type of xcalloc to
11639 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11641 * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
11642 * tree-vectorizer.c (vec_info::lookup_single_use): New function.
11643 * tree-vect-loop.c (vectorizable_reduction): Use it instead of
11644 a single_imm_use-based sequence.
11645 * tree-vect-stmts.c (supportable_widening_operation): Likewise.
11647 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11649 * tree-vectorizer.h (vec_info::lookup_def): Declare.
11650 * tree-vectorizer.c (vec_info::lookup_def): New function.
11651 * tree-vect-patterns.c (vect_get_internal_def): Use it.
11652 (vect_widened_op_tree): Likewise.
11653 * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11654 * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
11655 (vectorizable_reduction): Likewise.
11656 (vect_valid_reduction_input_p): Take a stmt_vec_info instead
11658 (vect_is_slp_reduction): Update calls accordingly. Use
11659 vec_info::lookup_def.
11660 (vect_is_simple_reduction): Likewise
11661 * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
11663 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11665 * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
11666 * tree-vectorizer.c (vec_info::lookup_stmt): New function.
11667 * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
11669 (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
11670 (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
11671 (vect_update_vf_for_slp, vect_analyze_loop_operations)
11672 (vect_is_slp_reduction, vectorizable_induction)
11673 (vect_transform_loop_stmt, vect_transform_loop): Likewise.
11674 * tree-vect-patterns.c (vect_init_pattern_stmt):
11675 (vect_determine_min_output_precision_1, vect_determine_precisions)
11676 (vect_pattern_recog): Likewise.
11677 * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
11678 * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
11679 * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
11680 * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11681 (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
11682 (vect_detect_hybrid_slp): Likewise. Change the walk_stmt_info
11683 info field from a loop to a loop_vec_info.
11685 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11687 * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
11688 (vec_info::add_stmt): Declare.
11689 * tree-vectorizer.c (vec_info::add_stmt): New function.
11690 * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
11691 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
11692 (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11693 (vectorizable_induction): Likewise.
11694 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
11695 * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11696 (vectorizable_simd_clone_call, vectorizable_store): Likewise.
11697 (vectorizable_load): Likewise.
11698 * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11699 (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
11700 (vect_recog_gather_scatter_pattern): Likewise.
11701 (append_pattern_def_seq): Likewise. Remove a check that is
11702 performed by add_stmt itself.
11704 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11706 * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
11707 which make_ssa_name was called with new_stmt before new_stmt
11710 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11712 * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
11714 (vect_is_slp_reduction): ...here...
11715 (vect_is_simple_reduction): ...and here. Remove repetition of tests
11716 that are already known to be false.
11718 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11720 * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
11721 * tree-vect-slp.c (vect_free_slp_tree): Likewise. Don't update
11722 STMT_VINFO_NUM_SLP_USES when it's true.
11723 (vect_free_slp_instance): Add a final_p parameter and pass it to
11724 vect_free_slp_tree.
11725 (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
11726 (vect_analyze_slp_instance): Likewise.
11727 (vect_slp_analyze_operations): Likewise.
11728 (vect_slp_analyze_bb_1): Likewise.
11729 * tree-vectorizer.c (vec_info): Likewise.
11730 * tree-vect-loop.c (vect_transform_loop): Likewise.
11732 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11734 * tree-vect-loop.c (vectorizable_reduction): Assert that the
11735 function is not called for second and subsequent members of
11738 2018-07-31 Richard Sandiford <richard.sandiford@arm.com>
11740 * tree-vect-loop.c (get_initial_def_for_reduction): Move special
11741 cases for nested loops from here to ...
11742 (vect_create_epilog_for_reduction): ...here. Only call
11743 vect_is_simple_use for inner-loop reductions.
11745 2018-07-31 Martin Liska <mliska@suse.cz>
11747 PR gcov-profile/85338
11748 PR gcov-profile/85350
11749 PR gcov-profile/85372
11750 * profile.c (struct location_triplet): New.
11751 (struct location_triplet_hash): Likewise.
11752 (output_location): Do not output a BB that
11753 is already recorded for a line.
11754 (branch_prob): Use streamed_locations.
11756 2018-07-31 Martin Liska <mliska@suse.cz>
11758 PR gcov-profile/85370
11759 * coverage.c (coverage_begin_function): Do not mark target
11760 clones as artificial functions.
11762 2018-07-31 Martin Liska <mliska@suse.cz>
11764 PR gcov-profile/83813
11765 PR gcov-profile/84758
11766 PR gcov-profile/85217
11767 PR gcov-profile/85332
11768 * profile.c (branch_prob): Do not record GOTO expressions
11769 for GIMPLE statements which locations are already streamed.
11771 2018-07-31 Olivier Hainque <hainque@adacore.com>
11773 * gcc.c (handle_spec_function): Accept a soft_matched_part
11774 argument, as do_spec_1. Pass it down to ...
11775 (eval_spec_function): Accept a soft_matched_part argument,
11776 and pass it down to ...
11777 (do_spec_2): Accept a soft_matched_part argument, and pass
11778 it down to do_spec_1.
11779 (do_spec_1): Pass soft_matched_part to handle_spec_function.
11780 (handle_braces): Update call to handle_spec_function.
11781 (driver::set_up_specs): Update calls to do_spec_2.
11782 (compare_debug_dump_opt_spec_function): Likewise.
11783 (compare_debug_self_opt_spec_function): Likewise.
11785 2018-07-31 Olivier Hainque <hainque@adacore.com>
11787 * common.opt (nolibc): New option.
11788 * doc/invoke.texi (Link Options): Document it.
11789 * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
11790 * config/alpha/linux.h: Likewise.
11791 * config/arc/elf.h: Likewise.
11792 * config/arm/uclinux-elf.h: Likewise.
11793 * config/arm/unknown-elf.h: Likewise.
11794 * config/avr/avrlibc.h: Likewise.
11795 * config/bfin/bfin.h: Likewise.
11796 * config/bfin/linux.h: Likewise.
11797 * config/bfin/uclinux.h: Likewise.
11798 * config/darwin.h: Likewise.
11799 * config/darwin10.h: Likewise.
11800 * config/darwin12.h: Likewise.
11801 * config/gnu-user.h: Likewise.
11802 * config/lm32/uclinux-elf.h: Likewise.
11803 * config/pa/pa-hpux11.h: Likewise.
11804 * config/pa/pa64-hpux.h: Likewise.
11805 * config/sparc/sparc.h: Likewise.
11807 2018-07-31 Olivier Hainque <hainque@adacore.com>
11809 * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
11810 undefined variables.
11812 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
11815 * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
11816 instead of GEN_INT.
11818 2018-07-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
11820 * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
11821 terminated string literal.
11823 2018-07-30 Segher Boessenkool <segher@kernel.crashing.org>
11825 PR rtl-optimization/85160
11826 * combine.c (is_just_move): New function.
11827 (try_combine): Allow combining two instructions into two if neither of
11828 the original instructions was a move.
11830 2018-07-30 Alexander Monakov <amonakov@ispras.ru>
11833 * doc/extend.texi (Global Register Variables): Discourage use of type
11835 (Local Register Variables): Likewise.
11837 2018-07-30 Richard Sandiford <richard.sandiford@arm.com>
11839 PR tree-optimization/86506
11840 * hwint.h (ceil_log2): Resync with hwint.c implementation.
11842 2018-07-30 Ilya Leoshkevich <iii@linux.ibm.com>
11845 * lra-constraints.c (spill_hard_reg_in_range): When selecting the
11846 hard_regno, make sure no insn between `from` and `to` clobbers it.
11848 2018-07-30 Cesar Philippidis <cesar@codesourcery.com>
11849 Tom de Vries <tdevries@suse.de>
11851 * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
11852 (PTX_DEFAULT_RUNTIME_DIM): ... this.
11853 (nvptx_goacc_validate_dims): Set default worker and gang dims to
11854 PTX_DEFAULT_RUNTIME_DIM.
11855 (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
11857 2018-07-29 John David Anglin <danglin@gcc.gnu.org>
11859 * config/pa/pa.c (pa_output_addr_vec): Align address table.
11860 * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
11861 * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
11863 2018-07-27 Michael Meissner <meissner@linux.ibm.com>
11865 * config/rs6000/constraints.md (wG constraint): Delete, no longer
11867 * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
11868 predicate to reflect toc fusion has been deleted.
11869 (toc_fusion_mem_raw): Delete, no longer used.
11870 (toc_fusion_mem_wrapped): Likewise.
11871 * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
11873 * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
11874 Delete, no longer used.
11875 * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
11876 meant to be used for toc fusion.
11877 (rs6000_debug_print_mode): Delete toc fusion debugging.
11878 (rs6000_debug_reg_global): Likewise.
11879 (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
11880 fusion and secondary reload support that were never used.
11881 (rs6000_option_override_internal): Delete TOC fusion, that was only
11882 partially defined, and it did not work unless you also used the
11884 (rs6000_legitimate_address_p): Delete TOC fusion support.
11885 (rs6000_opt_masks): Likewise.
11886 (fusion_wrap_memory_address): Delete function, no longer used.
11887 (fusion_split_address); Delete TOC fusion support.
11888 * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
11889 longer used with toc fusion being deleted.
11890 (TARGET_TOC_FUSION_FP): Likewise.
11891 * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
11893 (toc fusion spliter): Delete TOC fusion support.
11894 (toc_fusionload_<mode>): Likewise.
11895 (toc_fusionload_di): Likewise.
11896 (fusion_gpr_load_<mode>): Delete generator function, this insn no
11897 longer needs to be named. Rename predicate to delete TOC fusion.
11898 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11899 (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11900 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11901 (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11902 (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
11904 2018-07-27 Kelvin Nilsen <kelvin@gcc.gnu.org>
11906 * doc/extend.texi (Basic PowerPC Built-in Functions Available on
11907 ISA 2.05): Replace __uint128_t with __uint128 and __int128_t with
11908 __int128 in built-in function prototypes.
11909 (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
11910 (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
11912 2018-07-27 Martin Sebor <msebor@redhat.com>
11914 PR tree-optimization/86696
11915 * tree-ssa-strlen.c (get_min_string_length): Handle all integer
11916 types, including enums.
11917 (handle_char_store): Be prepared for the above function to fail.
11919 2018-07-26 Qing Zhao <qing.zhao@oracle.com>
11921 * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
11922 when optimization level is lower than 2 or optimize for size.
11924 2018-07-26 Martin Sebor <msebor@redhat.com>
11926 PR tree-optimization/86043
11927 PR tree-optimization/86042
11928 * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
11929 (get_string_cst_length): Rename...
11930 (get_min_string_length): ...to this. Add argument.
11931 (handle_char_store): Extend to handle multi-character stores by
11933 * tree.c (initializer_zerop): Use new argument. Handle MEM_REF.
11934 * tree.h (initializer_zerop): Add argument.
11936 2018-07-26 Jakub Jelinek <jakub@redhat.com>
11938 PR middle-end/86660
11939 * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
11940 declare target to variables if they have always,{to,from,tofrom} map
11943 2018-07-26 Martin Liska <mliska@suse.cz>
11946 * lto-wrapper.c: Add linker_output as prefix
11947 for ltrans_output_file.
11949 2018-07-26 Segher Boessenkool <segher@kernel.crashing.org>
11951 PR rtl-optimization/85805
11952 * combine.c (reg_nonzero_bits_for_combine): Only use the last set
11953 value for hard registers if that was written in the same mode.
11955 2018-07-26 Martin Liska <mliska@suse.cz>
11957 PR gcov-profile/86536
11958 * gcov.c (format_gcov): Use printf format %.*f directly
11959 and do not handle special values.
11961 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
11963 * common/config/arc/arc-common.c (arc_option_optimization_table):
11964 Update default optimizations for size.
11966 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
11968 * config/arc/arc.md (movsf_insn): Add short instruction selection.
11969 * config/arc/constraints.md (CfZ): New constraint.
11970 * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
11971 (subsf3_fpu): Likewise.
11972 (cmpsf_fpu): Likewise.
11973 (cmpsf_fpu_uneq): Likewise.
11975 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
11977 * config/arc/arc.c (compact_memory_operand_p): Check for uncached
11979 (arc_is_uncached_mem_p): uncached applies to both the variable and
11982 2018-07-25 Claudiu Zissulescu <claziss@synopsys.com>
11984 * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
11987 2018-07-25 David Malcolm <dmalcolm@redhat.com>
11989 * optinfo-emit-json.cc (class optrecord_json_writer): Convert
11990 field "m_scopes" from vec to auto_vec.
11992 2018-07-25 Martin Liska <mliska@suse.cz>
11994 * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
11997 2018-07-25 Richard Biener <rguenther@suse.de>
12000 * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
12001 special wrt context_die late.
12002 (gen_subprogram_die): Re-use DIEs in local scope.
12004 2018-07-25 Richard Sandiford <richard.sandiford@arm.com>
12006 PR tree-optimization/86644
12007 * hwint.c (ceil_log2): Fix comment. Return 0 for 0.
12009 2018-07-25 Martin Liska <mliska@suse.cz>
12011 PR middle-end/86645
12012 * dumpfile.c: And excluded values with TDF_ALL_VALUES.
12013 * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
12015 2018-07-25 Martin Liska <mliska@suse.cz>
12018 * params.def: Explain ASan abbreviation and provide
12019 a documentation link.
12021 2018-07-24 Martin Sebor <msebor@redhat.com>
12023 PR tree-optimization/86622
12024 PR tree-optimization/86532
12025 * builtins.h (string_length): Declare.
12026 * builtins.c (c_strlen): Correct handling of non-constant offsets.
12027 (check_access): Be prepared for non-constant length ranges.
12028 (string_length): Make extern.
12029 * expr.c (string_constant): Only handle the minor non-constant
12030 array index. Use string_constant to compute the length of
12031 a generic string constant.
12033 2018-07-24 Richard Sandiford <richard.sandiford@arm.com>
12035 PR tree-optimization/86618
12036 * tree-vect-stmts.c (vectorizable_call): Don't take the address
12037 of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
12039 2018-07-24 David Malcolm <dmalcolm@redhat.com>
12041 PR tree-optimization/86636
12042 * json.cc (json::object::set): Fix comment. Add assertions.
12043 (json::array::append): Move here from json.h. Add comment and an
12045 (json::string::string): Likewise.
12046 * json.h (json::array::append): Move to json.cc.
12047 (json::string::string): Likewise.
12048 * optinfo-emit-json.cc
12049 (optrecord_json_writer::impl_location_to_json): Assert that we
12050 aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
12051 wrapper around it. Expand the location once, rather than three
12053 (optrecord_json_writer::inlining_chain_to_json): Fix the check for
12054 UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
12056 (optrecord_json_writer::optinfo_to_json): Likewise, in four
12057 places. Fix some overlong lines.
12059 2018-07-24 Matthew Malcomson <matthew.malcomson@arm.com>
12061 * config/aarch64/aarch64-simd.md
12062 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
12063 (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
12064 (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
12065 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
12066 (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
12067 (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
12068 (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
12069 (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
12070 (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
12072 2018-07-24 Jakub Jelinek <jakub@redhat.com>
12074 PR middle-end/86627
12075 * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
12076 and size > HOST_BITS_PER_WIDE_INT. For size > HOST_BITS_PER_WIDE_INT
12077 and abs_d == d, do the power of two handling if profitable.
12079 2018-07-24 Richard Biener <rguenther@suse.de>
12081 * match.pd: Add BIT_FIELD_REF canonicalizations.
12083 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
12086 * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
12088 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
12090 * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
12091 terminated STRING_CST object.
12093 2018-07-23 Bernd Edlinger <bernd.edlinger@hotmail.de>
12095 hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
12097 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
12099 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
12100 * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
12101 * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
12102 the elements into a register.
12103 (rs6000_split_v4si_init_di_reg): Delete.
12104 (rs6000_split_v4si_init): Delete.
12105 * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
12106 (vsx_init_v4si): Rewrite as a define_expand.
12108 2018-07-23 Segher Boessenkool <segher@kernel.crashing.org>
12110 * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
12111 zero_extend argument from memory): New.
12113 2018-07-22 Martin Sebor <msebor@redhat.com>
12116 * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
12117 diagnosing calls with unknown arguments unless -Walloca-larger-than
12118 is restricted to less than PTRDIFF_MAX bytes.
12120 2018-07-22 Gerald Pfeifer <gerald@pfeifer.com>
12122 * doc/gcov.texi (Invoking Gcov): Editorial changes.
12124 2018-07-20 David Malcolm <dmalcolm@redhat.com>
12126 * pretty-print.c (text_info::set_location): Remove redundant
12127 "line_table" parameter from call to rich_location::set_range.
12129 2018-07-20 Martin Sebor <msebor@redhat.com>
12131 PR middle-end/82063
12132 * builtins.c (expand_builtin_alloca): Adjust.
12133 * calls.c (alloc_max_size): Simplify.
12134 * cgraphunit.c (cgraph_node::expand): Adjust.
12135 * common.opt (larger_than_size, warn_frame_larger_than): Remove
12137 (frame_larger_than_size): Same.
12138 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
12139 to take a HOST_WIDE_INT argument and accept a byte-size suffix.
12141 * doc/invoke.texi (GCC Command Options): Document option arguments.
12142 Explain byte-size arguments and suffixes.
12143 (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
12144 (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
12145 (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
12146 * doc/options.texi (UInteger): Expand.
12147 (Host_Wide_Int, ByteSize): Document new properties.
12148 * final.c (final_start_function_1): Include sizes in an error message.
12149 * function.c (frame_offset_overflow): Same.
12150 * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
12151 (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
12152 Diagnose unbounded alloca calls only for limits of less than
12154 (alloca_call_type): Adjust. Diagnose possibly out-of-bounds alloca
12155 calls and VLA size only for limits of less than PTRDIFF_MAX. Same
12157 (pass_walloca::execute): Adjust. Diagnose alloca calls in loops
12158 only for limits of less than PTRDIFF_MAX.
12159 * langhooks-def.h (lhd_handle_option): Change function argument
12161 * langhooks.c (lhd_handle_option): Same.
12162 * langhooks.h (handle_option): Same.
12163 * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
12165 (var_type, var_type_struct): Same.
12166 (var_set): Handle ByteSize flag.
12167 * optc-gen.awk: Add comments to output to ease debugging. Make
12168 use of HOST_WIDE_INT where appropriate.
12169 * opts-gen-save.awk: Use %lx to format unsigned long.
12170 * opth-gen.awk: Change function argument to HOST_WIDE_INT.
12171 * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
12172 arguments. Parse bytes-size suffixes.
12173 (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
12174 (enum_value_to_arg): Same.
12175 (decode_cmdline_option): Handle cl_host_wide_int. Adjust.
12176 (handle_option): Adjust.
12177 (generate_option): Change function argument to HOST_WIDE_INT.
12178 (cmdline_handle_error): Adjust.
12179 (read_cmdline_option): Change function argument to HOST_WIDE_INT.
12180 (set_option): Change function argument to HOST_WIDE_INT.
12181 (option_enabled): Handle cl_host_wide_int.
12182 (get_option_state): Handle CLVC_SIZE.
12183 (control_warning_option): Same.
12184 * opts.c (common_handle_option): Change function argument to
12185 HOST_WIDE_INT. Remove handling of OPT_Walloca_larger_than_ and
12186 OPT_Wvla_larger_than_.
12187 * opts.h (enum cl_var_type): Add an enumerator.
12188 * stor-layout.c (layout_decl): Print a more meaningful warning.
12189 * toplev.c (output_stack_usage): Adjust.
12191 2018-07-20 Qing Zhao <qing.zhao@oracle.com>
12193 * builtins.c (expand_builtin_memcmp): Delete the last parameter for
12194 call to inline_expand_builtin_string_cmp.
12195 (expand_builtin_strcmp): Likewise.
12196 (expand_builtin_strncmp): Likewise.
12197 (inline_string_cmp): Delete the last parameter, change char_type_node
12198 to unsigned_char_type_node for strcmp/strncmp, add conversions to the
12200 (inline_expand_builtin_string_cmp): Delete the last parameter, give up
12201 the inlining expansion on target where the type of the call has same or
12202 narrower precision than unsigned char.
12204 2018-07-20 David Malcolm <dmalcolm@redhat.com>
12206 * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
12207 (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
12208 * common.opt (fsave-optimization-record): New option.
12209 * coretypes.h (struct kv_pair): Move here from dumpfile.c.
12210 * doc/invoke.texi (-fsave-optimization-record): New option.
12211 * dumpfile.c: Include "optinfo-emit-json.h".
12212 (struct kv_pair): Move to coretypes.h.
12213 (optgroup_options): Make non-static.
12214 (dump_context::end_scope): Call
12215 optimization_records_maybe_pop_dump_scope.
12216 * dumpfile.h (optgroup_options): New decl.
12217 * json.cc: New file.
12218 * json.h: New file.
12219 * optinfo-emit-json.cc: New file.
12220 * optinfo-emit-json.h: New file.
12221 * optinfo.cc: Include "optinfo-emit-json.h".
12222 (optinfo::emit): Call optimization_records_maybe_record_optinfo.
12223 (optinfo_enabled_p): Check optimization_records_enabled_p.
12224 (optinfo_wants_inlining_info_p): Likewise.
12225 * optinfo.h: Update comment.
12226 * profile-count.c (profile_quality_as_string): New function.
12227 * profile-count.h (profile_quality_as_string): New decl.
12228 (profile_count::quality): New accessor.
12229 * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
12230 and optinfo_emit_json_cc_tests.
12231 * selftest.h (selftest::json_cc_tests): New decl.
12232 (selftest::optinfo_emit_json_cc_tests): New decl.
12233 * toplev.c: Include "optinfo-emit-json.h".
12234 (compile_file): Call optimization_records_finish.
12235 (do_compile): Call optimization_records_start.
12236 * tree-ssa-live.c: Include optinfo.h.
12237 (remove_unused_scope_block_p): Retain inlining information if
12238 optinfo_wants_inlining_info_p returns true.
12240 2018-07-20 Richard Biener <rguenther@suse.de>
12243 * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
12244 to cover -flto-partition=none.
12246 2018-07-20 Martin Liska <mliska@suse.cz>
12248 * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
12249 (get_decl_source_range): Remove unused function.
12251 2018-07-20 Richard Biener <rguenther@suse.de>
12253 * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
12254 (struct vn_phi_s): Likewise.
12255 (struct vn_reference_s): Likewise.
12256 * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
12257 for searching the slot of an entry known to be in the hash itself.
12258 (vn_phi_hasher::equal): Likewise.
12259 (vn_reference_hasher::equal): Likewise.
12260 (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
12262 (optimistic_info, current_info): Remove, keeping only valid_info.
12263 (vn_reference_lookup_1): Remove fallback lookup.
12264 (vn_reference_lookup_2): Likewise.
12265 (vn_nary_op_lookup_1): Likewise.
12266 (vn_phi_lookup): Likewise.
12267 (vn_nary_build_or_lookup_1): Make sure to not chain the built
12269 (vn_reference_insert): Adjust, chain the inserted hash element
12270 at last_inserted_ref.
12271 (vn_reference_insert_pieces): Likewise.
12272 (visit_reference_op_call): Likewise.
12273 (vn_nary_op_insert_into): Chain the inserted hash element at
12274 last_inserted_nary.
12275 (vn_nary_op_insert_pieces): Adjust.
12276 (vn_nary_op_insert): Likewise.
12277 (vn_nary_op_insert_stmt): Likewise.
12278 (vn_phi_insert): Adjust, chain the inserted hash element at
12280 (process_scc): Remove clearing and copying the optimistic
12281 table. Instead remove elements inserted during an optimistic
12282 iteration from the single table we maintain.
12283 (init_scc_vn): Adjust.
12284 (free_scc_vn): Likewise.
12285 (sccvn_dom_walker::record_cond): Likewise.
12286 (sccvn_dom_walker::after_dom_children): Likewise.
12288 2018-07-19 Martin Sebor <msebor@redhat.com>
12290 PR tree-optimization/84047
12291 PR tree-optimization/83776
12292 * tree-vrp.c (vrp_prop::check_mem_ref): New function.
12293 (check_array_bounds): Call it.
12295 2018-07-19 Martin Sebor <msebor@redhat.com>
12297 * align.h (align_flags): Use member initialization.
12299 2018-07-19 David Malcolm <dmalcolm@redhat.com>
12301 * Makefile.in (OBJS): Add optinfo.o.
12302 * coretypes.h (class symtab_node): New forward decl.
12303 (struct cgraph_node): New forward decl.
12304 (class varpool_node): New forward decl.
12305 * dump-context.h: New file.
12306 * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
12308 (refresh_dumps_are_enabled): Use optinfo_enabled_p.
12309 (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
12310 (set_alt_dump_file): Likewise.
12311 (dump_context::~dump_context): New dtor.
12312 (dump_gimple_stmt): Move implementation to...
12313 (dump_context::dump_gimple_stmt): ...this new member function.
12314 Add the stmt to any pending optinfo, creating one if need be.
12315 (dump_gimple_stmt_loc): Move implementation to...
12316 (dump_context::dump_gimple_stmt_loc): ...this new member function.
12317 Start a new optinfo and add the stmt to it.
12318 (dump_gimple_expr): Move implementation to...
12319 (dump_context::dump_gimple_expr): ...this new member function.
12320 Add the stmt to any pending optinfo, creating one if need be.
12321 (dump_gimple_expr_loc): Move implementation to...
12322 (dump_context::dump_gimple_expr_loc): ...this new member function.
12323 Start a new optinfo and add the stmt to it.
12324 (dump_generic_expr): Move implementation to...
12325 (dump_context::dump_generic_expr): ...this new member function.
12326 Add the tree to any pending optinfo, creating one if need be.
12327 (dump_generic_expr_loc): Move implementation to...
12328 (dump_context::dump_generic_expr_loc): ...this new member
12329 function. Add the tree to any pending optinfo, creating one if
12331 (dump_printf): Move implementation to...
12332 (dump_context::dump_printf_va): ...this new member function. Add
12333 the text to any pending optinfo, creating one if need be.
12334 (dump_printf_loc): Move implementation to...
12335 (dump_context::dump_printf_loc_va): ...this new member function.
12336 Start a new optinfo and add the stmt to it.
12337 (dump_dec): Move implementation to...
12338 (dump_context::dump_dec): ...this new member function. Add the
12339 value to any pending optinfo, creating one if need be.
12340 (dump_context::dump_symtab_node): New member function.
12341 (dump_context::get_scope_depth): New member function.
12342 (dump_context::begin_scope): New member function.
12343 (dump_context::end_scope): New member function.
12344 (dump_context::ensure_pending_optinfo): New member function.
12345 (dump_context::begin_next_optinfo): New member function.
12346 (dump_context::end_any_optinfo): New member function.
12347 (dump_context::s_current): New global.
12348 (dump_context::s_default): New global.
12349 (dump_scope_depth): Delete global.
12350 (dumpfile_ensure_any_optinfo_are_flushed): New function.
12351 (dump_symtab_node): New function.
12352 (get_dump_scope_depth): Reimplement in terms of dump_context.
12353 (dump_begin_scope): Likewise.
12354 (dump_end_scope): Likewise.
12355 (selftest::temp_dump_context::temp_dump_context): New ctor.
12356 (selftest::temp_dump_context::~temp_dump_context): New dtor.
12357 (selftest::verify_item): New function.
12358 (ASSERT_IS_TEXT): New macro.
12359 (ASSERT_IS_TREE): New macro.
12360 (ASSERT_IS_GIMPLE): New macro.
12361 (selftest::test_capture_of_dump_calls): New test.
12362 (selftest::dumpfile_c_tests): Call it.
12363 * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
12364 (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
12365 (dump_gimple_stmt, dump_dec): Gather these related decls and add a
12366 descriptive comment.
12367 (dump_function, print_combine_total_stats, enable_rtl_dump_file)
12368 (dump_node, dump_bb): Move these unrelated decls.
12369 (class dump_manager): Add leading comment.
12370 * optinfo.cc: New file.
12371 * optinfo.h: New file.
12373 2018-07-19 Michael Collison <michael.collison@arm.com>
12374 Richard Henderson <rth@redhat.com>
12376 * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
12377 (subti): Handle op1 zero.
12378 (subvti4, usub4ti4): New.
12379 (*sub<GPI>3_compare1_imm): New.
12380 (sub<GPI>3_carryinCV): New.
12381 (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
12382 (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
12384 2018-07-19 Michael Collison <michael.collison@arm.com>
12385 Richard Henderson <rth@redhat.com>
12387 * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
12388 (addti3): Create simpler code if low part is already known to be 0.
12389 (addvti4, uaddvti4): New.
12390 (*add<GPI>3_compareC_cconly_imm): New.
12391 (*add<GPI>3_compareC_cconly): New.
12392 (*add<GPI>3_compareC_imm): New.
12393 (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
12394 handle constants within this pattern..
12395 (*add<GPI>3_compareV_cconly_imm): New.
12396 (*add<GPI>3_compareV_cconly): New.
12397 (*add<GPI>3_compareV_imm): New.
12398 (add<GPI>3_compareV): New.
12399 (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
12400 (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
12401 (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
12402 ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
12403 with 'comparison' operator.
12404 (*add<GPI>3_compareV_cconly_imm): Ditto.
12405 (*add<GPI>3_compareV_cconly): Ditto.
12406 (*add<GPI>3_compareV_imm): Ditto.
12407 (add<GPI>3_compareV): Ditto.
12408 (add<mode>3_carryinC): Ditto.
12409 (*add<mode>3_carryinC_zero): Ditto.
12410 (*add<mode>3_carryinC): Ditto.
12411 (add<mode>3_carryinV): Ditto.
12412 (*add<mode>3_carryinV_zero): Ditto.
12413 (*add<mode>3_carryinV): Ditto.
12415 2018-07-19 Michael Collison <michael.collison@arm.com>
12416 Richard Henderson <rth@redhat.com>
12418 * config/aarch64/aarch64-modes.def (CC_V): New.
12419 * config/aarch64/aarch64-protos.h
12420 (aarch64_addti_scratch_regs): Declare
12421 (aarch64_subvti_scratch_regs): Declare.
12422 (aarch64_expand_subvti): Declare.
12423 (aarch64_gen_unlikely_cbranch): Declare
12424 * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
12425 for signed overflow using CC_Vmode.
12426 (aarch64_get_condition_code_1): Handle CC_Vmode.
12427 (aarch64_gen_unlikely_cbranch): New function.
12428 (aarch64_addti_scratch_regs): New function.
12429 (aarch64_subvti_scratch_regs): New function.
12430 (aarch64_expand_subvti): New function.
12432 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
12434 * config/aarch64/aarch64-option-extensions.def: New entry for profile
12436 * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
12437 * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
12440 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
12443 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12444 address check not strict.
12446 2018-07-19 Andre Vieira <andre.simoesdiasvieira@arm.com>
12448 * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
12450 (store_pair_lanes<mode>): Likewise.
12451 * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
12452 enum value 'ADDR_QUERY_LDP_STP_N'.
12453 * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
12454 (aarch64_print_address_internal): Add declaration.
12455 (aarch64_print_ldpstp_address): Remove.
12456 (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
12457 (aarch64_print_operand): Change printing of 'y'.
12458 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
12459 new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
12460 'true' rather than '1'.
12461 * config/aarch64/constraints.md (Uml): Likewise.
12462 (Uml): Rename to Umn.
12465 2018-07-19 Richard Biener <rguenther@suse.de>
12467 * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
12469 * tree-ssa-sccvn.c: Remove alloc-pool.h use.
12470 (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
12471 (vn_reference_hasher): Likewise.
12472 (struct vn_tables_s): Remove obstack and alloc-pool members.
12473 (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
12474 (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
12475 (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
12476 (vn_reference_insert_pieces): Likewise.
12477 (alloc_vn_nary_op_noinit): Adjust.
12478 (vn_nary_op_insert_stmt): Allocate phiargs in-place.
12479 (vn_phi_eq): Adjust.
12480 (shared_lookup_phiargs): Remove.
12481 (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
12482 (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
12483 (visit_reference_op_call): Likewise.
12484 (copy_nary, copy_phi, copy_reference): Remove.
12485 (process_scc): Rewind the obstack when iterating. Do not
12486 copy the elements to valid_info but just move them from one
12487 hashtable to the other.
12488 (allocate_vn_table): Adjust.
12489 (free_vn_table): Likewise.
12490 (init_scc_vn): Likewise.
12491 (free_scc_vn): Likewise.
12493 2018-07-19 H.J. Lu <hongjiu.lu@intel.com>
12496 * config/i386/i386.c (rest_of_insert_endbranch): Lookup
12497 indirect_return as function type attribute.
12498 (ix86_attribute_table): Change indirect_return to function
12500 * doc/extend.texi: Update indirect_return attribute.
12502 2018-07-19 Aldy Hernandez <aldyh@redhat.com>
12504 * wide-int.h (widest2_int): New.
12505 * gimple-fold.c (arith_overflowed_p): Use it.
12506 * tree.h (widest2_int_cst): New.
12507 * tree-vrp.c (wide_int_binop_overflow): Rename from
12508 vrp_int_const_binop.
12509 Rewrite to work on trees.
12510 (extract_range_from_multiplicative_op_1): Abstract code to...
12511 (wide_int_range_min_max): ...here.
12512 (wide_int_range_cross_product): ...and here.
12513 (extract_range_from_binary_expr_1): Abstract overflow code to...
12514 (wide_int_range_mult_wrapping): ...here.
12515 * tree-vrp.h (wide_int_range_cross_product): New.
12516 (wide_int_range_mult_wrapping): New.
12518 2018-07-19 Andrew Senkevich <andrew.senkevich@intel.com>
12519 Julia Koval <julia.koval@intel.com>
12521 * config/i386/x86-tune-costs.h (skylake_memcpy,
12522 skylake_memset): Replace rep_prefix with unrolling for size 512.
12524 2018-07-18 Kugan Vivekanandarajah <kuganv@linaro.org>
12526 PR middle-end/86544
12527 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
12528 comparision with EQ_EXPR in last stmt.
12530 2018-07-18 Kelvin Nilsen <kelvin@gcc.gnu.org>
12532 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
12533 this subsection to "PowerPC AltiVec/VSX Built-in Functions".
12534 (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
12535 previously known as "PowerPC AltiVec Built-in Functions". Move
12536 some material to new subsubsections "PowerPC AltiVec Built-in
12537 Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
12539 (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
12540 (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
12541 (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
12542 (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
12544 2018-07-18 Richard Biener <rguenther@suse.de>
12546 PR tree-optimization/86557
12547 * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
12550 2018-07-18 Ilya Leoshkevich <iii@linux.ibm.com>
12552 * config/s390/s390.c (s390_function_profiler): Generate CFI.
12554 2018-07-17 Jeff Law <law@redhat.com>
12556 * config/arm/arm.c (get_label_padding): Update for recent
12557 changes to label_to_alignment.
12559 PR tree-optimization/86010
12560 * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
12562 * config/mips/mips.c (vr4130_align_insns): Update for recent
12563 changes to label_to_alignment.
12565 * config/frv/frv.c (frv_label_align): Update for recent changes
12566 to label_to_alignment.
12568 * config/nios2/nios2.c (nios2_label_align): Update for recent
12569 changes which dropped ALIGN_LABELS_LOG.
12571 2018-07-17 Andreas Schwab <schwab@linux-m68k.org>
12573 * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
12574 (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
12576 2018-07-17 Claudiu Zissulescu <claziss@synopsys.com>
12578 * config/arc/arc.c (arc_label_align): Use align_labels instead of
12579 deprecated align_labels_log.
12581 2018-07-17 Richard Biener <rguenther@suse.de>
12584 * dwarf2out.c (init_sections_and_labels): Always generate
12585 a debug_line_str_section for early LTO debug.
12586 (dwarf2out_finish): Reset debug_line_str_hash output early.
12587 Bump counter for extra dwarf5 .debug_loc labels to not conflict
12589 (dwarf2out_early_finish): Output debug_line_str.
12591 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
12593 * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
12594 index register on z196 or later.
12596 2018-07-17 Robin Dapp <rdapp@linux.ibm.com>
12598 * config/s390/s390.c (s390_default_align): Set default function
12600 (s390_override_options_after_change): Call s390_default align.
12601 (s390_option_override_internal): Call s390_default_align.
12602 (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
12604 2018-07-17 Jakub Jelinek <jakub@redhat.com>
12606 PR middle-end/86542
12607 * omp-low.c (create_task_copyfn): Copy over also fields corresponding
12608 to _looptemp_ clauses, other than the first two.
12610 2018-07-17 Martin Liska <mliska@suse.cz>
12612 * opts.c: Do not enable OPT_falign_* for -Os.
12614 2018-07-17 Martin Liska <mliska@suse.cz>
12616 * align.h (MAX_CODE_ALIGN): New.
12617 (MAX_CODE_ALIGN_VALUE): New.
12618 * common/config/i386/i386-common.c (ix86_handle_option):
12619 (MAX_CODE_ALIGN): Moved to align.h.
12620 * final.c (MAX_CODE_ALIGN): Likewise.
12621 * opts.c (parse_and_check_align_values):
12622 (MAX_CODE_ALIGN): Likewise.
12623 (MAX_CODE_ALIGN_VALUE): Likewise.
12625 2018-07-17 Martin Liska <mliska@suse.cz>
12627 * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
12628 in order to fulfil coding style.
12629 * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
12630 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12631 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12632 * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
12633 * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
12634 * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
12635 * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
12636 * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
12637 (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12639 2018-07-17 Martin Liska <mliska@suse.cz>
12641 * align.h: New file.
12642 * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
12644 * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
12645 align_flags of label_to_alignment.
12646 * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
12648 * config/m68k/m68k.c: Do not use removed align_labels_value and
12650 * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
12651 (LOOP_ALIGN): Likewise.
12652 (LABEL_ALIGN): Likewise.
12653 * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12654 Remove not used macro.
12655 (rs6000_loop_align): Change return type to align_flags.
12656 (rs6000_loop_align_max_skip): Remove.
12657 * config/rs6000/rs6000-protos.h (rs6000_loop_align):
12658 Change return type to align_flags.
12659 * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12660 Remove not used macro.
12661 (rs6000_loop_align): Change return type to align_flags.
12662 (rs6000_loop_align_max_skip): Remove.
12663 * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
12664 * config/rx/rx-protos.h (rx_align_for_label): Make it
12666 * config/rx/rx.c (rx_align_for_label): Change return type
12668 (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
12670 into align_flags class.
12671 (LABEL_ALIGN): Likewise.
12672 (LOOP_ALIGN): Likewise.
12673 * config/s390/s390.c (s390_label_align): Use align_flags
12675 (s390_asm_output_function_label): Likewise.
12676 * config/sh/sh.c (sh_override_options_after_change):
12677 Use align_flags class directly without macros.
12678 (find_barrier): Likewise.
12679 (barrier_align): Likewise.
12680 (sh_loop_align): Likewise.
12681 * config/spu/spu.c (spu_option_override):
12682 Use align_flags_tuple::get_value instead of removed macros.
12683 (spu_sched_init): Likewise.
12684 * config/spu/spu.h (GTY): Likewise.
12685 * config/visium/visium.c (visium_option_override):
12686 Set "8" as default secondary alignment.
12687 * config/visium/visium.h (SUBALIGN_LOG): Define to 3
12688 in order to guarantee secondary alignment of 8.
12689 * coretypes.h: Include align.h header file.
12690 * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
12691 TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
12692 and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
12693 * doc/tm.texi.in: Likewise.
12694 * final.c (struct label_alignment): Remove not used structure.
12695 (LABEL_ALIGN): Change type to align_flags.
12696 (LOOP_ALIGN): Likewise.
12697 (JUMP_ALIGN): Likewise.
12698 (default_loop_align_max_skip): Remove.
12699 (default_label_align_max_skip): Likewise.
12700 (default_jump_align_max_skip): Likewise.
12701 (default_label_align_after_barrier_max_skip):
12702 (LABEL_TO_ALIGNMENT): Change to access label_align vector.
12703 (LABEL_TO_MAX_SKIP): Remove.
12704 (label_to_alignment): Return align_flags type instead of integer.
12705 (label_to_max_skip): Remove.
12706 (align_fuzz): Use align_flags type.
12707 (compute_alignments): Use align_flags type and use align_flags::max
12708 to combine multiple alignments.
12709 (grow_label_align): Grow vec instead of C array.
12710 (update_alignments): Assign just LABEL_TO_ALIGNMENT.
12711 (shorten_branches): Use align_flags type and use align_flags::max
12712 to combine multiple alignments.
12713 (final_scan_insn_1): Remove usage of secondary alignment that comes
12714 from label alignment, but instead use proper secondary alignment
12715 which is computed in grow_label_align.
12716 * flags.h (struct align_flags_tuple): Move to align.h.
12717 (struct align_flags): Likewise.
12718 (state_align_loops): Rename to align_loops.
12719 (state_align_jumps): Rename to align_jumps.
12720 (state_align_labels): Rename to align_labels.
12721 (state_align_functions): Rename to align_functions.
12722 (align_loops_log): Remove.
12723 (align_jumps_log): Remove.
12724 (align_labels_log): Remove.
12725 (align_functions_log): Remove.
12726 (align_loops_max_skip): Remove.
12727 (align_jumps_max_skip): Remove.
12728 (align_labels_max_skip): Remove.
12729 (align_functions_max_skip): Remove.
12730 (align_loops_value): Remove.
12731 (align_jumps_value): Remove.
12732 (align_labels_value): Remove.
12733 (align_functions_value): Remove.
12734 * output.h (label_to_alignment): Change return type to align_flags.
12735 (label_to_max_skip): Remove.
12736 * target.def: Remove loop_align_max_skip, label_align_max_skip,
12737 jump_align_max_skip macros.
12738 * targhooks.h (default_loop_align_max_skip): Remove.
12739 (default_label_align_max_skip): Likewise.
12740 (default_jump_align_max_skip): Likewise.
12741 (default_label_align_after_barrier_max_skip): Remove.
12742 * toplev.c (read_log_maxskip): Use ::normalize function.
12743 (parse_N_M): Remove not used argument and also call ::normalize.
12744 (parse_alignment_opts): Do not pass unused arguments.
12745 * varasm.c (assemble_start_function): Use directly align_functions
12746 instead of removed macros.
12747 * system.h: Do not poison removed macros.
12749 2018-07-17 Jakub Jelinek <jakub@redhat.com>
12751 PR middle-end/86539
12752 * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
12753 and cond temporaries don't have reference type if iterator has
12754 pointer type. For init use &for_pre_body instead of pre_p if
12755 for_pre_body is non-empty.
12757 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
12759 * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
12760 double-double modes to SFmode directly directly.
12761 (trunc<mode>sf2_fprs): Delete.
12763 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
12765 * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
12766 for conversions between IFmode and the decimal floating point modes.
12767 (init_float128_ieee): Use the correct names for conversions between
12768 KFmode and the decimal floating point modes.
12770 2018-07-16 Segher Boessenkool <segher@kernel.crashing.org>
12772 * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
12773 for the conversions between TDmode and IFmode.
12774 (init_float128_ieee): Use more correct names for the conversions
12775 between TDmode and KFmode.
12777 2018-07-16 Jakub Jelinek <jakub@redhat.com>
12779 PR tree-optimization/86526
12780 * builtins.c (expand_builtin_memcmp): Formatting fixes.
12781 (inline_expand_builtin_string_cmp): Likewise.
12782 (inline_string_cmp): Likewise. Use c_readstr instead of
12783 builtin_memcpy_read_str. Add unit_mode temporary.
12785 2018-07-16 Bernd Edlinger <bernd.edlinger@hotmail.de>
12787 PR middle-end/86528
12788 * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
12789 * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
12791 2018-07-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12793 * doc/extend.texi (PowerPC AltiVec Built-in Functions):
12794 Alphabetize prototypes of built-in functions, separating out
12795 built-in functions that are listed in this section but should be
12796 described elsewhere.
12798 2018-07-16 Uros Bizjak <ubizjak@gmail.com>
12801 * expmed.c (emit_store_flag): Do not emit setcc followed by a
12802 conditional move when trapping comparison was split to a
12803 non-trapping one (and vice versa).
12805 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
12807 * config/s390/s390.c (s390_function_profiler): Generate nops
12808 instead of profiler call sequences.
12809 * config/s390/s390.opt: Add the new option.
12811 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
12813 * config/s390/s390.c (s390_function_profiler): Generate
12814 __mcount_loc section.
12815 * config/s390/s390.opt: Add the new option.
12817 2018-07-16 Ilya Leoshkevich <iii@linux.ibm.com>
12819 * common.opt: Add the new warning.
12820 * config/s390/s390.c (s390_function_profiler): Emit "brasl
12821 %r0,__fentry__" when -mfentry is specified.
12822 (s390_option_override_internal): Disallow -mfentry for 31-bit
12824 * config/s390/s390.opt: Add the new option.
12826 2018-07-16 Richard Biener <rguenther@suse.de>
12829 * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
12830 for function-local FUNCTION_DECL and RESULT_DECL immediately.
12832 2018-07-16 Martin Liska <mliska@suse.cz>
12835 * ipa-pure-const.c (malloc_candidate_p): Revert ::get
12838 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
12840 * config/arc/arcHS.md: Update ARCHS scheduling rules.
12842 2017-07-16 Claudiu Zissulescu <claziss@synopsys.com>
12844 * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
12846 * config/arc/arc-cpus.def (hs4x): New cpu.
12848 * config/arc/arc-tables.opt: Regenerate.
12849 * config/arc/arc.c (arc_sched_issue_rate): New function.
12850 (TARGET_SCHED_ISSUE_RATE): Define.
12851 (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
12852 * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
12854 (attr tune): Add ARCHS4x tune values.
12855 (attr tune_dspmpy): Define.
12856 (*tst): Correct instruction type.
12857 * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
12858 * config/arc/arcHS4x.md: New file.
12859 * config/arc/fpu.md: Update instruction type attributes.
12860 * config/arc/t-multilib: Regenerate.
12862 2018-07-16 Tom de Vries <tdevries@suse.de>
12865 * var-tracking.c (vt_initialize): Fix pre_dec handling.
12867 2018-07-16 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
12869 * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
12872 2018-07-16 Eric Botcazou <ebotcazou@adacore.com>
12874 PR tree-optimization/86514
12875 * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
12876 conversion to a boolean type from a type with greater precision.
12878 2018-07-16 Tom de Vries <tdevries@suse.de>
12880 * var-tracking.c (vt_initialize): Print adjusted insn slim if
12881 dump_flags request TDF_SLIM.
12883 2018-07-16 Aldy Hernandez <aldyh@redhat.com>
12885 * fold-const.c (int_const_binop_1): Abstract...
12886 (wide_int_binop): ...wide int code here.
12887 (poly_int_binop): ...poly int code here.
12888 Abstract the rest of int_const_binop_1 into int_const_binop.
12889 * fold-const.h (wide_int_binop): New.
12890 * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
12891 Remove useless PLUS/MINUS_EXPR case.
12892 (zero_nonzero_bits_from_vr): Move wide int code...
12893 (zero_nonzero_bits_from_bounds): ...here.
12894 (extract_range_from_binary_expr_1): Move mask optimization code...
12895 (range_easy_mask_min_max): ...here.
12896 * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
12897 (range_easy_mask_min_max): New.
12899 2018-07-15 Jeff Law <law@redhat.com>
12902 * config/sh/sh.c (output_mi_thunk): Remove dead conditional
12905 2018-07-14 Jim Wilson <jimw@sifive.com>
12907 * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
12909 2018-07-14 Paul Koning <ni1d@arrl.net>
12911 * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
12913 2018-07-13 Jan Hubicka <hubicka@ucw.cz>
12915 * lto-streamer-out.c (copy_function_or_variable): Dump info about
12918 2018-07-13 Bill Schmidt <wschmidt@linux.ibm.com>
12919 Steve Munroe <munroesj52@gmail.com>
12921 * config/rs6000/emmintrin.h (_mm_and_si128): New function.
12922 (_mm_andnot_si128): Likewise.
12923 (_mm_or_si128): Likewise.
12924 (_mm_xor_si128): Likewise.
12926 2018-07-13 Qing Zhao <qing.zhao@oracle.com>
12928 PR middle-end/78809
12929 * builtins.c (expand_builtin_memcmp): Inline the calls first
12930 when result_eq is false.
12931 (expand_builtin_strcmp): Inline the calls first.
12932 (expand_builtin_strncmp): Likewise.
12933 (inline_string_cmp): New routine. Expand a string compare
12934 call by using a sequence of char comparison.
12935 (inline_expand_builtin_string_cmp): New routine. Inline expansion
12936 a call to str(n)cmp/memcmp.
12937 * doc/invoke.texi (--param builtin-string-cmp-inline-length):
12939 * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
12941 2018-07-13 Richard Earnshaw <rearnsha@arm.com>
12943 * config/arm/driver-arm.c: Include arm-native.h.
12944 (host_detect_local_cpu): Use auto-generated data tables.
12945 (vendors, arm_cpu_table): Delete. Move part information to ...
12946 * config/arm/arm-cpus.in: ... here.
12947 * config/arm/parsecpu.awk (gen_native): New function.
12948 (vendor, part): New CPU fields.
12949 (END): Add support for building the native CPU detection tables.
12950 * config/arm/t-arm (arm-native.h): Add build rule.
12951 (driver-arm.o): Add dependency on arm-native.h.
12953 2018-07-13 Richard Biener <rguenther@suse.de>
12955 PR middle-end/85974
12956 * match.pd (addr1 - addr2): Allow either of the operand to
12959 2018-07-13 Tom de Vries <tdevries@suse.de>
12961 * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
12964 2018-07-13 Jackson Woodruff <jackson.woodruff@arm.com>
12966 * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
12967 arrays instead of numbered variables.
12969 2018-07-13 Eric Botcazou <ebotcazou@adacore.com>
12971 * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
12972 * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
12974 2018-07-13 Richard Biener <rguenther@suse.de>
12977 * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
12978 instead of get_context_die.
12980 2018-07-13 Kugan Vivekanandarajah <kuganv@linaro.org>
12981 Richard Biener <rguenther@suse.de>
12983 PR middle-end/86489
12984 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
12985 that the loop latch destination where phi is defined.
12987 2018-07-12 Kito Cheng <kito.cheng@gmail.com>
12989 * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
12990 (riscv_expand_epilogue): Add assertion to check interrupt mode.
12991 (riscv_set_current_function): Extract getting interrupt type to new
12993 (riscv_get_interrupt_type): New function.
12994 (riscv_merge_decl_attributes): New function, checking interrupt type is
12996 (TARGET_MERGE_DECL_ATTRIBUTES): Define.
12998 2018-07-12 Paul Koning <ni1d@arrl.net>
13000 * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
13003 2018-07-12 Paul Koning <ni1d@arrl.net>
13005 * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
13006 zero reference, add doloop_end instead.
13007 * doc/md.texi (decrement_and_branch_until_zero): Remove.
13008 (Looping patterns): Remove decrement_and_branch_until_zero. Add
13009 detail for doloop_end.
13011 2018-07-12 Martin Sebor <msebor@redhat.com>
13014 * attribs.c (decl_attributes): Reject conflicting attributes before
13015 calling attribute handlers.
13017 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
13019 * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
13021 (gcc::dump_manager::get_dump_file_name): likewise.
13022 (dump_begin): Likewise.
13023 * dumpfile.h (dump_begin): Update prototype.
13024 (gcc::dump_manager::get_dump_file_name,
13025 gcc::dump_manager::get_dump_file_name): Update prototype.
13027 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
13029 * internal-fn.h (vectorizable_internal_fn_p): New function.
13030 * tree-vect-slp.c (compatible_calls_p): Likewise.
13031 (vect_build_slp_tree_1): Remove nops argument. Handle calls
13032 to internal functions.
13033 (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
13035 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
13037 * fold-const.h (inverse_conditions_p): Declare.
13038 * fold-const.c (inverse_conditions_p): New function.
13039 * match.pd: Use inverse_conditions_p. Add folds of view_converts
13040 that test the inverse condition of a conditional internal function.
13041 * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
13042 * internal-fn.c (internal_fn_mask_index): Handle conditional
13043 internal functions.
13044 (vectorized_internal_fn_supported_p): New function.
13045 * tree-if-conv.c: Include internal-fn.h and fold-const.h.
13046 (any_pred_load_store): Replace with...
13047 (need_to_predicate): ...this new variable.
13048 (redundant_ssa_names): New variable.
13049 (ifcvt_can_use_mask_load_store): Move initial checks to...
13050 (ifcvt_can_predicate): ...this new function. Handle tree codes
13051 for which a conditional internal function exists.
13052 (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
13053 instead of ifcvt_can_use_mask_load_store. Update after variable
13055 (predicate_load_or_store): New function, split out from
13056 predicate_mem_writes.
13057 (check_redundant_cond_expr): New function.
13058 (value_available_p): Likewise.
13059 (predicate_rhs_code): Likewise.
13060 (predicate_mem_writes): Rename to...
13061 (predicate_statements): ...this. Use predicate_load_or_store
13062 and predicate_rhs_code.
13063 (combine_blocks, tree_if_conversion): Update after above name changes.
13064 (ifcvt_local_dce): Handle redundant_ssa_names.
13065 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
13066 general conditional functions.
13067 * tree-vect-stmts.c (vectorizable_call): Likewise.
13069 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
13070 Alan Hayward <alan.hayward@arm.com>
13071 David Sherwood <david.sherwood@arm.com>
13073 * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
13074 * internal-fn.c (can_interpret_as_conditional_op_p): New function.
13075 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
13076 plus and minus and convert them into IFN_COND_FMA-based sequences.
13077 (convert_mult_to_fma): Handle conditional plus and minus.
13079 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
13081 * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
13082 * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
13083 (cond_fnms_optab): New optabs.
13084 * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
13085 internal functions.
13086 (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
13087 * internal-fn.h (get_conditional_internal_fn): Declare.
13088 (get_unconditional_internal_fn): Likewise.
13089 * internal-fn.c (cond_ternary_direct): New macro.
13090 (expand_cond_ternary_optab_fn): Likewise.
13091 (direct_cond_ternary_optab_supported_p): Likewise.
13092 (FOR_EACH_COND_FN_PAIR): Likewise.
13093 (get_conditional_internal_fn): New function.
13094 (get_unconditional_internal_fn): Likewise.
13095 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
13096 (gimple_match_op::gimple_match_op): Add a new overload for 5
13098 (gimple_match_op::set_op): Likewise.
13099 (gimple_resimplify5): Declare.
13100 * genmatch.c (decision_tree::gen): Generate simplifications for
13102 * gimple-match-head.c (gimple_simplify): Define an overload for
13103 5 operands. Handle calls with 5 arguments in the top-level overload.
13104 (convert_conditional_op): Handle conversions from unconditional
13105 internal functions to conditional ones.
13106 (gimple_resimplify5): New function.
13107 (build_call_internal): Pass a fifth operand.
13108 (maybe_push_res_to_seq): Likewise.
13109 (try_conditional_simplification): Try converting conditional
13110 internal functions to unconditional internal functions.
13111 Handle 3-operand unconditional forms.
13112 * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
13113 Define ternary equivalents of the current rules for binary conditional
13114 internal functions.
13115 * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
13116 ternary operations.
13117 * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
13118 (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
13119 (optab): Handle them.
13120 (SVE_COND_FP_TERNARY): New int iterator.
13121 (sve_fmla_op, sve_fmad_op): New int attributes.
13122 * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
13123 (*cond_<optab><mode>_2, *cond_<optab><mode_4)
13124 (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
13126 2018-07-12 Richard Sandiford <richard.sandiford@linaro.org>
13128 * target.def (preferred_else_value): New target hook.
13129 * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
13130 * doc/tm.texi: Regenerate.
13131 * targhooks.h (default_preferred_else_value): Declare.
13132 * targhooks.c (default_preferred_else_value): New function.
13133 * internal-fn.h (conditional_internal_fn_code): Declare.
13134 * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
13135 (get_conditional_internal_fn): Use it.
13136 (conditional_internal_fn_code): New function.
13137 * gimple-match.h (gimple_match_cond): New struct.
13138 (gimple_match_op): Add a cond member function.
13139 (gimple_match_op::gimple_match_op): Update all forms to take a
13141 * genmatch.c (expr::gen_transform): Use the same condition as res_op
13142 for the suboperation, but don't specify a particular else_value.
13143 * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
13144 (visit_nary_op, visit_reference_op_load): Pass
13145 gimple_match_cond::UNCOND to the gimple_match_op constructor.
13146 * gimple-match-head.c: Include tree-eh.h
13147 (convert_conditional_op): New function.
13148 (maybe_resimplify_conditional_op): Likewise.
13149 (gimple_resimplify1): Call maybe_resimplify_conditional_op.
13150 (gimple_resimplify2): Likewise.
13151 (gimple_resimplify3): Likewise.
13152 (gimple_resimplify4): Likewise.
13153 (maybe_push_res_to_seq): Return null for conditional operations.
13154 (try_conditional_simplification): New function.
13155 (gimple_simplify): Call it. Pass conditions to the gimple_match_op
13157 * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
13159 * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
13161 (TARGET_PREFERRED_ELSE_VALUE): Redefine.
13163 2018-07-12 Jan Hubicka <hubicka@ucw.cz>
13165 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
13167 (hash_tree): Do not hash DECL_FCONTEXT
13168 * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
13169 Do not stream DECL_FCONTEXT.
13170 * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
13171 * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
13173 2018-07-12 Richard Biener <rguenther@suse.de>
13176 * dwarf2out.c (gen_block_die): Only output blocks when they have
13177 at least one !DECL_IGNORED_P variable.
13179 2018-07-12 Richard Biener <rguenther@suse.de>
13182 * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
13183 Remove -mieee-fp handling.
13185 2018-07-12 Richard Biener <rguenther@suse.de>
13187 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
13188 left-over from last patch.
13190 2018-07-12 Jakub Jelinek <jakub@redhat.com>
13192 PR tree-optimization/86492
13193 * gimple-ssa-store-merging.c
13194 (imm_store_chain_info::coalesce_immediate_stores): Call
13195 check_no_overlap even for the merge_overlapping case. Formatting fix.
13197 2018-07-12 Richard Biener <rguenther@suse.de>
13199 PR middle-end/86479
13200 * fold-const.c (fold_binary_op_with_conditional_arg): Do not
13201 move possibly trapping operations into the conditional.
13203 2018-07-12 Richard Biener <rguenther@suse.de>
13205 * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
13206 (vn_lookup_simplify_result): Remove recursion limit applied
13208 (vn_nary_build_or_lookup_1): Adjust.
13209 (try_to_simplify): Likewise.
13210 * gimple-match-head.c (gimple_resimplify1): Instead apply one
13212 (gimple_resimplify2): Likewise.
13213 (gimple_resimplify3): Likewise.
13214 (gimple_resimplify4): Likewise.
13216 2018-07-11 Jakub Jelinek <jakub@redhat.com>
13218 * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
13219 Use __mmask64 type instead of __mmask8 for __M argument.
13220 * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
13221 _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
13223 (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
13224 __mmask16 for __M argument.
13225 (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
13226 _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
13227 to __mmask16 instead of __mmask8.
13228 * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
13229 _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
13230 _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
13231 _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
13232 instead of __mmask16 for __U argument.
13233 * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
13234 __mmask16 instead of __mmask8 for __U argument.
13235 (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
13237 (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
13239 (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
13241 (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
13243 (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
13245 (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
13247 (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
13249 (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
13251 (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
13253 (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
13255 (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
13256 _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
13257 return type as well as __M argument type and all casts from __mmask8
13259 (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
13260 _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
13261 return type as well as __M argument type and all casts from __mmask8
13263 (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
13264 _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
13265 return type as well as __M argument type and all casts from __mmask8
13267 (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
13268 _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
13269 return type as well as __M argument type and all casts from __mmask8
13271 * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
13272 _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
13275 2018-07-11 Grazvydas Ignotas <notasas@gmail.com>
13277 * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
13278 _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
13281 2018-07-11 Paul Koning <ni1d@arrl.net>
13283 * doc/md.texi (define_subst): Document how multiple occurrences of
13284 the same argument in the replacement pattern are handled.
13286 2018-07-11 Paul Koning <ni1d@arrl.net>
13288 * doc/extend.texi (Common Variable Attributes): Move "mode" into
13289 alphabetical order.
13290 (Common Type Attributes): Add "mode" attribute.
13292 2018-07-11 Jan Hubicka <hubicka@ucw.cz>
13294 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
13295 stream DECL_ORIGINAL_TYPE.
13296 (DFS::DFS_write_tree_body): Drop hack handling local external decls.
13297 (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
13298 * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
13299 Do not walk original type.
13300 * tree-streamer-out.c (streamer_write_chain): Drop hack handling
13302 (write_ts_decl_non_common_tree_pointers): Do not stream
13304 * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
13305 (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
13307 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
13309 * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
13310 thread twice from the same starting edge.
13312 2018-07-11 Aldy Hernandez <aldyh@redhat.com>
13314 * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
13315 * gimple.c (gimple_call_nonnull_result_p): ...here...
13316 (gimple_call_nonnull_arg): ...and here.
13317 * gimple.h (gimple_call_nonnull_result_p): New.
13318 (gimple_call_nonnull_arg): New.
13320 2018-07-11 Richard Earnshaw <rearnsha@arm.com>
13322 * config/arm/arm-cpus.in: Move information from fpu field of each
13323 cpu definition to the isa field.
13324 * config/arm/parsecpu.awk (fpu): Delete match rule.
13325 (gen_comm_data): Don't add bits from the CPU's FPU entry.
13327 2018-07-11 Richard Biener <rguenther@suse.de>
13330 * dwarf2out.c (init_sections_and_labels): Use
13331 output_asm_line_debug_info consistently.
13332 (dwarf2out_early_finish): Likewise.
13333 (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
13336 2018-07-11 Richard Biener <rguenther@suse.de>
13338 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
13339 Rework father_bb setting in a way to avoid propagating constants
13340 multiple times on a loop body.
13342 2018-07-10 Mark Wielaard <mark@klomp.org>
13345 * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
13347 2018-07-10 Richard Biener <rguenther@suse.de>
13349 * hash-map.h (hash_map::iterator::operator*): Return
13350 references to key and value.
13352 2018-07-10 Jakub Jelinek <jakub@redhat.com>
13355 * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
13356 to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
13357 BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
13358 (gimplify_omp_for): For composite loops, move outer
13359 OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
13360 OMP_FOR/OMP_SIMD if there are any non-trivial wrappers. For class
13361 iterators add any needed clauses. Allow OMP_FOR_ORIG_DECLS to contain
13362 TREE_LIST for both the original class iterator and the "last" helper
13363 var. Gimplify OMP_FOR_PRE_BODY before the outermost composite
13364 loop, remember has_decl_expr from outer composite loops for the
13365 innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
13367 2018-07-09 Martin Sebor <msebor@redhat.com>
13369 PR middle-end/77357
13370 PR middle-end/86428
13371 * builtins.c (c_strlen): Avoid out-of-bounds warnings when
13372 accessing implicitly initialized array elements.
13373 * expr.c (string_constant): Handle string initializers of
13374 character arrays within aggregates.
13375 * gimple-fold.c (fold_array_ctor_reference): Add argument.
13376 Store element offset. As a special case, handle zero size.
13377 (fold_nonarray_ctor_reference): Same.
13378 (fold_ctor_reference): Add argument. Store subobject offset.
13379 * gimple-fold.h (fold_ctor_reference): Add argument.
13381 2018-07-09 Paul Koning <ni1d@arrl.net>
13383 * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
13384 (pdp11_insn_cost): New function.
13385 (pdp11_md_asm_adjust): New function.
13386 (TARGET_INVALID_WITHIN_DOLOOP): Define.
13387 (pdp11_rtx_costs): Update to match machine better.
13388 (output_addr_const_pdp11): Correct format mismatch warnings.
13389 * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
13390 * config/pdp11/pdp11.md: General change to add base_cost and/or
13391 length attributes for use by new pdp11_insn_cost function.
13392 (MIN_BRANCH): Correct definition.
13394 (doloop_end): Use standard pattern name for looping pattern.
13395 (doloop_end_nocc): New.
13396 (movsf): Add another constraint alternative.
13397 (zero_extendqihi2): Add constraint alternatives for not in place
13399 (zero_extendhisi2): Remove.
13400 (shift patterns): Add CC handling variants.
13404 (define_peephole2): New peephole to recognize mov that sets CC for
13407 2018-07-09 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13409 * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
13410 SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
13413 2018-07-09 Jakub Jelinek <jakub@redhat.com>
13416 * real.c (real_nextafter): Return true if result is denormal.
13418 2018-07-09 Martin Liska <mliska@suse.cz>
13420 * common.opt: Add back wrongly removed attribute.
13422 2018-07-09 Richard Biener <rguenther@suse.de>
13425 * dwarf2out.c (gen_block_die): For an early generated DIE
13426 always output high/low PC attributes.
13428 2018-07-09 Tom de Vries <tdevries@suse.de>
13430 * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
13431 * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
13432 onto VAR_DECL with abstract origin.
13434 2018-07-07 Jim Wilson <jimw@sifive.com>
13436 * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
13438 2018-07-07 Tom de Vries <tdevries@suse.de>
13440 * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
13442 2018-07-07 Aldy Hernandez <aldyh@redhat.com>
13444 * tree-vrp.c (vrp_int_const_binop): Change overflow type to
13446 (combine_bound): Use wide-int overflow calculation instead of
13448 * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
13450 * fold-const.c (int_const_binop_2): Same.
13451 (extract_muldiv_1): Same.
13452 (fold_div_compare): Same.
13453 (fold_abs_const): Same.
13455 * poly-int.h (add): Same.
13459 * predict.c (predict_iv_comparison): Same.
13460 * profile-count.c (slow_safe_scale_64bit): Same.
13461 * simplify-rtx.c (simplify_const_binary_operation): Same.
13462 * tree-chrec.c (tree_fold_binomial): Same.
13463 * tree-data-ref.c (split_constant_offset_1): Same.
13464 * tree-if-conv.c (idx_within_array_bound): Same.
13465 * tree-scalar-evolution.c (iv_can_overflow_p): Same.
13466 * tree-ssa-phiopt.c (minmax_replacement): Same.
13467 * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
13468 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
13469 * vr-values.c (vr_values::adjust_range_with_scev): Same.
13470 * wide-int.cc (wi::add_large): Same.
13471 (wi::mul_internal): Same.
13472 (wi::sub_large): Same.
13473 (wi::divmod_internal): Same.
13474 * wide-int.h: Change overflow type to overflow_type for neg, add,
13475 mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
13476 mod_trunc, mod_ceil, mod_round, add_large, sub_large,
13477 mul_internal, divmod_internal.
13478 (overflow_type): New enum.
13479 (accumulate_overflow): New.
13481 2018-07-06 Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
13483 * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
13484 (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
13486 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
13488 * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
13489 argument is checked for zero before entering loop, avoid checking again.
13491 2018-07-06 Kugan Vivekanandarajah <kuganv@linaro.org>
13493 * gimplify.h (generic_expr_could_trap_p): Set as global function.
13494 * gimplify.h (generic_expr_could_trap_p): Likwise.
13495 * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
13497 2018-07-06 Jakub Jelinek <jakub@redhat.com>
13499 PR tree-optimization/86401
13500 * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
13501 ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
13502 (fold_bit_and_mask): ... here. New helper function for match.pd.
13503 * fold-const.h (fold_bit_and_mask): Declare.
13504 * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
13506 2018-07-06 Peter Bergner <bergner@linux.ibm.com>
13509 * target.def (translate_mode_attribute): New hook.
13510 * targhooks.h (default_translate_mode_attribute): Declare.
13511 * targhooks.c (default_translate_mode_attribute): New function.
13512 * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
13513 * doc/tm.texi: Regenerate.
13514 * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
13515 (rs6000_translate_mode_attribute): New function.
13517 2018-07-06 Paul Koning <ni1d@arrl.net>
13519 * doc/md.texi (define_split): Document DONE and FAIL.
13520 (define_peephole2): Ditto.
13522 2018-07-05 Jeff Law <law@redhat.com>
13524 PR tree-optimization/86010
13525 * tree-ssa-dse.c (compute_trims): More aggressively trim at
13526 both the head and tail of mem* and str* calls.
13528 2018-07-05 Jim Wilson <jimw@sifive.com>
13530 * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
13532 2018-07-05 Indu Bhagat <indu.bhagat@oracle.com>
13534 * config/aarch64/aarch64-simd.md: correct flags text for
13535 MIN_EXPR replacement.
13537 2018-07-05 James Clarke <jrtc27@jrtc27.com>
13539 * configure: Regenerated.
13541 2018-07-05 Carl Love <cel@us.ibm.com>
13543 * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
13544 float argument to VSX_BUILTIN_DOUBLEH_V4SF.
13545 Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
13546 VSX_BUILTIN_DOUBLEL_V4SF.
13548 2018-07-05 Martin Sebor <msebor@redhat.com>
13551 * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
13552 than its domain to compute its the upper bound of a char array.
13554 2018-07-05 Nathan Sidwell <nathan@acm.org>
13556 Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
13557 * doc/cpp.texi: Update comment.
13558 * doc/tm.texi: Rebuilt.
13559 * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
13560 (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
13561 * doc/extend.texi (Backwards Compatibility): Clarify it is system
13562 headers affected by extern "C".
13563 * system.h: Poison NO_IMPLICIT_EXTERN_C.
13564 * config/alpha/alpha.h, config/arm/uclinux-elf.h,
13565 config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
13566 config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
13567 config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
13568 config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
13569 config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
13570 config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
13571 config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
13572 config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
13573 config/sparc/openbsd64.h, config/sparc/sp-elf.h,
13574 config/sparc/sp64-elf.h, config/spu/spu.h,
13575 config/stormy16/stormy16.h, config/v850/v850.h,
13576 config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
13577 define NO_IMPLICIT_EXTERN_C.
13578 * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
13580 2018-07-05 Tamar Christina <tamar.christina@arm.com>
13583 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
13584 instead of GET_MODE_SIZE when comparing Units.
13586 2018-07-05 Tamar Christina <tamar.christina@arm.com>
13589 * rtlanal.c (set_noop_p): Constrain on mode change,
13590 include hard-reg-set.h
13592 2018-07-05 Tamar Christina <tamar.christina@arm.com>
13594 * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
13596 2018-07-05 Jakub Jelinek <jakub@redhat.com>
13599 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
13602 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13604 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13606 2018-07-04 Maxim Ostapenko <m.ostapenko@samsung.com>
13609 * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13611 * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13613 2018-07-04 Eric Botcazou <ebotcazou@adacore.com>
13615 PR middle-end/86380
13616 * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
13618 2018-07-04 Aldy Hernandez <aldyh@redhat.com>
13620 * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
13621 neg_*_op* variables.
13623 2018-07-04 Martin Liska <mliska@suse.cz>
13625 * tree-switch-conversion.c: Define
13626 max_ratio_for_speed and max_ratio_for_size constants.
13628 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
13629 Martin Liska <mliska@suse.cz>
13631 PR middle-end/66240
13634 * common.opt: Rename align options with 'str_' prefix.
13635 * common/config/i386/i386-common.c (set_malign_value): New
13637 (ix86_handle_option): Use it to set -falign-* options/
13638 * config/aarch64/aarch64-protos.h (struct tune_params): Change
13639 type from int to string.
13640 * config/aarch64/aarch64.c: Update default values from int
13642 * config/alpha/alpha.c (alpha_override_options_after_change):
13644 * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
13645 * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13646 max skip conditionally.
13647 * config/i386/freebsd.h (SUBALIGN_LOG): New.
13648 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13649 max skip conditionally.
13650 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13651 max skip conditionally.
13652 * config/i386/gnu-user.h (SUBALIGN_LOG): New.
13653 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13654 max skip conditionally.
13655 * config/i386/i386.c (struct ptt): Change type from int to
13657 (ix86_default_align): Set default values.
13658 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
13659 max skip conditionally.
13660 * config/i386/iamcu.h (SUBALIGN_LOG): New.
13661 (ASM_OUTPUT_MAX_SKIP_ALIGN):
13662 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
13663 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13664 max skip conditionally.
13665 * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
13666 (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
13667 * config/i386/x86-64.h (SUBALIGN_LOG): New.
13668 (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13669 max skip conditionally.
13670 (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13671 * config/ia64/ia64.c (ia64_option_override): Set default values
13672 for alignment options.
13673 * config/m68k/m68k.c: Handle new str_align_* options.
13674 * config/mips/mips.c (mips_set_compression_mode): Change
13676 (mips_option_override): Set default values for options.
13677 * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
13679 * config/rs6000/rs6000.c (rs6000_option_override_internal):
13681 * config/rx/rx.c (rx_option_override): Likewise.
13682 * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
13683 (LABEL_ALIGN): Use align_labels_log.
13684 (LOOP_ALIGN): Use align_loops_align.
13685 * config/s390/s390.c (s390_asm_output_function_label): Use new
13687 * config/sh/sh.c (sh_override_options_after_change):
13688 Change type of constants.
13689 * config/spu/spu.c (spu_sched_init): Likewise.
13690 * config/sparc/sparc.c (sparc_option_override): Set default
13691 values for options.
13692 * config/visium/visium.c (visium_option_override): Likewise.
13693 * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
13694 emit p2align format with last argument if it's not needed.
13695 * doc/invoke.texi: Document extended format of -falign-*.
13696 * final.c: Use align_labels alignment.
13697 * flags.h (struct target_flag_state): Change type to use
13699 (struct align_flags_tuple): New.
13700 (struct align_flags): Likewise.
13701 (align_loops_log): Redefine macro to use new types.
13702 (align_loops_max_skip): Redefine macro to use new types.
13703 (align_jumps_log): Redefine macro to use new types.
13704 (align_jumps_max_skip): Redefine macro to use new types.
13705 (align_labels_log): Redefine macro to use new types.
13706 (align_labels_max_skip): Redefine macro to use new types.
13707 (align_functions_log): Redefine macro to use new types.
13708 (align_loops): Redefine macro to use new types.
13709 (align_jumps): Redefine macro to use new types.
13710 (align_labels): Redefine macro to use new types.
13711 (align_functions): Redefine macro to use new types.
13712 (align_functions_max_skip): Redefine macro to use new types.
13713 (align_loops_value): New macro.
13714 (align_jumps_value): New macro.
13715 (align_labels_value): New macro.
13716 (align_functions_value): New macro.
13717 * function.c (invoke_set_current_function_hook): Propagate
13718 alignment values from flags to global variables default in
13720 * ipa-icf.c (sem_function::equals_wpa): Use
13721 cl_optimization_option_eq instead of memcmp.
13722 * lto-streamer.h (cl_optimization_stream_out): Support streaming
13724 (cl_optimization_stream_in): Likewise.
13725 * optc-save-gen.awk: Support strings in cl_optimization.
13726 * opth-gen.awk: Likewise.
13727 * opts.c (finish_options): Remove error checking of invalid
13729 (MAX_CODE_ALIGN): Remove.
13730 (MAX_CODE_ALIGN_VALUE): Likewise.
13731 (parse_and_check_align_values): New function.
13732 (check_alignment_argument): Likewise.
13733 (common_handle_option): Use check_alignment_argument.
13734 * opts.h (parse_and_check_align_values): Declare.
13735 * toplev.c (init_alignments): Remove.
13736 (read_log_maxskip): New.
13737 (parse_N_M): Likewise.
13738 (parse_alignment_opts): Likewise.
13739 (backend_init_target): Remove usage of init_alignments.
13740 * toplev.h (parse_alignment_opts): Declare.
13741 * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
13743 * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
13744 * tree.c (cl_option_hasher::equal): New.
13745 * varasm.c: Use new global macros.
13747 2018-07-04 Denys Vlasenko <dvlasenk@redhat.com>
13749 * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
13750 Use a simpler align directive also if MAXSKIP = ALIGN-1.
13751 * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13752 * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13753 * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13754 * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13755 * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
13756 is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
13757 define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
13758 * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13759 * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13760 * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13761 * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13763 2018-07-04 Martin Liska <mliska@suse.cz>
13764 Jonathan Wakely <jwakely@redhat.com>
13766 * coverage.c: Use correct type.
13767 * doc/invoke.texi: Language correction.
13769 2018-07-03 H.J. Lu <hongjiu.lu@intel.com>
13772 * config/i386/i386.c (rest_of_insert_endbranch): Also generate
13773 ENDBRANCH for non-tail call which may return via indirect branch.
13774 * doc/extend.texi: Document indirect_return attribute.
13776 2018-07-03 Martin Sebor <msebor@redhat.com>
13778 PR tree-optimization/86274
13779 * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
13781 (format_floating): Correct handling of infinities and NaNs.
13783 2018-07-03 Martin Sebor <msebor@redhat.com>
13785 * print-tree.c (print_real_cst): New function.
13786 (print_node_brief): Call it.
13787 (print_node): Ditto.
13789 2018-07-03 Jeff Law <law@redhat.com>
13791 * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
13792 into a single pattern.
13794 * config/h8300/h8300.md (ors code_iterator): New.
13795 (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
13796 a single pattern and single splitter.
13797 (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
13798 (iorqi3_1, xorqi3_1): Likewise.
13799 (iorqi3, xorqi3 expanders): Similarly.
13801 * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
13802 (movmd_internal) into a single pattern using the P mode iterator.
13803 (movmd splitters): Similarly.
13804 (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
13805 (movsd splitters): Similarly.
13807 * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
13808 ADDB, ADDW and ADDL into a single ADD attribute which selects the
13809 right table based on the size of the operand.
13810 * config/h8300/h8300.md (length_table): Corresponding changes. All
13811 references to "addb", "addw" and "addl" changed to "add".
13812 (btst patterns): Merge two variants into a single pattern.
13813 (tstqi, tsthi): Likewise.
13814 (addhi3_incdec, addsi3_incdec): Likewise.
13815 (subhi3_h8300hs, subsi3_h8300hs): Likewise.
13816 (mulhi3, mulsi3): Likewise.
13817 (udivhi3, udivsi3): Likewise.
13818 (divhi3, divsi3): Likewise.
13819 (andorqi3, andorhi3, andorsi3): Likewise.
13821 2018-07-03 Uros Bizjak <ubizjak@gmail.com>
13824 * config/i386/sse.md (uavg<mode>3_ceil): New expander.
13825 (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
13827 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
13829 PR tree-optimization/85694
13830 * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
13831 (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
13833 * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
13834 (<u>avg<mode>3_ceil): New patterns.
13836 2018-07-03 David Malcolm <dmalcolm@redhat.com>
13838 * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
13839 scan-tree-dump directive.
13840 * gcc.dg/vect/slp-perm-2.c: Likewise.
13841 * gcc.dg/vect/slp-perm-3.c: Likewise.
13842 * gcc.dg/vect/slp-perm-5.c: Likewise.
13843 * gcc.dg/vect/slp-perm-6.c: Likewise.
13844 * gcc.dg/vect/slp-perm-7.c: Likewise.
13845 * gcc.dg/vect/slp-perm-8.c: Likewise.
13847 2018-07-03 Marek Polacek <polacek@redhat.com>
13849 PR middle-end/86202
13850 * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
13852 2018-07-03 Richard Biener <rguenther@suse.de>
13855 * tree-ssa-structalias.c (find_func_clobbers): Properly
13856 handle indirect calls.
13858 2018-07-03 Jeff Law <law@redhat.com>
13860 * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
13861 (shifts): New code iterator.
13862 (movqi, movhi, movsi, movsf expanders): Consolidate into a single
13863 expander. Fix HImode handling on H8/SX.
13864 (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
13865 (subqi3, subhi3, subsi3 expanders): Likewise.
13866 (andqi3, andhi3, andsi3 expanders): Likewise.
13867 (iorqi3, iorhi3, iorsi3 expanders): Likewise.
13868 (xorqi3, xorhi3, xorsi3 expanders): Likewise.
13869 (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
13870 (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
13871 (zero_extendqihi2, zero_extendqisi2): Likewise.
13872 (extendqihi2, extendqisi2): Likewise.
13873 (rotlqi3, rotlhi3, rotlsi3): Likewise.
13874 (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
13875 (rotlqi3_1, rotlhi3_1): Likewise.
13876 (logicalhi3_sn, logicalsi3_sn): Likewise.
13877 (logicalhi3, logicalsi3): Likewise.
13879 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
13881 * tree-vect-patterns.c (vect_recog_rotate_pattern)
13882 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
13883 (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
13884 (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
13885 type to append_pattern_def_seq instead of creating a stmt_vec_info
13887 (build_mask_conversion): Likewise. Remove vinfo argument.
13888 (vect_add_conversion_to_patterm): Likewise, renaming to...
13889 (vect_add_conversion_to_pattern): ...this.
13890 (vect_recog_mask_conversion_pattern): Update call to
13891 build_mask_conversion. Pass the vector type to
13892 append_pattern_def_seq here too.
13893 (vect_recog_gather_scatter_pattern): Update call to
13894 vect_add_conversion_to_pattern.
13896 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
13898 * tree-vect-patterns.c (new_pattern_def_seq): Delete.
13899 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13900 (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
13901 (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
13902 STMT_VINFO_PATTERN_DEF_SEQ to null here.
13903 (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
13904 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
13905 append_pattern_def_seq instead of new_pattern_def_seq.
13906 (vect_recog_divmod_pattern): Do both of the above.
13907 (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
13910 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
13912 * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
13913 (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
13914 (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
13915 (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
13916 (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
13917 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
13918 (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
13919 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
13920 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
13921 (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
13922 parameter with a single stmt_vec_info.
13923 (vect_recog_func_ptr): Likewise.
13924 (vect_recog_gather_scatter_pattern): Likewise, folding in...
13925 (vect_try_gather_scatter_pattern): ...this.
13926 (vect_pattern_recog_1): Remove stmts_to_replace and just pass
13927 the stmt_vec_info of the statement to be matched. Don't clear
13928 STMT_VINFO_RELATED_STMT.
13929 (vect_pattern_recog): Update call accordingly.
13931 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
13933 PR tree-optimization/85694
13934 * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
13935 (uavgM3_ceil): Document new optabs.
13936 * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
13937 * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
13939 * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
13940 (savg_ceil_optab): New optabs.
13941 * tree-vect-patterns.c (vect_recog_average_pattern): New function.
13942 (vect_vect_recog_func_ptrs): Add it.
13943 * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
13944 constant directly from the associated lhs.
13946 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
13948 * tree-vect-patterns.c (vect_split_statement): New function.
13949 (vect_convert_input): Use it to try to split an existing cast.
13951 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
13953 * poly-int.h (print_hex): New function.
13954 * dumpfile.h (dump_dec, dump_hex): Declare.
13955 * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
13956 * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
13957 min_input_precision, operation_precision and operation_sign.
13958 * tree-vect-patterns.c (vect_get_range_info): New function.
13959 (vect_same_loop_or_bb_p, vect_single_imm_use)
13960 (vect_operation_fits_smaller_type): Delete.
13961 (vect_look_through_possible_promotion): Add an optional
13962 single_use_p parameter.
13963 (vect_recog_over_widening_pattern): Rewrite to use new
13964 stmt_vec_info infomration. Handle one operation at a time.
13965 (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
13966 (vect_truncatable_operation_p, vect_set_operation_type)
13967 (vect_set_min_input_precision): New functions.
13968 (vect_determine_min_output_precision_1): Likewise.
13969 (vect_determine_min_output_precision): Likewise.
13970 (vect_determine_precisions_from_range): Likewise.
13971 (vect_determine_precisions_from_users): Likewise.
13972 (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
13973 (vect_vect_recog_func_ptrs): Put over_widening first.
13975 (vect_pattern_recog): Call vect_determine_precisions.
13977 2018-07-03 Richard Sandiford <richard.sandiford@arm.com>
13979 * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
13980 statements that have been replaced by further pattern statements.
13981 (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
13983 2018-07-03 Richard Biener <rguenther@suse.de>
13985 * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
13986 always set *dt. Dump vectype in vectype overload.
13987 * dumpfile.h (dump_gimple_expr): New function.
13988 (dump_gimple_expr_loc): Likewise.
13989 * dumpfile.c (dump_gimple_expr): New function.
13990 (dump_gimple_expr_loc): Likewise.
13992 2018-07-02 Jeff Law <law@redhat.com>
13994 * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
13995 the H8/300, H8/300H and H8/S variants into a single pattern.
13996 (movhi_h8300, movqi_h8300hs): Similarly.
13997 (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
13998 (QHI mode iterator): New.
14000 * config/h8300/h8300.md: Remove trailing whitespace.
14002 2018-07-02 Jim Wilson <jimw@sifive.com>
14004 * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
14005 instead of emit_insn for interrupt returns.
14006 * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
14007 (riscv_sret, riscv_uret): Likewise.
14009 2018-07-02 David Malcolm <dmalcolm@redhat.com>
14011 * pretty-print.c (selftest::test_pp_format): Move save and restore
14012 of quotes to class auto_fix_quotes, and add an instance.
14013 * selftest.c: Include "intl.h".
14014 (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
14015 (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
14016 * selftest.h (selftest::auto_fix_quotes): New class.
14018 2018-07-02 Richard Henderson <richard.henderson@linaro.org>
14020 * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
14021 (aarch64_sve_prepare_conditional_op): Remove.
14022 * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
14023 Allow aarch64_simd_reg_or_zero as select operand; remove
14024 the aarch64_sve_prepare_conditional_op call.
14025 (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
14026 (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
14027 (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
14028 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
14029 (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
14030 (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
14031 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
14032 (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
14033 and a splitters to match all of the *_any patterns.
14034 * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
14036 * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
14037 (SVE_COND_FP_BINARY_REV): Remove.
14038 (sve_int_op_rev, sve_fp_op_rev): New.
14039 * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
14040 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
14041 (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
14042 (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
14043 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
14044 (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
14045 (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
14046 (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
14047 (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
14049 * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
14050 Remove match_dup 1 from the inner unspec.
14051 (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
14053 * config/aarch64/aarch64.md (movprfx): New attr.
14054 (length): Default movprfx to 8.
14055 * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
14056 (*madd<SVE_I>, *msub<SVE_I): Likewise.
14057 (*<su>mul<SVE_I>3_highpart): Likewise.
14058 (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
14059 (*v<ASHIFT><SVE_I>3): Likewise.
14060 (*<su><MAXMIN><SVE_I>3): Likewise.
14061 (*<su><MAXMIN><SVE_F>3): Likewise.
14062 (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
14063 (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
14064 (*div<SVE_F>4): Likewise.
14066 2018-07-02 Richard Sandiford <richard.sandiford@arm.com>
14068 * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
14071 2018-07-02 Richard Biener <rguenther@suse.de>
14073 PR tree-optimization/86363
14074 * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
14075 memset argument refers to a non-variable address.
14077 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
14079 * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
14080 {PLUS,MINUS}_EXPR code to...
14081 (adjust_symbolic_bound): ...here,
14082 (combine_bound): ...here,
14083 (set_value_range_with_overflow): ...and here.
14085 2018-07-02 Aldy Hernandez <aldyh@redhat.com>
14087 * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
14089 (extract_range_from_abs_expr): ...here.
14091 2018-07-02 Eric Botcazou <ebotcazou@adacore.com>
14093 * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
14094 -fno-omit-frame-pointer when not optimizing.
14096 2018-07-02 Martin Liska <mliska@suse.cz>
14099 * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
14100 (propagate_nothrow): Likewise.
14102 2018-07-02 Martin Liska <mliska@suse.cz>
14105 * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
14107 2018-07-02 David Malcolm <dmalcolm@redhat.com>
14109 * dumpfile.c (dump_generic_expr_loc): Undo removal of this
14110 function in r262149, changing "loc" param from source_location to
14111 const dump_location_t &.
14112 * dumpfile.h (dump_generic_expr_loc): Undo removal of this
14113 declaration, as above.
14115 2018-07-01 Paul Koning <ni1d@arrl.net>
14117 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
14118 -munit-asm, -mgnu-asm, -mdec-asm.
14119 * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
14120 (pdp11_output_labelref): New.
14121 (pdp11_output_def): New.
14122 (pdp11_output_addr_vec_elt): New.
14123 * config/pdp11/pdp11.c: Use tab between opcode and operands. Use
14124 %# and %@ format codes.
14125 (pdp11_option_override): New.
14126 (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
14127 (pdp11_output_ident): New.
14128 (pdp11_asm_named_section): New.
14129 (pdp11_asm_init_sections): New.
14130 (pdp11_file_start): New.
14131 (pdp11_file_end): New.
14132 (output_ascii): Use .ascii/.asciz for -mdec-asm.
14133 (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm. Add
14134 %o, like %c but octal.
14135 (pdp11_option_override): New.
14136 * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
14138 (DATA_SECTION_ASM_OP): Ditto.
14139 (READONLY_DATA_SECTION_ASM_OP): New.
14140 (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
14141 (ASM_GENERATE_INTERNAL_LABEL): Use new function.
14142 (ASM_OUTPUT_LABELREF): Ditto.
14143 (ASM_OUTPUT_DEF): Ditto.
14144 (ASM_OUTPUT_EXTERNAL): New.
14145 (ASM_OUTPUT_SOURCE_FILENAME): New.
14146 (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
14147 (ASM_OUTPUT_SKIP): Update for -mdec-asm.
14148 * config/pdp11/pdp11.md: Use tab between opcode and operands. Use
14149 %# and %@ format codes.
14150 * config/pdp11/pdp11.opt (mgnu-asm): New.
14151 (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
14152 (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
14153 * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
14155 2018-07-01 Aldy Hernandez <aldyh@redhat.com>
14157 * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
14158 dereferencing path[] beyond its length.
14160 (debug_all_paths): New.
14161 (rewire_first_differing_edge): New.
14162 (adjust_paths_after_duplication): New.
14163 (duplicate_thread_path): Call adjust_paths_after_duplication.
14165 (thread_through_all_blocks): Add new argument to
14166 duplicate_thread_path.
14168 2018-06-30 Jim Wilson <jimw@sifive.com>
14170 * config/riscv/predicates.md (p2m1_shift_operand): New.
14171 (high_mask_shift_operand): New.
14172 * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
14173 pattern using p2m1_shift_operand.
14174 (lshsi3_zero_extend_3+2): New combiner pattern using
14175 high_mask_shift_operand.
14177 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
14179 * tree-vect-patterns.c (vect_get_external_def_edge): New function,
14181 (vect_recog_rotate_pattern): ...here.
14182 (vect_convert_input): Try to insert casts of invariants in the
14184 * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
14185 preheader to be empty.
14187 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
14189 * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
14190 vector type. If given, install it in the new statement's
14191 STMT_VINFO_VECTYPE.
14192 (vect_element_precision): New function.
14193 (vect_unpromoted_value): New struct.
14194 (vect_unpromoted_value::vect_unpromoted_value): New function.
14195 (vect_unpromoted_value::set_op): Likewise.
14196 (vect_look_through_possible_promotion): Likewise.
14197 (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
14198 (vect_widened_op_tree, vect_convert_input): Likewise.
14199 (vect_convert_inputs, vect_convert_output): Likewise.
14200 (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
14201 to handle the optional cast of the multiplication result and
14202 vect_widened_op_tree to detect the widened multiplication itself.
14203 Do not require the input and output of promotion casts to have
14204 the same sign, but base the signedness of the operation on the
14205 input rather than the result. If the pattern includes two
14206 promotions, check that those promotions have the same sign.
14207 Do not restrict the MULT_EXPR handling to a double-width result;
14208 handle quadruple-width results and wider. Use vect_convert_inputs
14209 to convert the inputs to the common type.
14210 (vect_recog_sad_pattern): Use vect_look_through_possible_promotion
14211 to handle the optional cast of the ABS result. Also allow a sign
14212 change or a sign extension between the ABS and MINUS.
14213 Use vect_widened_op_tree to detect the widened subtraction and use
14214 vect_convert_inputs to convert the inputs to the common type.
14215 (vect_handle_widen_op_by_const): Delete.
14216 (vect_recog_widen_op_pattern): New function.
14217 (vect_recog_widen_mult_pattern): Use it.
14218 (vect_recog_widen_shift_pattern): Likewise.
14219 (vect_recog_widen_sum_pattern): Use
14220 vect_look_through_possible_promotion to handle the promoted
14223 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
14225 * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
14226 the containing gimple_seq *.
14227 * gimple-iterator.h (gsi_for_stmt): Declare it.
14228 * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
14229 (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
14230 (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
14231 (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
14232 (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
14234 (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
14236 (vect_mark_pattern_stmts): ...here. Handle cases in which the
14237 statement being replaced is part of an existing pattern
14238 definition sequence, inserting the new pattern statements before
14240 (vect_pattern_recog_1): Don't return a bool. If the statement
14241 is already part of a pattern, instead apply pattern matching
14242 to the pattern definition statements. Don't clear the
14243 STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
14244 (vect_pattern_recog): Don't break after the first match;
14245 continue processing the pattern definition statements instead.
14246 Don't bail out for STMT_VINFO_IN_PATTERN_P here.
14248 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
14250 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14251 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
14252 (vect_recog_widen_sum_pattern): Use it.
14254 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
14256 * tree-vect-loop.c (vectorizable_reduction): Assert that the
14257 phi is not a pattern statement and has not been replaced by
14258 a pattern statement.
14259 * tree-vect-patterns.c (type_conversion_p): Don't check
14260 STMT_VINFO_IN_PATTERN_P.
14261 (vect_recog_vector_vector_shift_pattern): Likewise.
14262 (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
14263 the pattern statement rather than the original statement; check
14264 directly for a WIDEN_MULT_EXPR here.
14265 * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
14266 vect_is_simple_use to return the pattern statement rather
14267 than the original statement; use is_pattern_stmt_p to check
14268 for such a pattern statement.
14269 * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
14270 to return the pattern statement rather than the original statement;
14271 don't do the same transformation here.
14272 (vect_is_simple_use): If the defining statement has been replaced
14273 by a pattern statement, return the pattern statement instead.
14274 Remove the corresponding (local) transformation from the vectype
14277 2018-06-30 Richard Sandiford <richard.sandiford@arm.com>
14279 * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
14280 end and default to null.
14281 * tree-vect-loop.c (vect_create_epilog_for_reduction)
14282 (vectorizable_reduction): Update calls accordingly, dropping the
14283 gimple ** argument if the passed-back statement isn't needed.
14284 * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
14285 (vect_recog_rotate_pattern): Likewise.
14286 (vect_recog_mask_conversion_pattern): Likewise.
14287 * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
14288 (vect_mask_constant_operand_p): Likewise.
14289 * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
14290 (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
14291 (get_group_load_store_type, get_load_store_type): Likewise.
14292 (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
14293 (vectorizable_call, vectorizable_simd_clone_call): Likewise.
14294 (vectorizable_conversion, vectorizable_assignment): Likewise.
14295 (vectorizable_shift, vectorizable_operation): Likewise.
14296 (vectorizable_store, vect_is_simple_cond): Likewise.
14297 (vectorizable_condition, vectorizable_comparison): Likewise.
14298 (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
14299 (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
14300 and move it to the end. Cope with null def_stmt_outs.
14302 2018-06-30 Bernd Edlinger <bernd.edlinger@hotmail.de>
14304 * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
14306 2018-06-29 Jeff Law <law@redhat.com>
14308 * config/v850/v850.c (v850_legitimate_address_p): Handle large
14309 displacements for TARGET_V850E2V3 and newer.
14310 (TARGET_LRA_P): Remove. Defaults to LRA now.
14311 * config/v850/v850.md (sign23byte_load): Remove.
14312 (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
14313 (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
14315 2018-06-29 Martin Liska <mliska@suse.cz>
14318 * coverage.c (coverage_init): Mangle full path name.
14319 * doc/invoke.texi: Document the change.
14320 * gcov-io.c (mangle_path): New.
14321 * gcov-io.h (mangle_path): Likewise.
14322 * gcov.c (mangle_name): Use mangle_path for path mangling.
14324 2018-06-29 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14326 * config/arm/arm.c (output_move_double): Don't allow STRD instructions
14327 if starting source register is not even.
14329 2018-06-29 Martin Liska <mliska@suse.cz>
14331 PR tree-optimization/86263
14332 * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
14333 Make edge redirection.
14335 2018-06-29 David Malcolm <dmalcolm@redhat.com>
14337 * dumpfile.c (dump_loc): Add indentation based on scope depth.
14338 (dump_scope_depth): New variable.
14339 (get_dump_scope_depth): New function.
14340 (dump_begin_scope): New function.
14341 (dump_end_scope): New function.
14342 * dumpfile.h (get_dump_scope_depth): New declaration.
14343 (dump_begin_scope): New declaration.
14344 (dump_end_scope): New declaration.
14345 (class auto_dump_scope): New class.
14346 (AUTO_DUMP_SCOPE): New macro.
14347 * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
14350 2018-06-29 Richard Biener <rguenther@suse.de>
14352 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
14353 compute_all_dependences succeeds.
14354 * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
14355 exceed --param loop-max-datarefs-for-datadeps.
14357 2018-06-29 Jakub Jelinek <jakub@redhat.com>
14359 * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
14361 2018-06-28 Uros Bizjak <ubizjak@gmail.com>
14364 * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
14365 alternative 0 in preferred_for_speed attribute.
14367 2018-06-28 Paul Koning <ni1d@arrl.net>
14369 * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
14370 * config/pdp11/pdp11.c (pdp11_shift_length): New function.
14371 * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
14372 * config/pdp11/pdp11.md: Correct "length" attribute calculation
14373 for shift insn patterns.
14375 2018-06-28 David Malcolm <dmalcolm@redhat.com>
14377 * cgraph.c (cgraph_node::get_body): Replace assignments to
14378 "dump_file" with calls to set_dump_file.
14379 * dumpfile.c (alt_dump_file): Make static, and group with...
14380 (alt_flags): ...this definition.
14381 (dumps_are_enabled): New variable.
14382 (refresh_dumps_are_enabled): New function.
14383 (set_dump_file): New function.
14384 (set_alt_dump_file): New function.
14385 (gcc::dump_manager::dump_start): Replace assignments to
14386 "dump_file" and "alt_dump_file" with calls to set_dump_file and
14388 (gcc::dump_manager::dump_finish): Likewise.
14389 * dumpfile.h (alt_dump_file): Delete decl.
14390 (dumps_are_enabled): New variable decl.
14391 (set_dump_file): New function decl.
14392 (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
14394 * tree-nested.c (lower_nested_functions): Replace assignments to
14395 "dump_file" with calls to set_dump_file.
14397 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
14399 * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
14400 goto_locus of each outgoing edge of each basic block.
14402 2018-06-28 Richard Biener <rguenther@suse.de>
14404 * dwarf2out.c (decl_scope_table): Remove.
14405 (push_decl_scope): Likewise.
14406 (pop_decl_scope): Likewise.
14407 (gen_type_die_for_member): Do not call push/pop_decl_scope.
14408 (gen_struct_or_union_type_die): Likewise.
14409 (gen_tagged_type_die): Likewise.
14410 (dwarf2out_init): Do not initialize decl_scope_table.
14411 (dwarf2out_c_finalize): Do not free it.
14413 2018-06-28 Richard Biener <rguenther@suse.de>
14415 * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
14416 deciding whether to not re-use a DIE.
14418 2018-06-28 Richard Biener <rguenther@suse.de>
14420 * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
14421 DW_AT_abstract_origin attribute.
14423 2018-06-28 Martin Liska <mliska@suse.cz>
14425 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
14426 Use newly introduced constants.
14427 * tree-switch-conversion.h (struct jump_table_cluster):
14428 Define max_ratio_for_size and max_ratio_for_speed.
14430 2018-06-28 Martin Liska <mliska@suse.cz>
14432 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14433 Add new checking assert to catch invalid state.
14434 (jump_table_cluster::can_be_handled): Handle single case
14436 (jump_table_cluster::is_beneficial): Bail out for such case.
14437 (bit_test_cluster::find_bit_tests):
14438 Add new checking assert to catch invalid state.
14439 (bit_test_cluster::can_be_handled): Handle single case
14441 (bit_test_cluster::is_beneficial): Bail out for such case.
14442 (switch_decision_tree::analyze_switch_statement):
14445 2018-06-28 Martin Liska <mliska@suse.cz>
14447 * common.opt: Introduce -completion option.
14448 * gcc.c (driver_handle_option): Handle it.
14449 (driver::main): Print completions if completion
14451 * opt-suggestions.c (option_proposer::get_completions):
14453 (option_proposer::suggest_completion): Likewise.
14454 (option_proposer::find_param_completions): Likewise.
14455 (verify_autocompletions): Likewise.
14456 (test_completion_valid_options): Likewise.
14457 (test_completion_valid_params): Likewise.
14458 (in_completion_p): Likewise.
14459 (empty_completion_p): Likewise.
14460 (test_completion_partial_match): Likewise.
14461 (test_completion_garbage): Likewise.
14462 (opt_proposer_c_tests): Likewise.
14463 * opt-suggestions.h: Declare new functions.
14464 * opts.c (common_handle_option): Handle OPT__completion_.
14465 * selftest-run-tests.c (selftest::run_tests): Add
14466 opt_proposer_c_tests.
14467 * selftest.c (assert_str_startswith): New.
14468 * selftest.h (assert_str_startswith): Likewise.
14469 (opt_proposer_c_tests): New.
14470 (ASSERT_STR_STARTSWITH): Likewise.
14472 2018-06-28 Martin Liska <mliska@suse.cz>
14474 * Makefile.in: Add opt-suggestions.o.
14475 * gcc-main.c: Include opt-suggestions.h.
14476 * gcc.c (driver::driver): Likewise.
14477 (driver::~driver): Remove m_option_suggestions.
14478 (driver::build_option_suggestions): Moved to option_proposer.
14479 (driver::suggest_option): Likewise.
14480 (driver::handle_unrecognized_options): Use option_proposer.
14481 * gcc.h (class driver): Add new memver m_option_proposer.
14482 * opt-suggestions.c: New file.
14483 * opt-suggestions.h: New file.
14485 2018-06-28 Martin Liska <mliska@suse.cz>
14487 * vec.h (class auto_string_vec): New (moved from auto_argvec).
14488 (auto_string_vec::~auto_string_vec): Likewise.
14490 2018-06-28 Eric Botcazou <ebotcazou@adacore.com>
14492 * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
14493 prevent_decl_creation_for_types fields up and add reset_location field.
14494 * tree-inline.c (remap_gimple_stmt): Force input_location on the new
14495 statement if id->reset_location is true.
14496 (copy_edges_for_bb): Do not set goto_locus on the new edges if
14497 id->reset_location is true.
14498 (copy_phis_for_bb): Force input_location on the arguments if
14499 id->reset_location is true.
14500 (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
14501 is set on the function to be inlined.
14503 2018-06-27 Stephan Bergmann <sbergman@redhat.com>
14505 * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
14507 2018-06-27 Dimitar Dimitrov <dimitar@dinux.eu>
14509 * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
14510 registers for Pmode.
14511 * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
14512 hard registers for the clobbered pseudo.
14514 2018-06-27 Paul Koning <ni1d@arrl.net>
14516 * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
14517 mutually exclusive options.
14518 * config/pdp11/constraints.md (h): New constraint.
14519 (O): Update definition to match shift code generation.
14520 (D): New constraint.
14521 * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
14523 * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
14525 (output_jump): Change arguments.
14526 (pdp11_fixed_cc_regs): New function.
14527 (pdp11_cc_mode): Ditto.
14528 (pdp11_expand_shift): Ditto.
14529 (pdp11_assemble_shift): Ditto.
14530 (pdp11_small_shift): Ditto.
14531 (pdp11_branch_cost): Remove.
14532 * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
14534 (pdp11_register_move_cost): Update for CC registers.
14535 (pdp11_rtx_costs): Add case for LSHIFTRT.
14536 (pdp11_output_jump): Add CCNZ mode conditional branches.
14537 (notice_update_cc_on_set): Remove.
14538 (pdp11_cc_mode): New function.
14539 (simple_memory_operand): Correct pre/post decrement case.
14540 (no_side_effect_operand): New function.
14541 (pdp11_regno_reg_class): Add CC_REGS class.
14542 (pdp11_fixed_cc_regs): New function.
14543 (pdp11_small_shift): New function.
14544 (pdp11_expand_shift): New function to expand shift insns.
14545 (pdp11_assemble_shift): New function to output shifts.
14546 (pdp11_branch_cost): Remove.
14547 (pdp11_modes_tieable_p): Make QI/HI modes tieable.
14548 * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
14549 (WCHAR_TYPE): Ditto.
14550 (PTRDIFF_TYPE): Ditto.
14551 (ADJUST_INSN_LENGTH): New macro.
14552 (FIXED_REGISTERS): Add CC registers.
14553 (CALL_USED_REGISTERS): Ditto.
14554 (reg_class): Ditto.
14555 (REG_CLASS_NAMES): Ditto.
14556 (REG_CLASS_CONTENTS): Ditto.
14557 (SELECT_CC_MODE): Use new function.
14558 (TARGET_FLAGS_REGNUM): New macro.
14559 (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
14560 (cc0_reg_rtx): Remove.
14561 (CC_STATUS_MDEP): Remove.
14562 (CC_STATUS_MDEFP_INIT): Remove.
14563 (CC_IN_FPU): Remove.
14564 (NOTICE_UPDATE_CC): Remove.
14565 (REGISTER_NAMES): Add CC registers.
14566 (BRANCH_COST): Change to constant 1.
14567 * config/pdp11/pdp11.md: Rewrite for CCmode condition code
14569 * config/pdp11/pdp11.opt (mbcopy): Remove.
14570 (mbcopy-builtin): Remove.
14571 (mbranch-cheap): Remove.
14572 (mbranch-expensive): Remove.
14573 * config/pdp11/predicates.md (expand_shift_operand): Update to
14574 match shift code generation.
14575 (ccnz_operator): New predicate.
14576 * doc/invoke.texi (PDP-11 Options): Remove deleted options
14577 -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
14578 Remove non-existent option -mabshi, -mno-abshi. Document mutually
14580 * doc/md.texi (PDP-11): Document new D and h constraints. Update
14581 description of O constraint.
14583 2018-06-27 Jeff Law <law@redhat.com>
14584 Austin Law <austinklaw@gmail.com>
14586 * config/v850/v850.md (addsi3_set_flags): New pattern.
14587 (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
14588 (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
14589 (zero_extendhisi2_v850_set_flags): Likewise.
14590 (zero_extendqisi2_v850_set_flags): Likewise.
14591 (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
14592 (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
14593 (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
14595 * config/v850/v850-protos.h (notice_update_cc): Remove.
14596 * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
14597 (v850_print_operand): Handle 'D' and "d".
14598 (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
14599 Add handling of arithmetic/logical operations compared against zero.
14600 (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
14601 Do not look at v850_compare_op, instead get mode from last argument.
14602 (v850_gen_compare): Remove
14603 (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
14604 after reload for prologue insns.
14605 (expand_prologue): Account for CLOBBER of CC_REGNUM in various
14607 (construct_save_jarl): Likewise.
14608 (TARGET_FLAGS_REGNUM): Define.
14609 * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
14610 (NOTICE_UPDATE_CC): Remove.
14611 * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
14612 than cc0. Conditionalize on reload_completed.
14613 (cmpsi_insn, setfcc_insn): Likewise.
14614 (tst1 splitter): Turn into define_and_split which sets the flags
14616 (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
14617 (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
14618 (cstoresf4, cstoredf4): Clobber the flags.
14619 (cmpsi, cmpsf, cmpdf): Remove expanders.
14620 (setf_insn): Remove pattern.
14621 (addsi3): Turn into define_and_split which clobbers the flags after
14622 reload and a suitable pattern (addsi3_clobber_flags) for use after
14624 (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
14625 (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
14626 (ashrsi3, ashrsi3_v850e2): Likewise.
14627 (bins): Clobber the flags.
14628 (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
14629 (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
14630 (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
14631 (call_value_internal_short, call_value_internal_long): Likewise.
14632 (callt_save_interrupt, callt_return_interrupt): Likewise.
14633 (save_interrupt, return_interrupt): Likewise.
14634 (callt_save_all_interrupt, save_all_interrupt): Likewise.
14635 (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
14636 (restore_all_interrupt, _restore_all_interrupt): Likewise.
14637 (All FP comparisons): Only allow after reload has completed.
14639 (divh, divhu): Tweak output template.
14640 (branch_z_normal, branch_z_invert): Remove
14641 (branch_nz_normal, branch_nz_invert): Likewise.
14642 (extendhisi_insn, extendqisi_insn): Do not clobber flags.
14644 * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
14645 * config/v850/v850.c (notice_update_cc): Remove.
14646 * config/v850/v850.h (CC_OVERFLOW_UNUSABLE): Remove
14647 (CC_NO_CARRY): Likewise.
14648 (NOTICE_UPDATE_CC): Define to nothing.
14649 * config/v850/v850.md: Remove block comment on cc0 handling
14650 Remove "cc" attribute from all patterns. Remove cc_status handling
14651 from all patterns. Minor formatting fixes.
14653 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14655 * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
14656 (cortex-a76.cortex-a55): Likewise.
14657 * config/aarch64/aarch64-tune.md: Regenerate.
14658 * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
14659 cortex-a76.cortex-a55.
14661 2018-06-27 Jeff Law <law@redhat.com>
14663 * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
14664 (MULTILIB_DIRNAMES): Similarly.
14666 2018-06-27 Eric Botcazou <ebotcazou@adacore.com>
14668 * gimple.h (gimple_return_retbnd): Delete.
14669 (gimple_return_set_retbnd): Likewise.
14670 * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
14671 gimple_return_set_retbnd.
14672 * gimple-pretty-print.c (dump_gimple_return): Remove call to
14673 gimple_return_retbnd and adjust.
14674 * tree-inline.h (struct copy_body_data): Remove retbnd field.
14675 * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
14676 Explicitly return NULL in a couple more cases. Move assertion
14677 on debug statements and remove unreachable code.
14678 (reset_debug_binding): Do not test id->retbnd.
14679 (expand_call_inline): Do not set it.
14681 2018-06-27 Rasmus Villemoes <rv@rasmusvillemoes.dk>
14683 * configure.ac: Add --disable-gcov option.
14684 * configure: Regenerate.
14685 * Makefile.in: Honour @enable_gcov@.
14686 * doc/install.texi: Document --disable-gcov.
14688 2018-06-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14690 * config/arm/arm-cpus.in (cortex-a76): New entry.
14691 (cortex-a76.cortex-a55): Likewise.
14692 * config/arm/arm-tables.opt: Regenerate.
14693 * config/arm/arm-tune.md: Likewise.
14694 * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
14695 * doc/invoke.texi (ARM Options): Document cortex-a76 and
14696 cortex-a76.cortex-a55.
14698 2018-06-27 Tamar Christina <tamar.christina@arm.com>
14701 * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
14703 2018-06-27 Siddhesh Poyarekar <siddhesh@sourceware.org>
14705 * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
14707 (EPILOGUE_USES): Likewise.
14709 2018-06-26 Eric Botcazou <ebotcazou@adacore.com>
14711 * tree-inline.c (remap_location): New function extracted from...
14712 (copy_edges_for_bb): Add ID parameter. Remap goto_locus.
14713 (copy_phis_for_bb): ...here. Call remap_location.
14714 (copy_cfg_body): Adjust call to copy_edges_for_bb.
14716 2018-06-26 Aaron Sawdey <acsawdey@linux.ibm.com>
14718 * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
14719 unaligned vsx for 16B memset.
14721 2018-06-26 Segher Boessenkool <segher@kernel.crashing.org>
14724 * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
14725 ieee128_float_type_node to long_double_type_node unless
14726 TARGET_LONG_DOUBLE_128 is set.
14728 2018-06-26 David Malcolm <dmalcolm@redhat.com>
14730 * cfgloop.c (get_loop_location): Convert return type from
14731 location_t to dump_user_location_t, replacing INSN_LOCATION lookups
14732 by implicit construction from rtx_insn *, and using
14733 dump_user_location_t::from_function_decl for the fallback case.
14734 * cfgloop.h (get_loop_location): Convert return type from
14735 location_t to dump_user_location_t.
14736 * cgraphunit.c (walk_polymorphic_call_targets): Update call to
14737 dump_printf_loc to pass in a dump_location_t rather than a
14738 location_t, via the gimple stmt.
14739 * coverage.c (get_coverage_counts): Update calls to
14740 dump_printf_loc to pass in dump_location_t rather than a
14742 * doc/optinfo.texi (Dump types): Convert example of
14743 dump_printf_loc from taking "locus" to taking "insn". Update
14744 description of the "_loc" calls to cover dump_location_t.
14745 * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
14747 (dump_user_location_t::dump_user_location_t): New constructors,
14748 from gimple *stmt and rtx_insn *.
14749 (dump_user_location_t::from_function_decl): New function.
14750 (dump_loc): Make static.
14751 (dump_gimple_stmt_loc): Convert param "loc" from location_t to
14752 const dump_location_t &.
14753 (dump_generic_expr_loc): Delete.
14754 (dump_printf_loc): Convert param "loc" from location_t to
14755 const dump_location_t &.
14756 (selftest::test_impl_location): New function.
14757 (selftest::dumpfile_c_tests): New function.
14758 * dumpfile.h: Include "profile-count.h".
14759 (class dump_user_location_t): New class.
14760 (struct dump_impl_location_t): New struct.
14761 (class dump_location_t): New class.
14762 (dump_printf_loc): Convert 2nd param from source_location to
14763 const dump_location_t &.
14764 (dump_generic_expr_loc): Delete.
14765 (dump_gimple_stmt_loc): Convert 2nd param from source_location to
14766 const dump_location_t &.
14767 * gimple-fold.c (fold_gimple_assign): Update call to
14768 dump_printf_loc to pass in a dump_location_t rather than a
14769 location_t, via the gimple stmt.
14770 (gimple_fold_call): Likewise.
14771 * gimple-loop-interchange.cc
14772 (loop_cand::analyze_iloop_reduction_var): Update for change to
14773 check_reduction_path.
14774 (tree_loop_interchange::interchange): Update for change to
14775 find_loop_location.
14776 * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
14777 change in return-type of find_loop_location.
14778 (graphite_regenerate_ast_isl): Likewise.
14779 * graphite-optimize-isl.c (optimize_isl): Likewise.
14780 * graphite.c (graphite_transform_loops): Likewise.
14781 * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
14782 pass in a dump_location_t rather than a location_t, via the
14784 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14785 * ipa.c (walk_polymorphic_call_targets): Likewise.
14786 * loop-unroll.c (report_unroll): Convert "locus" param from
14787 location_t to dump_location_t.
14788 (decide_unrolling): Update for change to get_loop_location's
14790 * omp-grid.c (struct grid_prop): Convert field "target_loc" from
14791 location_t to dump_user_location_t.
14792 (grid_find_single_omp_among_assignments_1): Updates calls to
14793 dump_printf_loc to pass in a dump_location_t rather than a
14794 location_t, via the gimple stmt.
14795 (grid_parallel_clauses_gridifiable): Convert "tloc" from
14796 location_t to dump_location_t. Updates calls to dump_printf_loc
14797 to pass in a dump_location_t rather than a location_t, via the
14799 (grid_inner_loop_gridifiable_p): Likewise.
14800 (grid_dist_follows_simple_pattern): Likewise.
14801 (grid_gfor_follows_tiling_pattern): Likewise.
14802 (grid_target_follows_gridifiable_pattern): Likewise.
14803 (grid_attempt_target_gridification): Convert initialization
14804 of local "grid" from memset to zero-initialization; FIXME: does
14805 this require C++11? Update call to dump_printf_loc to pass in a
14806 optinfo_location rather than a location_t, via the gimple stmt.
14807 * profile.c (read_profile_edge_counts): Updates call to
14808 dump_printf_loc to pass in a dump_location_t rather than a
14810 (compute_branch_probabilities): Likewise.
14811 * selftest-run-tests.c (selftest::run_tests): Call
14813 * selftest.h (dumpfile_c_tests): New decl.
14814 * tree-loop-distribution.c (pass_loop_distribution::execute):
14815 Update for change in return type of find_loop_location.
14816 * tree-parloops.c (parallelize_loops): Likewise.
14817 * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
14818 "locus" from location_t to dump_user_location_t.
14819 (canonicalize_loop_induction_variables): Likewise.
14820 * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
14821 for change in return type of find_loop_location.
14822 * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
14823 to dump_printf_loc to pass in a dump_location_t rather than a
14824 location_t, via the stmt.
14825 * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
14827 * tree-vect-loop-manip.c (find_loop_location): Convert return
14828 type from source_location to dump_user_location_t.
14829 (vect_do_peeling): Update for above change.
14830 (vect_loop_versioning): Update for change in type of
14832 * tree-vect-loop.c (check_reduction_path): Convert "loc" param
14833 from location_t to dump_user_location_t.
14834 (vect_estimate_min_profitable_iters): Update for change in type
14836 * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
14837 location_t to dump_location_t.
14838 (vect_slp_bb): Update for change in type of vect_location.
14839 * tree-vectorizer.c (vect_location): Convert from source_location
14840 to dump_user_location_t.
14841 (try_vectorize_loop_1): Update for change in vect_location's type.
14842 (vectorize_loops): Likewise.
14843 (increase_alignment): Likewise.
14844 * tree-vectorizer.h (vect_location): Convert from source_location
14845 to dump_user_location_t.
14846 (find_loop_location): Convert return type from source_location to
14847 dump_user_location_t.
14848 (check_reduction_path): Convert 1st param from location_t to
14849 dump_user_location_t.
14850 * value-prof.c (check_counter): Update call to dump_printf_loc to
14851 pass in a dump_user_location_t rather than a location_t; update
14852 call to error_at for change in type of "locus".
14853 (check_ic_target): Update call to dump_printf_loc to
14854 pass in a dump_user_location_t rather than a location_t, via the
14857 2018-06-26 Robin Dapp <rdapp@linux.vnet.ibm.com>
14859 * config/s390/s390.h (enum processor_flags): Do not use
14860 default tune parameter when -march was specified.
14862 2018-06-26 Jakub Jelinek <jakub@redhat.com>
14865 * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
14866 Check reg_overlap_mentioned_p in addition to reg_set_p with the same
14869 2018-06-26 Richard Biener <rguenther@suse.de>
14871 PR tree-optimization/86287
14873 * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
14874 (vect_analyze_loop): Initialize n_stmts.
14876 2018-06-26 Richard Biener <rguenther@suse.de>
14878 PR middle-end/86271
14879 * fold-const.c (fold_convertible_p): Pointer extension
14882 2018-06-26 Alexandre Oliva <aoliva@redhat.com>
14885 * dwarf2out.c (loc_list_has_views): Adjust comments.
14886 (dw_loc_list): Split single cross-partition range with
14889 2018-06-25 Jeff Law <law@redhat.com>
14891 * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
14892 on -mbig-switch by default.
14894 * config/v850/predicates.md (const_float_1_operand): Fix match_code
14896 (const_float_0_operand): Remove unused predicate.
14897 * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
14898 (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
14899 (recipsf2): New expander. Original pattern now called
14901 (recipdf2, recipdf2_insn): Similarly.
14902 (rsqrtsf2, rsqrtsf2_insn): Similarly
14903 (rsqrtdf2, rsqrtdf2_insn): Similarly
14905 2018-06-26 Gerald Pfeifer <gerald@pfeifer.com>
14907 * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
14908 Simplify logic for FreeBSD (twice).
14910 2018-06-25 Martin Sebor <msebor@redhat.com>
14912 PR tree-optimization/86204
14913 * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
14914 a strnlen result if it's less than the length of the string.
14916 2018-06-25 Martin Sebor <msebor@redhat.com>
14918 PR tree-optimization/85700
14919 * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
14920 * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
14921 (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
14923 2018-06-25 Martin Sebor <msebor@redhat.com>
14925 * doc/extend.texi (Zero-length arrays): Update and clarify.
14927 2018-06-25 Michael Meissner <meissner@linux.ibm.com>
14929 * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
14930 added IEEE/IBM long double multilib support on PowerPC little
14931 endian Linux systems.
14932 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
14933 (MULTILIB_DEFAULTS): Likewise.
14934 * config/rs6000/rs6000.c (rs6000_option_override_internal):
14936 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
14937 * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
14938 * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
14940 2018-06-25 Alexander Monakov <amonakov@ispras.ru>
14942 PR middle-end/86311
14943 * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
14944 (REORDER_45): Likewise.
14946 2018-06-25 Jeff Law <law@redhat.com>
14948 * config/v850/v850.md (divmodhi4): Make sure to sign extend the
14949 dividend to 32 bits. Adjust length.
14950 (udivmodhi4): Cleanup output template. Fix length.
14952 2018-06-25 Carl Love <cel@us.ibm.com>
14954 * config/rs6000/vsx.md: Change word selector to prefered location.
14956 2018-06-25 Richard Biener <rguenther@suse.de>
14958 PR tree-optimization/86304
14959 * tree-vectorizer.c (vectorize_loops): Walk over new possibly
14960 epilogue-if-converted loops as well.
14962 2018-06-25 Jan Hubicka <hubicka@ucw.cz>
14964 * lto-section-out.c (lto_begin_section): Do not print section
14965 name for noaddr and unnumbered dumps.
14967 2018-06-25 Richard Biener <rguenther@suse.de>
14969 * tree-vectorizer.h (struct vec_info_shared): New structure
14970 with parts split out from struct vec_info and loop_nest from
14971 struct _loop_vec_info.
14972 (struct vec_info): Adjust accordingly.
14973 (struct _loop_vec_info): Likewise.
14974 (LOOP_VINFO_LOOP_NEST): Adjust.
14975 (LOOP_VINFO_DATAREFS): Likewise.
14976 (LOOP_VINFO_DDRS): Likewise.
14977 (struct _bb_vec_info): Likewise.
14978 (BB_VINFO_DATAREFS): Likewise.
14979 (BB_VINFO_DDRS): Likewise.
14980 (struct _stmt_vec_info): Add dr_aux member.
14981 (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
14982 (DR_MISALIGNMENT_UNINITIALIZED): New.
14983 (set_dr_misalignment): Adjust.
14984 (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
14985 (vect_analyze_loop): Adjust prototype.
14986 (vect_analyze_loop_form): Likewise.
14987 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
14988 Compute dependences lazily.
14989 (vect_record_base_alignments): Use shared datarefs/ddrs.
14990 (vect_verify_datarefs_alignment): Likewise.
14991 (vect_analyze_data_refs_alignment): Likewise.
14992 (vect_analyze_data_ref_accesses): Likewise.
14993 (vect_analyze_data_refs): Likewise.
14994 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
14995 constructor parameter for shared part.
14996 (vect_analyze_loop_form): Pass in shared part and adjust.
14997 (vect_analyze_loop_2): Pass in storage for the number of
14998 stmts. Move loop nest finding to the caller. Compute
15000 (vect_analyze_loop): Pass in shared part.
15001 (vect_transform_loop): Verify shared datarefs are unchanged.
15002 * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
15003 constructor parameter for shared part.
15004 (vect_slp_analyze_bb_1): Pass in shared part and adjust.
15005 (vect_slp_bb): Verify shared datarefs are unchanged before
15007 * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
15009 (new_stmt_vec_info): Initialize DR_AUX misalignment to
15010 DR_MISALIGNMENT_UNINITIALIZED.
15011 * tree-vectorizer.c (vec_info::vec_info): Add constructor
15012 parameter for shared part.
15013 (vec_info::~vec_info): Adjust.
15014 (vec_info_shared::vec_info_shared): New.
15015 (vec_info_shared::~vec_info_shared): Likewise.
15016 (vec_info_shared::save_datarefs): Likewise.
15017 (vec_info_shared::check_datarefs): Likewise.
15018 (try_vectorize_loop_1): Construct shared part live for analyses
15019 of a single loop for multiple vector sizes.
15020 * tree-parloops.c (gather_scalar_reductions): Adjust.
15022 2018-06-25 Richard Biener <rguenther@suse.de>
15024 * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
15025 DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
15026 (vect_analyze_data_refs): Remove similar code from here and
15027 simplify accordingly.
15029 2018-06-25 Richard Biener <rguenther@suse.de>
15031 * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
15032 for reverse storage order accesses rather than asserting
15033 they cannot happen here.
15035 2018-06-25 Tom de Vries <tdevries@suse.de>
15038 * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
15039 Use data16 instead of .byte for insn prefix.
15041 2018-06-25 Andreas Krebbel <krebbel@linux.ibm.com>
15044 * parser.c (make_char_string_pack): Pass this literal chars
15045 through cpp_interpret_string.
15046 (cp_parser_userdef_numeric_literal): Check the result of
15047 make_char_string_pack.
15049 2018-06-24 Maya Rashish <coypu@sdf.org>
15051 * ginclude/stddef.h: Simplify conditions around avoiding
15052 re-definition of __size_t.
15054 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
15056 * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
15057 unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
15059 2018-06-22 Maya Rashish <coypu@sdf.org>
15061 * doc/invoke.texi (mno-fancy-math-387): Update for changes
15062 made to OpenBSD and NetBSD through the years.
15064 2018-06-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
15066 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
15067 behavior of vec_pack (vector double, vector double) to match
15068 behavior of vec_float2 (vector double, vector double).
15070 2018-06-22 Olivier Hainque <hainque@adacore.com>
15072 * gimplify.c (gimplify_function_tree): Prevent creation
15073 of a trampoline for the address of the current function
15074 passed to entry/exit instrumentation hooks.
15076 2018-06-22 Aaron Sawdey <acsawdey@linux.ibm.com>
15079 * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
15082 2018-06-22 Martin Liska <mliska@suse.cz>
15084 PR tree-optimization/86263
15085 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
15086 Bail out if is_enabled is false.
15087 * tree-switch-conversion.h (jump_table_cluster::is_enabled):
15089 (jump_table_cluster::is_enabled): New function.
15091 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
15093 * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
15094 BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
15095 * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
15096 (lto_input_ts_binfo_tree_pointers): Likewise.
15097 * tree-streamer-out.c (streamer_write_tree_bitfields,
15098 write_ts_binfo_tree_pointers): Likewise.
15099 * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
15101 2018-06-22 Jan Hubicka <hubicka@ucw.cz>
15103 * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
15105 2018-06-22 Martin Liska <mliska@suse.cz>
15107 * symbol-summary.h (get): Make it pure and inline move
15108 functionality from ::get function.
15109 (get): Remove and inline into ::get and ::get_create.
15110 (get_create): Move code from ::get function.
15112 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15115 * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
15116 -x assembler-with-cpp.
15118 2018-06-22 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15120 * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
15121 _FILE_OFFSET_BITS=64 for C++.
15123 2018-06-21 Michael Meissner <meissner@linux.ibm.com>
15125 * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
15126 conversion insn that shows up when pr85657-3.c is compiled using
15127 IEEE 128-bit long double.
15128 (neg<mode>2_internal): Use the correct mode to check whether the
15129 mode is IBM extended.
15130 * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
15131 multiply and divide external functions from being created more
15134 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
15136 * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
15138 (rtl_merge_blocks): Likewise. Do not emit a NOP if the location of
15139 the edge can be forwarded.
15140 (cfg_layout_merge_blocks): Likewise.
15142 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
15144 * except.c (finish_eh_generation): Commit edge insertions only after
15145 the EH edges have been redirected from post-landing to landing pads.
15147 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
15149 * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
15150 create_tmp_var_for to create the FRAME decl.
15151 (finalize_nesting_tree_1): Do not unchain the FRAME decl.
15153 2018-06-21 Eric Botcazou <ebotcazou@adacore.com>
15155 * tree-inline.c (copy_edges_for_bb): Minor tweak.
15156 (maybe_move_debug_stmts_to_successors): Also reset the locus of the
15157 debug statement when resetting its value.
15158 (expand_call_inline): Copy the locus of the call onto the assignment
15159 of the return value, if any. Use local variable in more cases.
15161 2018-06-21 Martin Liska <mliska@suse.cz>
15163 * ipa-pure-const.c (propagate_nothrow): Use
15164 funct_state_summaries->get.
15165 (dump_malloc_lattice): Likewise.
15166 (propagate_malloc): Likewise.
15168 2018-06-21 Richard Biener <rguenther@suse.de>
15170 * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
15171 comment. Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
15172 * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
15174 * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
15175 BLOCK_ABSTRACT_ORIGIN unconditionally.
15177 2018-06-21 David Malcolm <dmalcolm@redhat.com>
15179 * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
15181 * ipa-reference.c (ipa_reference_c_finalize): Delete
15182 ipa_ref_opt_sum_summaries and set it to NULL.
15184 2018-06-21 Tom de Vries <tdevries@suse.de>
15186 PR tree-optimization/85859
15187 * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
15188 test with comment from bb_no_side_effects_p.
15190 2018-06-21 Richard Biener <rguenther@suse.de>
15192 PR tree-optimization/86232
15193 * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
15194 max for constant niter.
15196 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
15198 * config/aarch64/aarch64-simd.md
15199 (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
15201 2018-06-21 Andre Vieira <andre.simoesdiasvieira@arm.com>
15203 * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
15204 Make opernads of the unspec commutative.
15206 2018-06-21 Richard Biener <rguenther@suse.de>
15208 * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
15209 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
15210 Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
15211 (vect_analyze_data_ref_dependence): Re-order checks to deal with
15213 (vect_record_base_alignments): Do not record base alignment
15214 for gathers or scatters.
15215 (vect_compute_data_ref_alignment): Drop return value that is always
15216 true. Bail out early for gathers or scatters.
15217 (vect_enhance_data_refs_alignment): Bail out early for gathers
15219 (vect_find_same_alignment_drs): Likewise.
15220 (vect_analyze_data_refs_alignment): Remove dead code.
15221 (vect_slp_analyze_and_verify_node_alignment): Likewise.
15222 (vect_analyze_data_refs): For possible gathers or scatters do
15223 not create an alternate DR, just check their possible validity
15224 and mark them. Adjust DECL_NONALIASED handling to not rely
15225 on DR_BASE_ADDRESS.
15226 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
15227 update inits of gathers or scatters.
15228 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
15229 Also copy gather/scatter flag to pattern vinfo.
15231 2018-06-20 Kelvin Nilsen <kelvin@gcc.gnu.org>
15233 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
15234 behavior of vec_packsu (vector unsigned long long, vector unsigned
15235 long long) to match behavior of vec_packs with same signature.
15237 2018-06-20 Chung-Lin Tang <cltang@codesourcery.com>
15238 Thomas Schwinge <thomas@codesourcery.com>
15239 Cesar Philippidis <cesar@codesourcery.com>
15241 * gimplify.c (gimplify_scan_omp_clauses): Add support for
15242 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
15243 (gimplify_adjust_omp_clauses): Likewise.
15244 (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
15245 support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
15246 (gimplify_omp_target_update): Update handling of acc update and
15248 * omp-low.c (install_var_field): Remove unused parameter
15249 base_pointers_restrict.
15250 (scan_sharing_clauses): Remove base_pointers_restrict parameter.
15251 Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
15253 (omp_target_base_pointers_restrict_p): Delete.
15254 (scan_omp_target): Update call to scan_sharing_clauses.
15255 * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
15257 * tree-nested.c (convert_nonlocal_omp_clauses): Handle
15258 OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
15259 (convert_local_omp_clauses): Likewise.
15260 * tree-pretty-print.c (dump_omp_clause): Likewise.
15261 * tree.c (omp_clause_num_ops): Add entries for OMP_CLAUSE_{IF_PRESENT,
15263 (omp_clause_code_name): Likewise.
15265 2018-06-20 Jakub Jelinek <jakub@redhat.com>
15268 * var-tracking.c (use_narrower_mode_test): Check if shift amount can
15271 PR tree-optimization/86231
15272 * tree-vrp.c (union_ranges): For ( [ ) ] or ( )[ ] range and
15273 anti-range don't overwrite *vr0min before using it to compute *vr0max.
15275 2018-06-20 Tom de Vries <tdevries@suse.de>
15277 PR tree-optimization/86097
15278 * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
15279 iv type if signedness of iv type is not the same as that of *nit.
15281 2018-06-20 Jakub Jelinek <jakub@redhat.com>
15283 * cfgrtl.c (rtl_verify_edges): Formatting fix. If bb->preds has any
15284 EDGE_EH edges, verify they are all EDGE_EH.
15286 2018-06-20 Maya Rashish <coypu@sdf.org>
15288 * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
15290 2018-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
15292 * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
15293 * config/aarch64/aarch64.c (xgene1_tunings): Add
15294 AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
15295 (aarch64_mode_valid_for_sched_fusion_p):
15296 Allow 16-byte modes.
15297 (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
15298 * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
15300 * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
15302 (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
15303 * config/aarch64/iterators.md (VQ2): New mode iterator.
15305 2018-06-20 Martin Liska <mliska@suse.cz>
15307 * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
15308 Change default ratio from 10 to 8.
15310 2018-06-20 Martin Liska <mliska@suse.cz>
15312 * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
15314 (bit_test_cluster::find_bit_tests): Likewise.
15315 (switch_decision_tree::analyze_switch_statement): Find clusters.
15316 * tree-switch-conversion.h (struct jump_table_cluster): Document
15319 2018-06-20 Martin Liska <mliska@suse.cz>
15321 * tree-switch-conversion.c (switch_conversion::collect):
15322 Record m_uniq property.
15323 (switch_conversion::expand): Bail out for special conditions.
15324 (group_cluster::~group_cluster): New.
15325 (group_cluster::group_cluster): Likewise.
15326 (group_cluster::dump): Likewise.
15327 (jump_table_cluster::emit): New.
15328 (switch_decision_tree::fix_phi_operands_for_edges): New.
15329 (struct case_node): Remove struct.
15330 (jump_table_cluster::can_be_handled): New.
15331 (case_values_threshold): Moved to header.
15332 (reset_out_edges_aux): Likewise.
15333 (jump_table_cluster::is_beneficial): New.
15334 (bit_test_cluster::can_be_handled): Likewise.
15335 (add_case_node): Remove.
15336 (bit_test_cluster::is_beneficial): New.
15337 (case_bit_test::cmp): New.
15338 (bit_test_cluster::emit): New.
15339 (expand_switch_as_decision_tree_p): Remove.
15340 (bit_test_cluster::hoist_edge_and_branch_if_true): New.
15341 (fix_phi_operands_for_edge): Likewise.
15342 (switch_decision_tree::analyze_switch_statement): New.
15343 (compute_cases_per_edge): Move ...
15344 (switch_decision_tree::compute_cases_per_edge): ... here.
15345 (try_switch_expansion): Likewise.
15346 (switch_decision_tree::try_switch_expansion): Likewise.
15347 (record_phi_operand_mapping): Likewise.
15348 (switch_decision_tree::record_phi_operand_mapping): Likewise.
15349 (emit_case_decision_tree): Likewise.
15350 (switch_decision_tree::emit): Likewise.
15351 (balance_case_nodes): Likewise.
15352 (switch_decision_tree::balance_case_nodes): Likewise.
15353 (dump_case_nodes): Likewise.
15354 (switch_decision_tree::dump_case_nodes): Likewise.
15355 (emit_jump): Likewise.
15356 (switch_decision_tree::emit_jump): Likewise.
15357 (emit_cmp_and_jump_insns): Likewise.
15358 (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
15359 (emit_case_nodes): Likewise.
15360 (switch_decision_tree::emit_case_nodes): Likewise.
15361 (conditional_probability): Remove.
15362 * tree-switch-conversion.h (enum cluster_type): New.
15364 (struct cluster): Likewise.
15365 (cluster::cluster): Likewise.
15366 (struct simple_cluster): Likewise.
15367 (simple_cluster::simple_cluster): Likewise.
15368 (struct group_cluster): Likewise.
15369 (struct jump_table_cluster): Likewise.
15370 (struct bit_test_cluster): Likewise.
15371 (struct min_cluster_item): Likewise.
15372 (struct case_tree_node): Likewise.
15373 (case_tree_node::case_tree_node): Likewise.
15374 (jump_table_cluster::case_values_threshold): Likewise.
15375 (struct case_bit_test): Likewise.
15376 (struct switch_decision_tree): Likewise.
15377 (struct switch_conversion): Likewise.
15378 (switch_decision_tree::reset_out_edges_aux): Likewise.
15380 2018-06-20 Martin Liska <mliska@suse.cz>
15382 * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
15383 (hoist_edge_and_branch_if_true): Likewise.
15384 (expand_switch_using_bit_tests_p): Likewise.
15385 (struct case_bit_test): Likewise.
15386 (case_bit_test_cmp): Likewise.
15387 (emit_case_bit_tests): Likewise.
15388 (switch_conversion::switch_conversion): New class.
15389 (struct switch_conv_info): Remove old struct.
15390 (collect_switch_conv_info): More to ...
15391 (switch_conversion::collect): ... this.
15392 (check_range): Likewise.
15393 (switch_conversion::check_range): Likewise.
15394 (check_all_empty_except_final): Likewise.
15395 (switch_conversion::check_all_empty_except_final): Likewise.
15396 (check_final_bb): Likewise.
15397 (switch_conversion::check_final_bb): Likewise.
15398 (create_temp_arrays): Likewise.
15399 (switch_conversion::create_temp_arrays): Likewise.
15400 (free_temp_arrays): Likewise.
15401 (gather_default_values): Likewise.
15402 (switch_conversion::gather_default_values): Likewise.
15403 (build_constructors): Likewise.
15404 (switch_conversion::build_constructors): Likewise.
15405 (constructor_contains_same_values_p): Likewise.
15406 (switch_conversion::contains_same_values_p): Likewise.
15407 (array_value_type): Likewise.
15408 (switch_conversion::array_value_type): Likewise.
15409 (build_one_array): Likewise.
15410 (switch_conversion::build_one_array): Likewise.
15411 (build_arrays): Likewise.
15412 (switch_conversion::build_arrays): Likewise.
15413 (gen_def_assigns): Likewise.
15414 (switch_conversion::gen_def_assigns): Likewise.
15415 (prune_bbs): Likewise.
15416 (switch_conversion::prune_bbs): Likewise.
15417 (fix_phi_nodes): Likewise.
15418 (switch_conversion::fix_phi_nodes): Likewise.
15419 (gen_inbound_check): Likewise.
15420 (switch_conversion::gen_inbound_check): Likewise.
15421 (process_switch): Use the newly created class.
15422 (switch_conversion::expand): New.
15423 (switch_conversion::~switch_conversion): New.
15424 * tree-switch-conversion.h: New file.
15426 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
15428 * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
15429 tree-vect-patterns.c.
15430 * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
15431 (vect_recog_dot_prod_pattern): Use it. Remove the type_in argument.
15432 (vect_recog_sad_pattern): Likewise.
15433 (vect_recog_widen_sum_pattern): Likewise.
15434 (vect_recog_pow_pattern): Likewise. Check for a null vectype.
15435 (vect_recog_widen_shift_pattern): Remove the type_in argument.
15436 (vect_recog_rotate_pattern): Likewise.
15437 (vect_recog_mult_pattern): Likewise.
15438 (vect_recog_vector_vector_shift_pattern): Likewise.
15439 (vect_recog_divmod_pattern): Likewise.
15440 (vect_recog_mixed_size_cond_pattern): Likewise.
15441 (vect_recog_bool_pattern): Likewise.
15442 (vect_recog_mask_conversion_pattern): Likewise.
15443 (vect_try_gather_scatter_pattern): Likewise.
15444 (vect_recog_widen_mult_pattern): Likewise. Check for a null vectype.
15445 (vect_recog_over_widening_pattern): Likewise.
15446 (vect_recog_gather_scatter_pattern): Likewise.
15447 (vect_recog_func_ptr): Move from tree-vectorizer.h
15448 (vect_vect_recog_func_ptrs): Move further down the file.
15449 (vect_recog_func): Likewise. Remove the third argument.
15450 (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
15451 (vect_pattern_recog_1): Expect the pattern function to do any
15452 necessary target tests. Also expect it to provide a vector type.
15453 Remove the type_in handling.
15455 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
15457 * tree-vect-patterns.c (vect_pattern_detected): New function.
15458 (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
15459 (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
15460 (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
15461 (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
15462 (vect_recog_mult_pattern, vect_recog_divmod_pattern)
15463 (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
15464 (vect_recog_mask_conversion_pattern)
15465 (vect_try_gather_scatter_pattern): Likewise.
15467 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
15469 * tree-vect-patterns.c (vect_get_internal_def): New function.
15470 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
15471 (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
15472 (search_type_for_mask_1): Use it.
15474 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
15476 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
15477 redundant WIDEN_SUM_EXPR handling.
15478 (vect_recog_sad_pattern): Likewise.
15480 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
15482 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
15483 redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
15484 (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
15485 (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
15486 (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
15488 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
15490 * tree-vect-stmts.c (vectorizable_call): Make sure that we
15491 use the stmt_vec_info of the original bb statement for the
15492 new zero assignment, even if the call is part of a pattern.
15494 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
15496 * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
15497 that the sequence is attached to the original statement rather
15498 than the pattern statement.
15499 * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
15500 PATTERN_DEF_SEQ from the original statement rather than
15501 the main pattern statement.
15502 * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
15503 * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
15504 (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
15506 2018-06-20 Richard Sandiford <richard.sandiford@arm.com>
15508 * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
15509 definition statements before the early exit for statements that aren't
15511 * tree-vect-loop.c (vect_transform_loop_stmt): New function,
15513 (vect_transform_loop): ...here. Process pattern definition
15514 statements without first checking whether the main pattern
15515 statement is live or relevant.
15517 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
15519 * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
15520 -O0 if the locus represent UNKNOWN_LOCATION but have different values.
15522 2018-06-19 Aaron Sawdey <acsawdey@linux.ibm.com>
15524 * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
15525 TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
15526 (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
15527 (expand_block_compare): Change select_block_compare_mode call.
15528 (expand_strncmp_align_check): Use new functions, fix comment.
15529 (emit_final_str_compare_gpr): New function.
15530 (expand_strn_compare): Refactor and clean up code.
15531 * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
15533 2018-06-19 Tony Reix <tony.reix@atos.com>
15534 Damien Bergamini <damien.bergamini@atos.com>
15535 David Edelsohn <dje.gcc@gmail.com>
15537 * collect2.c (static_obj): New variable.
15538 (static_libs): New variable.
15539 (is_in_list): Uncomment declaration.
15540 (main): Track AIX libraries linked statically.
15541 (is_in_list): Uncomment definition.
15542 (scan_prog_file): Don't add AIX shared libraries initializer
15543 to constructor list if linking statically.
15545 2018-06-19 Max Filippov <jcmvbkbc@gmail.com>
15547 * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
15549 (allocate_stack, frame_blockage, *frame_blockage): New patterns.
15551 2018-06-19 Jan Hubicka <hubicka@ucw.cz>
15553 * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
15556 2018-06-19 Martin Liska <mliska@suse.cz>
15558 * config/i386/i386.c (ix86_can_inline_p): Do not use
15559 ipa_fn_summaries::get_create.
15560 * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
15562 (devirtualization_time_bonus): Likewise.
15563 (ipcp_propagate_stage): Likewise.
15564 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
15565 (edge_set_predicate): Likewise.
15566 (evaluate_conditions_for_known_args): Likewise.
15567 (evaluate_properties_for_edge): Likewise.
15568 (ipa_call_summary::reset): Tranform to ...
15569 (ipa_call_summary::~ipa_call_summary): ... this.
15570 (ipa_fn_summary::reset): Transform to ...
15571 (ipa_fn_summary::~ipa_fn_summary): ... this.
15572 (ipa_fn_summary_t::remove): Rename to ...
15573 (ipa_fn_summary_t::remove_callees): ... this.
15574 (ipa_fn_summary_t::duplicate): Use placement new
15575 instead of memory copy.
15576 (ipa_call_summary_t::duplicate): Likewise.
15577 (ipa_call_summary_t::remove): Remove.
15578 (dump_ipa_call_summary): Change get_create to get.
15579 (ipa_dump_fn_summary): Dump only when summary exists.
15580 (analyze_function_body): Use symbol_summary::get instead
15582 (compute_fn_summary): Likewise.
15583 (estimate_edge_devirt_benefit): Likewise.
15584 (estimate_edge_size_and_time): Likewise.
15585 (inline_update_callee_summaries): Likewise.
15586 (remap_edge_change_prob): Likewise.
15587 (remap_edge_summaries): Likewise.
15588 (ipa_merge_fn_summary_after_inlining): Likewise.
15589 (write_ipa_call_summary): Likewise.
15590 (ipa_fn_summary_write): Likewise.
15591 (ipa_free_fn_summary): Likewise.
15592 * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
15593 (struct ipa_call_summary): Likewise.
15594 * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
15596 * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
15597 (estimate_size_after_inlining): Likewise.
15598 (estimate_growth): Likewise.
15599 (growth_likely_positive): Likewise.
15600 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
15601 (inline_call): Likewise.
15602 * ipa-inline.c (caller_growth_limits): Likewise.
15603 (can_inline_edge_p): Likewise.
15604 (can_inline_edge_by_limits_p): Likewise.
15605 (compute_uninlined_call_time): Likewise.
15606 (compute_inlined_call_time): Likewise.
15607 (want_inline_small_function_p): Likewise.
15608 (edge_badness): Likewise.
15609 (update_caller_keys): Likewise.
15610 (update_callee_keys): Likewise.
15611 (inline_small_functions): Likewise.
15612 (inline_to_all_callers_1): Likewise.
15613 (dump_overall_stats): Likewise.
15614 (early_inline_small_functions): Likewise.
15615 (early_inliner): Likewise.
15616 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
15617 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15618 * ipa-pure-const.c (malloc_candidate_p): Likewise.
15619 * ipa-split.c (execute_split_functions): Likewise.
15620 * symbol-summary.h: Likewise.
15621 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
15623 2018-06-19 Richard Biener <rguenther@suse.de>
15625 * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
15626 (vectorize_loops): ... here. Fix dbgcnt handling.
15627 (try_vectorize_loop): Wrap try_vectorize_loop_1.
15629 2018-06-19 Segher Boessenkool <segher@kernel.crashing.org>
15632 * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
15633 ieee128 argument takes up only one (vector) register, not two (floating
15636 2018-06-19 Eric Botcazou <ebotcazou@adacore.com>
15638 * gimplify.c (gimplify_init_constructor): Really never clear for an
15639 incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
15641 2018-06-19 Richard Biener <rguenther@suse.de>
15643 PR tree-optimization/86179
15644 * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
15645 after failed recognition.
15647 2018-06-18 Martin Sebor <msebor@redhat.com>
15649 PR middle-end/85602
15650 * calls.c (maybe_warn_nonstring_arg): Handle strncat.
15651 * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
15652 Handle integer subtraction.
15653 (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
15654 * tree-ssa-strlen.h (is_strlen_related_p): Declare.
15656 2018-06-18 David Malcolm <dmalcolm@redhat.com>
15658 * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
15659 param from rtx to rtx_insn *.
15660 * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
15662 (frv_ifcvt_modify_insn): Likwise.
15663 (frv_ifcvt_modify_final): Likwise for local "existing_insn",
15664 adding an as_a <rtx_insn *> cast. Likewise for local "insn".
15665 * config/mips/mips.c (r10k_insert_cache_barriers): Add an
15666 as_a <rtx_insn *> cast to local "unprotected_region" once
15667 it's been established that it's not NULL or pc_rtx.
15668 * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
15669 param "sethi" from rtx to rtx_insn *.
15670 (nds32_group_float_insns): Likewise for param "insn".
15671 * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
15673 (vax_output_int_subtract): Likewise.
15674 * config/vax/vax.c (vax_output_int_add): Likewise for param
15676 (vax_output_int_subtract): Likewise.
15677 * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
15678 (emit_pattern_after): Likewise for param "after".
15679 (emit_insn_after): Likewise.
15680 (emit_jump_insn_after): Likewise.
15681 (emit_call_insn_after): Likewise.
15682 (emit_debug_insn_after): Likewise.
15683 (emit_pattern_before): Likewise for param "before".
15684 (emit_insn_before): Likewise.
15685 (emit_jump_insn_before): Likewise.
15686 * final.c (get_insn_template): Likewise for param "insn", removing
15688 * output.h (get_insn_template): Likewise for 2nd param.
15689 * rtl.h (emit_insn_before): Likewise.
15690 (emit_jump_insn_before): Likewise.
15691 (emit_debug_insn_before_noloc): Likewise.
15692 (emit_insn_after): Likewise.
15693 (emit_jump_insn_after): Likewise.
15694 (emit_call_insn_after): Likewise.
15695 (emit_debug_insn_after): Likewise.
15696 (set_insn_deleted): Likewise for param.
15698 2018-06-18 Michael Meissner <meissner@linux.ibm.com>
15701 * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
15702 floating point modes, so that IFmode is numerically greater than
15703 TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
15704 to declare the ordering. This prevents IFmode from being
15705 converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
15706 machine. Include rs6000-modes.h to share the fractional values
15707 between genmodes* and the rest of the compiler.
15708 (IFmode): Likewise.
15709 (KFmode): Likewise.
15710 (TFmode): Likewise.
15711 * config/rs6000/rs6000-modes.h: New file.
15712 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
15713 meaning of rs6000_long_double_size so that 126..128 selects an
15714 appropriate 128-bit floating point type.
15715 (rs6000_option_override_internal): Likewise.
15716 * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
15717 (TARGET_LONG_DOUBLE_128): Change the meaning of
15718 rs6000_long_double_size so that 126..128 selects an appropriate
15719 128-bit floating point type.
15720 (LONG_DOUBLE_TYPE_SIZE): Update comment.
15721 * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
15722 source and destination to match the standard usage.
15723 (truncifkf2): Likewise.
15724 (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
15725 ISA 2.07 to use an explicit clobber, instead of passing in a
15727 (copysign<mode>3_soft): Likewise.
15729 2018-06-18 David Malcolm <dmalcolm@redhat.com>
15731 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
15732 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
15733 (vect_slp_analyze_instance_dependence): Likewise.
15734 (vect_enhance_data_refs_alignment): Likewise.
15735 (vect_analyze_data_refs_alignment): Likewise.
15736 (vect_slp_analyze_and_verify_instance_alignment
15737 (vect_analyze_data_ref_accesses): Likewise.
15738 (vect_prune_runtime_alias_test_list): Likewise.
15739 (vect_analyze_data_refs): Likewise.
15740 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
15741 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15742 (vect_analyze_scalar_cycles_1): Likewise.
15743 (vect_get_loop_niters): Likewise.
15744 (vect_analyze_loop_form_1): Likewise.
15745 (vect_update_vf_for_slp): Likewise.
15746 (vect_analyze_loop_operations): Likewise.
15747 (vect_analyze_loop): Likewise.
15748 (vectorizable_induction): Likewise.
15749 (vect_transform_loop): Likewise.
15750 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
15751 * tree-vect-slp.c (vect_analyze_slp): Likewise.
15752 (vect_make_slp_decision): Likewise.
15753 (vect_detect_hybrid_slp): Likewise.
15754 (vect_slp_analyze_operations): Likewise.
15755 (vect_slp_bb): Likewise.
15756 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
15757 (vectorizable_bswap): Likewise.
15758 (vectorizable_call): Likewise.
15759 (vectorizable_simd_clone_call): Likewise.
15760 (vectorizable_conversion): Likewise.
15761 (vectorizable_assignment): Likewise.
15762 (vectorizable_shift): Likewise.
15763 (vectorizable_operation): Likewise.
15764 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
15766 2018-06-18 Martin Sebor <msebor@redhat.com>
15768 PR tree-optimization/81384
15769 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
15770 * builtins.c (expand_builtin_strnlen): New function.
15771 (expand_builtin): Call it.
15772 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
15773 * builtins.def (BUILT_IN_STRNLEN): New.
15774 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
15775 Warn for bounds in excess of maximum object size.
15776 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
15777 single-value ranges. Handle strnlen.
15778 (handle_builtin_strlen): Handle strnlen.
15779 (strlen_check_and_optimize_stmt): Same.
15780 * doc/extend.texi (Other Builtins): Document strnlen.
15782 2018-06-18 Maya Rashish <coypu@sdf.org>
15784 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
15785 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
15786 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
15788 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
15790 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
15792 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15794 * tree.c (escaped_string::escape): Replace cast to char * by
15795 const_cast<char *> (unescaped).
15797 2018-06-18 Nick Clifton <nickc@redhat.com>
15800 * tree.c (escaped_string): New class. Converts an unescaped
15801 string into its escaped equivalent.
15802 (warn_deprecated_use): Use the new class to convert the
15803 deprecation message, if present.
15804 (test_escaped_strings): New self test.
15805 (test_c_tests): Add test_escaped_strings.
15806 * doc/extend.texi (deprecated): Add a note that the
15807 deprecation message is affected by the -fmessage-length
15808 option, and that control characters will be escaped.
15809 (#pragma GCC error): Document this pragma.
15810 (#pragma GCC warning): Likewise.
15811 * doc/invoke.texi (-fmessage-length): Document this option's
15812 effect on the #warning and #error preprocessor directives and
15813 the deprecated attribute.
15815 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
15817 * tree.c (decl_value_expr_lookup): Revert latest change.
15818 (decl_value_expr_insert): Likewise.
15820 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
15822 * gimplify.c (nonlocal_vlas): Delete.
15823 (nonlocal_vla_vars): Likewise.
15824 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
15826 (gimplify_body): Do not create and destroy nonlocal_vlas.
15827 * tree-nested.c: Include diagnostic.h.
15828 (use_pointer_in_frame): Tweak.
15829 (lookup_field_for_decl): Add assertion and declare the transformation.
15830 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
15831 internal error when the reference is in a wrong context. Do not
15832 create a debug decl by default.
15833 (note_nonlocal_block_vlas): Delete.
15834 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
15835 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
15836 create a debug decl by default.
15837 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
15838 call is in a wrong context.
15839 (fixup_vla_decls): New function.
15840 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
15841 debug variables were created.
15842 * tree.c (decl_value_expr_lookup): Add checking assertion.
15843 (decl_value_expr_insert): Likewise.
15845 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
15847 PR middle-end/82479
15848 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
15849 * tree-scalar-evolution.c (interpret_expr): Likewise.
15850 (expression_expensive_p): Likewise.
15851 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
15852 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
15853 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
15854 (ssa_defined_by_minus_one_stmt_p): New.
15856 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
15858 PR middle-end/64946
15859 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
15860 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15861 * dojump.c (do_jump): Likewise.
15862 * expr.c (expand_expr_real_2): Check operand type's sign.
15863 * fold-const.c (const_unop): Handle ABSU_EXPR.
15864 (fold_abs_const): Likewise.
15865 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
15866 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
15867 (strip_sign_op_1): Likesise.
15868 * match.pd: Add new pattern to generate ABSU_EXPR.
15869 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
15870 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15871 * tree-eh.c (operation_could_trap_helper_p): Likewise.
15872 * tree-inline.c (estimate_operator_cost): Likewise.
15873 * tree-pretty-print.c (dump_generic_node): Likewise.
15874 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
15875 * tree.def (ABSU_EXPR): New.
15877 2018-06-16 Jakub Jelinek <jakub@redhat.com>
15879 PR middle-end/86095
15880 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
15881 documented as preserved for backward compatibility only.
15882 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
15884 PR rtl-optimization/86108
15885 * bb-reorder.c (create_forwarder_block): Renamed to ...
15886 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
15887 jump from new landing pad to the second part.
15888 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
15891 2018-06-15 Jakub Jelinek <jakub@redhat.com>
15893 PR middle-end/85878
15894 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
15895 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
15896 Only call store_expr for halves if the mode is the same.
15898 PR middle-end/86123
15899 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
15900 Fix up comment formatting.
15902 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
15904 * typed-splay-tree.h (typed_splay_tree::remove): New function.
15905 (typed_splay_tree::closure,
15906 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
15907 (typed_splay_tree::typed_splay_tree,
15908 typed_splay_tree::operator =): Declared private.
15909 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
15910 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
15911 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
15912 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
15913 typed_splay_tree::splay_tree_splay,
15914 typed_splay_tree::splay_tree_foreach_helper,
15915 typed_splay_tree::splay_tree_insert,
15916 typed_splay_tree::splay_tree_remove,
15917 typed_splay_tree::splay_tree_lookup,
15918 typed_splay_tree::splay_tree_predecessor,
15919 typed_splay_tree::splay_tree_successor,
15920 typed_splay_tree::splay_tree_min,
15921 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
15922 (typed_splay_tree::root, typed_splay_tree::comp,
15923 typed_splay_tree::delete_key,
15924 typed_splay_tree::delete_value): New data members.
15925 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
15926 typed_splay_tree::remove.
15928 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
15930 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
15931 -mginv and -mno-ginv to the assembler.
15932 * config/mips/mips.opt (-mcrc): New option.
15933 (-mginv): Likewise.
15934 * doc/invoke.text (-mcrc): Document.
15935 (-mginv): Likewise.
15937 2018-06-15 Nick Clifton <nickc@redhat.com>
15940 * tree.c (escaped_string): New class. Converts an unescaped
15941 string into its escaped equivalent.
15942 (warn_deprecated_use): Use the new class to convert the
15943 deprecation message, if present.
15944 (test_escaped_strings): New self test.
15945 (test_c_tests): Add test_escaped_strings.
15946 * doc/extend.texi (deprecated): Add a note that the
15947 deprecation message is affected by the -fmessage-length
15948 option, and that control characters will be escaped.
15949 (#pragma GCC error): Document this pragma.
15950 (#pragma GCC warning): Likewise.
15951 * doc/invoke.texi (-fmessage-length): Document this option's
15952 effect on the #warning and #error preprocessor directives and
15953 the deprecated attribute.
15955 2018-06-15 Richard Biener <rguenther@suse.de>
15957 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
15958 here, also noting vector size used.
15959 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
15960 size used in MSG_OPTIMIZED_LOCATIONS dump.
15961 (pass_slp_vectorize::execute): Adjust.
15963 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
15966 * config/arc/arc.c (arc_return_address_register): Fix
15969 2018-06-15 Richard Biener <rguenther@suse.de>
15971 PR middle-end/86159
15972 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
15973 leave useless conversion stripping to force_gimple_operand_gsi.
15974 (gimplify_build2): Likewise.
15975 (gimplify_build1): Likewise.
15977 2018-06-15 Richard Biener <rguenther@suse.de>
15979 PR middle-end/86076
15980 * tree-cfg.c (move_stmt_op): unshare invariant addresses
15981 before adjusting their block.
15983 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
15985 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
15986 multilibs for *-*-rtems*.
15987 * config/riscv/t-rtems: New file.
15989 2018-06-14 Jakub Jelinek <jakub@redhat.com>
15991 PR middle-end/86122
15992 * match.pd ((A +- CST1) +- CST2): Punt if last resort
15993 unsigned_type_for returns NULL.
15996 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
15997 subregs of multi-word pseudos unless the float mode has word size.
15999 2018-06-14 Richard Biener <rguenther@suse.de>
16001 PR middle-end/86139
16002 * tree-vect-generic.c (build_word_mode_vector_type): Remove
16003 duplicate and harmful type_hash_canon.
16004 * tree.c (type_hash_canon): Assert we didn't find ourselves.
16006 2018-06-14 Richard Biener <rguenther@suse.de>
16009 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
16012 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
16014 * config/rtems.h (STDINT_LONG32): Define.
16016 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
16017 Prachi Godbole <prachi.godbole@imgtec.com>
16019 * config/mips/mips-cpus.def: Define P6600.
16020 * config/mips/mips-tables.opt: Regenerate.
16021 * config/mips/mips.c (mips_ucbranch_type): New enum.
16022 (mips_rtx_cost_data): Add support for P6600.
16023 (mips_issue_rate): Likewise.
16024 (mips_multipass_dfa_lookahead): Likewise.
16025 (mips_avoid_hazard): Likewise.
16026 (mips_reorg_process_insns): Likewise.
16027 (mips_classify_branch_p6600): New function.
16028 * config/mips/mips.h (TUNE_P6600): New define.
16029 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
16030 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
16031 * config/mips/mips.md: Include p6600.md.
16032 (processor): Add p6600.
16033 * config/mips/p6600.md: New file.
16034 * doc/invoke.texi: Add p6600 to supported architectures.
16036 2018-06-13 Martin Sebor <msebor@redhat.com>
16038 PR tree-optimization/86114
16039 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
16041 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
16043 2018-06-13 Richard Biener <rguenther@suse.de>
16045 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
16046 Properly set vector type of the intermediate stmt.
16047 * tree-vect-stmts.c (vectorizable_operation): The destination
16048 var always has vectype_out type.
16050 2018-06-13 Jeff Law <law@redhat.com>
16052 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
16053 integer 0 for argument to print_rtl_with_bb.
16054 (rl78_reorg): Likewise.
16056 2018-06-13 David Malcolm <dmalcolm@redhat.com>
16058 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
16059 from rtx to rtx_insn *.
16060 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
16062 (add_sched_insns_for_speculation): Likewise for local "target",
16063 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
16064 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
16065 from rtx_insn ** to rtx_code_label **.
16066 (reorg_emit_nops): Likewise.
16067 (c6x_reorg): Likewise for local "call_labels".
16068 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
16070 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
16071 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
16072 the loops over LABEL_REFS.
16073 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
16075 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
16076 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
16077 (split_branches): Strengthen local "olabel" from rtx to
16078 rtx_insn *, adding a safe_as_a cast.
16079 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
16081 (add_insn_after): Likewise for first two params.
16082 (add_insn_before): Likewise.
16083 (remove_insn): Likewise for param.
16084 (emit_pattern_before_noloc): Likewise for second and third params.
16085 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
16086 (emit_call_insn_before_noloc): Likewise.
16087 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
16089 (emit_barrier_before): Likewise.
16090 (emit_label_before): Strengthen "label" param from "rtx" to
16091 "rtx_code_label *". Strengthen "before" param from "rtx" to
16093 (emit_insn_after_1): Strengthen "after" param from "rtx" to
16095 (emit_pattern_after_noloc): Likewise.
16096 (emit_insn_after_noloc): Likewise.
16097 (emit_jump_insn_after_noloc): Likewise.
16098 (emit_call_insn_after_noloc): Likewise.
16099 (emit_debug_insn_after_noloc): Likewise.
16100 (emit_barrier_after): Likewise.
16101 (emit_label_after): Likewise for both params.
16102 (emit_pattern_after_setloc): Likewise for "after" param. Convert
16103 "loc" param from "int" to "location_t".
16104 (emit_insn_after_setloc): Likewise.
16105 (emit_jump_insn_after_setloc): Likewise.
16106 (emit_call_insn_after_setloc): Likewise.
16107 (emit_debug_insn_after_setloc): Likewise.
16108 (emit_pattern_before_setloc): Likewise for "before" param. Convert
16109 "loc" param from "int" to "location_t".
16110 (emit_pattern_before): Convert NULL_RTX to NULL.
16111 (emit_insn_before_setloc): Convert "loc" param from "int" to
16113 (emit_jump_insn_before_setloc): Likewise.
16114 (emit_call_insn_before_setloc): Likewise.
16115 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
16116 rtx_insn *. Convert "loc" param from "int" to "location_t".
16117 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
16118 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
16119 Convert 3rd param from "int" to "location_t".
16120 (emit_barrier_before, emit_barrier_after, next_real_insn):
16121 Strengthen param from rtx to rtx_insn *.
16122 (emit_label_before): Strengthen 1st param from "rtx" to
16123 "rtx_code_label *". Strengthen 2nd param from "rtx" to
16125 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
16126 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
16127 Strengthen 2nd param from "rtx" to "rtx_insn *".
16128 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
16129 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
16130 Likewise. Convert 3rd param from "int" to "location_t".
16131 (emit_label_after): Strengthen 1st param from "rtx" to
16132 "rtx_code_label *".
16133 (next_real_insn, remove_insn): Strengthen param from "rtx" to
16135 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
16136 from "rtx" to "rtx_insn *".
16138 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
16140 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
16141 bodies streamed in with -Q.
16142 * dumpfile.c (dump_files): Add lto-stream-out dump file.
16143 * dumpfile.h (tree_dump_index): Add lto_stream_out.
16144 * gimple-streamer-out.c: Include gimple-pretty-print.h
16145 (output_bb): Dump stmts streamed.
16146 * lto-section-out.c: Include print-tree.h
16147 (lto_begin_section): Dump sections created.
16148 (lto_output_decl_index): Dump decl encoded.
16149 * lto-streamer-out.c: Include print-tree.h
16150 (create_output_block): Dump output block created.
16151 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
16152 (output_function): Dump function output.
16153 (output_constructor): Dump constructor streamed.
16154 (write_global_stream): Output indexes encoded.
16155 (produce_asm_for_decls): Dump streams encoded.
16156 * lto-streamer.c (streamer_dump_file): New global var.
16157 * lto-streamer.h (streamer_dump_file): Declare.
16158 * passes.c (ipa_write_summaries): Initialize streamer dump.
16159 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
16162 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
16165 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
16166 offsets for register save directives. Emit a second batch of save
16167 directives, if need be, when the function accesses prior frames.
16169 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
16171 * config/arc/fpu.md (fmasf4): Force operand to register.
16172 (fnmasf4): Likewise.
16174 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
16176 * config/arc/arc-protos.h (arc_pad_return): Remove.
16177 * config/arc/arc.c (machine_function): Remove force_short_suffix
16179 (arc_print_operand): Adjust printing of '&'.
16180 (arc_verify_short): Remove conditional printing of short suffix.
16181 (arc_final_prescan_insn): Remove reference to size_reason.
16182 (pad_return): New function.
16183 (arc_reorg): Call pad_return.
16184 (arc_pad_return): Remove.
16185 (arc_init_machine_status): Remove reference to force_short_suffix.
16186 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
16187 (attr length): When attribute iscompact is true force to 2
16188 regardless; in the case of maybe check if we want to force the
16189 instruction to have 4 bytes length.
16190 (nopv): Change it to generate 4 byte long nop as well.
16191 (blockage): New pattern.
16192 (simple_return): Remove call to arc_pad_return.
16193 (p_return_i): Likewise.
16195 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
16197 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
16199 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
16201 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
16204 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
16206 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
16207 for ARC700 and ARCv2.
16209 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
16212 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
16213 operands[2] instead of operands[1].
16216 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
16218 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
16219 case, check whether the outer register overlaps an unallocatable
16220 register, not just whether it fits the required class.
16222 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
16224 * poly-int.h (can_div_trunc_p): Add new overload in which all values
16226 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
16227 (memrefs_conflict_p): Likewise.
16228 (init_alias_analysis): Likewise.
16229 * cfgexpand.c (expand_debug_expr): Likewise.
16230 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
16231 * cse.c (fold_rtx): Likewise.
16232 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
16233 * expr.c (emit_block_move_hints): Likewise.
16234 (clear_storage_hints, push_block, emit_push_insn): Likewise.
16235 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
16236 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
16237 (emit_group_store): Likewise.
16238 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
16239 to read the PRE/POST_MODIFY increment.
16240 * calls.c (store_one_arg): Use strip_offset.
16241 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
16243 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
16245 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
16246 (simplify_binary_operation_1): Extend CONST_INT handling to
16248 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
16249 than a HOST_WIDE_INT.
16250 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
16252 (adjust_mems, add_stores): Update accodingly.
16253 (vt_canonicalize_addr): Track polynomial offsets.
16254 (emit_note_insn_var_location): Likewise.
16255 (vt_add_function_parameter): Likewise.
16256 (vt_initialize): Likewise.
16258 2018-06-12 Jeff Law <law@redhat.com>
16260 * config.gcc (alpha*-*-freebsd*): Remove.
16261 * config/alpha/freebsd.h: Remove.
16263 2018-06-12 David Malcolm <dmalcolm@redhat.com>
16266 * spellcheck-tree.c (levenshtein_distance): Rename to...
16267 (get_edit_distance): ...this, and update for underlying renaming.
16268 * spellcheck-tree.h (levenshtein_distance): Rename to...
16269 (get_edit_distance): ...this.
16270 * spellcheck.c (levenshtein_distance): Rename to...
16271 (get_edit_distance): ...this. Convert from Levenshtein distance
16272 to Damerau-Levenshtein distance by supporting transpositions of
16273 adjacent characters. Rename "v1" to "v_next" and "v0" to
16275 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
16276 (selftest::test_edit_distance_unit_test_oneway): ...this, and
16277 update for underlying renaming.
16278 (selftest::levenshtein_distance_unit_test): Rename to...
16279 (selftest::test_get_edit_distance_unit): ...this, and update for
16280 underlying renaming.
16281 (selftest::test_find_closest_string): Add example from PR 69968
16282 where transposition helps
16283 (selftest::test_metric_conditions): Update for renaming.
16284 (selftest::test_metric_conditions): Likewise.
16285 (selftest::spellcheck_c_tests): Likewise.
16286 * spellcheck.h (levenshtein_distance): Rename both overloads to...
16287 (get_edit_distance): ...this.
16288 (best_match::consider): Update for renaming.
16290 2018-06-12 Martin Sebor <msebor@redhat.com>
16292 PR tree-optimization/85259
16293 * builtins.c (compute_objsize): Handle constant offsets.
16294 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
16295 true iff a warning has been issued.
16296 * gimple.h (gimple_nonartificial_location): New function.
16297 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
16298 gimple_nonartificial_location and handle -Wno-system-headers.
16299 (handle_builtin_stxncpy): Same.
16301 2018-06-12 Martin Sebor <msebor@redhat.com>
16304 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
16306 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
16308 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16309 BUILTIN_VEC_XST entries for pointer to double and long long.
16311 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
16314 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
16316 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
16319 2018-06-12 Martin Liska <mliska@suse.cz>
16321 * doc/options.texi: Document IntegerRange.
16323 2018-06-12 Martin Liska <mliska@suse.cz>
16325 * config/i386/i386.opt: Make MPX-related options as Deprecated.
16326 * opt-functions.awk: Handle Deprecated flag.
16327 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
16329 (read_cmdline_option): Report warning for a deprecated option.
16330 * opts.h (struct cl_option): Add new field cl_deprecated.
16331 (CL_ERR_DEPRECATED): New.
16333 2018-06-12 Martin Liska <mliska@suse.cz>
16335 * doc/options.texi: Document Deprecated option flag.
16337 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
16339 * config/arc/arc-arch.h (arc_extras): New enum.
16340 (arc_cpu_t):Add field extra.
16341 (arc_cpu_types): Consider the extras.
16342 * config/arc/arc-cpus.def: Add extras info.
16343 * config/arc/arc-opts.h (processor_type): Consider extra field.
16344 * config/arc/arc.c (arc_override_options): Handle extra field.
16346 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
16348 * config/arc/arc-arch.h: Update ARC_OPTX macro.
16349 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
16351 * config/arc/arc.c (arc_init): Update pic warning.
16352 (irq_range): Update irq range parsing warnings.
16353 (arc_override_options): Update various warning messages.
16354 (arc_handle_aux_attribute): Likewise.
16356 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
16358 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
16360 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
16362 * doc/sourcebuild.texi: Document usage of line number 0 in verify
16363 compiler messages directives.
16365 2018-06-12 Matthew Fortune <mfortune@gmail.com>
16367 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
16368 * config/mips/mips-tables.opt: Regenerate.
16369 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
16371 * doc/invoke.texi: Document -march=i6500.
16373 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
16375 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
16376 (i6400_gpmul): Add cpu_unit.
16377 (i6400_gpdiv): Likewise.
16378 (i6400_msa_add_d): Update reservations.
16379 (i6400_msa_int_add) Likewise.
16380 (i6400_msa_short_logic3) Likewise.
16381 (i6400_msa_short_logic2) Likewise.
16382 (i6400_msa_short_logic) Likewise.
16383 (i6400_msa_move) Likewise.
16384 (i6400_msa_cmp) Likewise.
16385 (i6400_msa_short_float2) Likewise.
16386 (i6400_msa_div_d) Likewise.
16387 (i6400_msa_long_logic1) Likewise.
16388 (i6400_msa_long_logic2) Likewise.
16389 (i6400_msa_mult) Likewise.
16390 (i6400_msa_long_float2) Likewise.
16391 (i6400_msa_long_float4) Likewise.
16392 (i6400_msa_long_float5) Likewise.
16393 (i6400_msa_long_float8) Likewise.
16394 (i6400_fpu_fadd): Include frint type.
16395 (i6400_fpu_store): New define_insn_reservation.
16396 (i6400_fpu_load): Likewise.
16397 (i6400_fpu_move): Likewise.
16398 (i6400_fpu_fcmp): Likewise.
16399 (i6400_fpu_fmadd): Likewise.
16400 (i6400_int_mult): Include imul3nc type and update reservation.
16401 (i6400_int_div): Include idiv3 type and update reservation.
16402 (i6400_int_load): Update to check type not move_type.
16403 (i6400_int_store): Likewise.
16404 (i6400_int_prefetch): Set zero latency.
16406 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
16408 * gcc.c: Document new %@{...} sequence.
16409 (LINK_COMMAND_SPEC): Use it for the -L switches.
16410 (cpp_unique_options): Use it for the -I switches.
16411 (at_file_argbuf): New global variable.
16412 (in_at_file): Likewise.
16413 (alloc_args): Create at_file_argbuf.
16414 (clear_args): Truncate at_file_argbuf.
16415 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
16416 (open_at_file): New function.
16417 (close_at_file): Likewise.
16418 (create_at_file): Delete.
16419 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
16422 (validate_switches_from_spec): Deal with %@{...} sequence.
16423 (validate_switches): Likewise.
16424 (driver::finalize): Call clear_args.
16426 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
16428 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
16430 2018-06-11 Martin Sebor <msebor@redhat.com>
16432 * doc/invoke.texi (-Wall): List -Wc++17-compat.
16433 (Wno-class-memaccess): Add @opindex.
16434 (Wno-templates, Wno-multiple-inheritance): Same.
16435 (Wno-virtual-inheritance, Wno-namespaces): Same.
16436 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
16437 (Wno-format-overflow, Wno-format-truncation): Same.
16438 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
16439 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
16440 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
16441 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
16442 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
16443 (Wno-misspelled-isr): Same.
16445 2018-06-11 Martin Sebor <msebor@redhat.com>
16447 * PR tree-optimization/86083
16448 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
16450 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
16452 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
16454 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
16457 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
16458 on the correct operand.
16459 (*movdi_internal64): Ditto.
16461 2018-06-11 Martin Liska <mliska@suse.cz>
16463 PR tree-optimization/86089
16464 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
16466 2018-06-11 Julia Koval <julia.koval@intel.com>
16468 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
16469 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
16470 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
16472 2018-06-11 Olivier Hainque <hainque@adacore.com>
16474 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
16475 for Ada with strict dwarf2.
16477 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
16480 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
16483 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
16485 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
16487 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
16489 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
16492 2018-06-08 Martin Liska <mliska@suse.cz>
16494 * tree-cfg.h (debug_function): Fix argument type to match
16497 2018-06-08 Martin Liska <mliska@suse.cz>
16499 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
16500 Remove usage of MPX-related (and removed) fields.
16501 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
16503 2018-06-08 David Malcolm <dmalcolm@redhat.com>
16505 * cfg.c (debug): Use TDF_NONE rather than 0.
16506 * cfghooks.c (debug): Likewise.
16507 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
16508 (struct dump_option_value_info): Convert to...
16509 (struct kv_pair): ...this template type.
16510 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
16512 (optinfo_verbosity_options): Likewise.
16513 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
16515 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
16516 than int for "optgroup_flags" param.
16517 (dump_generic_expr_loc): Use dump_flags_t rather than int for
16519 (dump_dec): Likewise.
16520 (dump_finish): Use TDF_NONE rather than 0.
16521 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
16522 rather than int for "optgroup_flags" param. Use TDF_NONE rather
16523 than 0. Update for change to option_ptr.
16524 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
16525 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
16526 0. Update for changes to optinfo_verbosity_options and
16528 (opt_info_switch_p): Convert optgroup_flags from int to
16530 (dump_basic_block): Use dump_flags_t rather than int
16531 for "dump_kind" param.
16532 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
16533 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
16534 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
16535 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
16536 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
16537 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
16538 TDF_NONE): Convert from macros to...
16539 (enum dump_flag): ...this new enum.
16540 (dump_flags_t): Update to use enum.
16541 (operator|, operator&, operator~, operator|=, operator&=):
16542 Implement for dump_flags_t.
16543 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
16544 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
16545 Convert from macros to...
16546 (enum optgroup_flag): ...this new enum.
16547 (optgroup_flags_t): New typedef.
16548 (operator|, operator|=): Implement for optgroup_flags_t.
16549 (struct dump_file_info): Convert field "alt_flags" to
16550 dump_flags_t. Convert field "optgroup_flags" to
16552 (dump_basic_block): Use dump_flags_t rather than int for param.
16553 (dump_generic_expr_loc): Likewise.
16554 (dump_dec): Likewise.
16555 (dump_register): Convert param "optgroup_flags" to
16557 (opt_info_enable_passes): Likewise.
16558 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
16560 * gimple-pretty-print.c (debug): Likewise.
16561 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
16562 (merged_store_group::apply_stores): Likewise.
16563 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
16564 * gimple.c (verify_gimple_pp): Likewise.
16565 * graphite-poly.c (print_pbb_body): Likewise.
16566 * passes.c (pass_manager::register_one_dump_file): Convert
16567 local "optgroup_flags" to optgroup_flags_t.
16568 * print-tree.c (print_node): Use TDF_NONE rather than 0.
16570 (debug_body): Likewise.
16571 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
16572 to optgroup_flags_t.
16573 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
16575 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
16576 (convert_mult_to_fma): Likewise.
16577 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
16578 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
16579 * tree-vect-data-refs.c (dump_lower_bound): Convert param
16580 "dump_kind" to dump_flags_t.
16582 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
16584 * config/rs6000/rs6000.c (min, max): Delete.
16586 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
16588 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
16589 -mabi=spe and -mabi=no-spe.
16591 2018-06-08 Martin Liska <mliska@suse.cz>
16593 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
16594 where we expect an existing summary.
16596 2018-06-08 Martin Liska <mliska@suse.cz>
16598 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
16599 * ipa-inline.h (estimate_edge_growth): Likewise.
16601 2018-06-08 Martin Liska <mliska@suse.cz>
16603 * cgraph.c (function_version_hasher::hash): Use
16604 cgraph_node::get_uid ().
16605 (function_version_hasher::equal):
16606 * cgraph.h (cgraph_node::get_uid): New method.
16607 * ipa-inline.c (update_caller_keys): Use
16608 cgraph_node::get_uid ().
16609 (update_callee_keys): Likewise.
16610 * ipa-utils.c (searchc): Likewise.
16611 (ipa_reduced_postorder): Likewise.
16612 * lto-cgraph.c (input_node): Likewise.
16613 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
16614 * symbol-summary.h (symtab_insertion): Likewise.
16615 (symtab_removal): Likewise.
16616 (symtab_duplication): Likewise.
16617 * tree-pretty-print.c (dump_function_header): Likewise.
16618 * tree-sra.c (convert_callers_for_node): Likewise.
16620 2018-06-08 Martin Liska <mliska@suse.cz>
16622 * cgraph.c (symbol_table::create_edge): Always assign a new
16624 (symbol_table::free_edge): Do not recycle numbers.
16625 * cgraph.h (cgraph_edge::get): New method.
16626 * symbol-summary.h (symtab_removal): Use it.
16627 (symtab_duplication): Likewise.
16628 (call_summary::hashable_uid): Remove.
16630 2018-06-08 Martin Liska <mliska@suse.cz>
16632 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
16633 (initialize_growth_caches): Remove.
16634 (free_growth_caches): Likewise.
16635 (do_estimate_edge_time): Use edge_growth_cache.
16636 (do_estimate_edge_size): Likewise.
16637 (do_estimate_edge_hints): Likewise.
16638 * ipa-inline.c (reset_edge_caches): Likewise.
16639 (recursive_inlining): Likewise.
16640 (inline_small_functions): Likewise.
16641 * ipa-inline.h (initialize_growth_caches): Remove.
16642 (estimate_edge_size): Likewise.
16643 (estimate_edge_time): Likewise.
16644 (estimate_edge_hints): Likewise.
16645 (reset_edge_growth_cache): Likewise.
16646 * symbol-summary.h (call_summary::remove): New method.
16648 2018-06-08 Martin Liska <mliska@suse.cz>
16650 * ipa-cp.c (class edge_clone_summary): New summary.
16651 (grow_edge_clone_vectors): Remove.
16652 (ipcp_edge_duplication_hook): Remove.
16653 (class edge_clone_summary_t): New call_summary class.
16654 (ipcp_edge_removal_hook): Remove.
16655 (edge_clone_summary_t::duplicate): New function.
16656 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
16657 (create_specialized_node): Likewise.
16658 (ipcp_driver): Initialize edge_clone_summaries and do not
16661 2018-06-08 Martin Liska <mliska@suse.cz>
16663 * symbol-summary.h (get): New function.
16664 (call_summary::m_initialize_when_cloning): New class member.
16666 2018-06-08 Martin Liska <mliska@suse.cz>
16668 * cgraph.c (cgraph_node::remove): Do not recycle uid.
16669 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
16670 (symbol_table::allocate_cgraph_symbol): Do not set uid.
16671 * passes.c (uid_hash_t): Record removed_nodes by their uids.
16672 (remove_cgraph_node_from_order): Use the removed_nodes set.
16673 (do_per_function_toporder): Likwise.
16674 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
16675 instead of summary_uid.
16676 (symtab_removal): Likewise.
16677 (symtab_duplication): Likewise.
16679 2018-06-08 Martin Liska <mliska@suse.cz>
16681 * ipa-cp.c (ipcp_store_bits_results): Use
16682 ipcp_transformation_sum.
16683 (ipcp_store_vr_results): Likewise.
16684 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
16686 (ipcp_transformation_initialize): ... this.
16687 (ipa_set_node_agg_value_chain):
16688 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
16689 (write_ipcp_transformation_info): Likewise.
16690 (read_ipcp_transformation_info): Likewise.
16691 (ipcp_update_bits): Likewise.
16692 (ipcp_update_vr): Likewise.
16693 (ipcp_transform_function): Likewise.
16694 * ipa-prop.h: Rename ipcp_transformation_summary to
16695 ipcp_transformation.
16696 (class ipcp_transformation_t): New function summary.
16697 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
16698 (ipa_get_agg_replacements_for_node): Likewise.
16700 2018-06-08 Martin Liska <mliska@suse.cz>
16702 * ipa-pure-const.c (struct funct_state_d): Do it class instead
16704 (class funct_state_summary_t): New function_summary class.
16705 (has_function_state): Remove.
16706 (get_function_state): Likewise.
16707 (set_function_state): Likewise.
16708 (add_new_function): Likewise.
16709 (funct_state_summary_t::insert): New function.
16710 (duplicate_node_data): Remove.
16711 (remove_node_data): Remove.
16712 (funct_state_summary_t::duplicate): New function.
16713 (register_hooks): Create new funct_state_summaries.
16714 (pure_const_generate_summary): Use it.
16715 (pure_const_write_summary): Likewise.
16716 (pure_const_read_summary): Likewise.
16717 (propagate_pure_const): Likewise.
16718 (propagate_nothrow): Likewise.
16719 (dump_malloc_lattice): Likewise.
16720 (propagate_malloc): Likewise.
16721 (execute): Do not register hooks, just remove summary
16723 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
16726 2018-06-08 Martin Liska <mliska@suse.cz>
16728 * ipa-reference.c (remove_node_data): Remove.
16729 (duplicate_node_data): Likewise.
16730 (class ipa_ref_var_info_summary_t): New class.
16731 (class ipa_ref_opt_summary_t): Likewise.
16732 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
16733 (get_reference_optimization_summary): Use
16734 ipa_ref_opt_sum_summaries.
16735 (set_reference_vars_info): Remove.
16736 (set_reference_optimization_summary): Likewise.
16737 (ipa_init): Create summaries.
16738 (init_function_info): Use function summary.
16739 (ipa_ref_opt_summary_t::duplicate): New function.
16740 (ipa_ref_opt_summary_t::remove): New function.
16741 (get_read_write_all_from_node): Fix GNU coding style.
16742 (propagate): Use function summary.
16743 (write_node_summary_p): Fix GNU coding style.
16744 (stream_out_bitmap): Likewise.
16745 (ipa_reference_read_optimization_summary): Use function summary.
16746 (ipa_reference_c_finalize): Do not release hooks.
16748 2018-06-08 Martin Liska <mliska@suse.cz>
16750 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
16751 (analyze_function_body): Extract multiple calls of get_create.
16752 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16753 * ipa-inline.c (recursive_inlining): Use ::get method.
16754 * ipa-inline.h (estimate_edge_growth): Likewise.
16756 2018-06-08 Martin Liska <mliska@suse.cz>
16758 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
16760 (hsa_function_summary::hsa_function_summary): Use the new enum
16762 (hsa_gpu_implementation_p): Use hsa_summaries::get.
16763 * hsa-gen.c (hsa_get_host_function): Likewise.
16764 (get_brig_function_name): Likewise.
16765 * ipa-hsa.c (process_hsa_functions): Likewise.
16766 (ipa_hsa_write_summary): Likewise.
16767 * symbol-summary.h (symtab_duplication): Use ::get function/
16768 (get): New function.
16770 2018-06-08 Martin Liska <mliska@suse.cz>
16772 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
16774 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
16775 (hsa_register_kernel): Likewise.
16776 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
16777 * hsa-gen.c (hsa_get_host_function): Likewise.
16778 (get_brig_function_name): Likewise.
16779 (generate_hsa): Likewise.
16780 (pass_gen_hsail::execute): Likewise.
16781 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
16782 (devirtualization_time_bonus): Likewise.
16783 (ipcp_propagate_stage): Likewise.
16784 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
16785 (edge_set_predicate): Likewise.
16786 (evaluate_conditions_for_known_args): Likewise.
16787 (evaluate_properties_for_edge): Likewise.
16788 (ipa_fn_summary::reset): Likewise.
16789 (ipa_fn_summary_t::duplicate): Likewise.
16790 (dump_ipa_call_summary): Likewise.
16791 (ipa_dump_fn_summary): Likewise.
16792 (analyze_function_body): Likewise.
16793 (compute_fn_summary): Likewise.
16794 (estimate_edge_devirt_benefit): Likewise.
16795 (estimate_edge_size_and_time): Likewise.
16796 (estimate_calls_size_and_time): Likewise.
16797 (estimate_node_size_and_time): Likewise.
16798 (inline_update_callee_summaries): Likewise.
16799 (remap_edge_change_prob): Likewise.
16800 (remap_edge_summaries): Likewise.
16801 (ipa_merge_fn_summary_after_inlining): Likewise.
16802 (ipa_update_overall_fn_summary): Likewise.
16803 (read_ipa_call_summary): Likewise.
16804 (inline_read_section): Likewise.
16805 (write_ipa_call_summary): Likewise.
16806 (ipa_fn_summary_write): Likewise.
16807 (ipa_free_fn_summary): Likewise.
16808 * ipa-hsa.c (process_hsa_functions): Likewise.
16809 (ipa_hsa_write_summary): Likewise.
16810 (ipa_hsa_read_section): Likewise.
16811 * ipa-icf.c (sem_function::merge): Likewise.
16812 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16813 (do_estimate_edge_time): Likewise.
16814 (estimate_size_after_inlining): Likewise.
16815 (estimate_growth): Likewise.
16816 (growth_likely_positive): Likewise.
16817 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
16818 (inline_call): Likewise.
16819 * ipa-inline.c (caller_growth_limits): Likewise.
16820 (can_inline_edge_p): Likewise.
16821 (can_inline_edge_by_limits_p): Likewise.
16822 (compute_uninlined_call_time): Likewise.
16823 (compute_inlined_call_time): Likewise.
16824 (want_inline_small_function_p): Likewise.
16825 (edge_badness): Likewise.
16826 (update_caller_keys): Likewise.
16827 (update_callee_keys): Likewise.
16828 (recursive_inlining): Likewise.
16829 (inline_small_functions): Likewise.
16830 (inline_to_all_callers_1): Likewise.
16831 (dump_overall_stats): Likewise.
16832 (early_inline_small_functions): Likewise.
16833 (early_inliner): Likewise.
16834 * ipa-inline.h (estimate_edge_growth): Likewise.
16835 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
16836 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
16837 * ipa-prop.h (IPA_NODE_REF): Likewise.
16838 (IPA_EDGE_REF): Likewise.
16839 * ipa-pure-const.c (malloc_candidate_p): Likewise.
16840 (propagate_malloc): Likewise.
16841 * ipa-split.c (execute_split_functions): Likewise.
16842 * symbol-summary.h: Rename get to get_create.
16844 (get_create): Likewise.
16845 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
16847 2018-06-08 Martin Liska <mliska@suse.cz>
16849 * symbol-summary.h (release): Move definition out of class
16851 (symtab_removal): Likewise.
16852 (symtab_duplication): Likewise.
16854 2018-06-08 Martin Liska <mliska@suse.cz>
16856 * symbol-summary.h (function_summary): Move constructor
16857 implementation out of class declaration.
16858 (release): Likewise.
16859 (symtab_insertion): Likewise.
16860 (symtab_removal): Likewise.
16861 (symtab_duplication): Likewise.
16864 2018-06-08 Martin Liska <mliska@suse.cz>
16866 * Makefile.in: Remove support for MPX (macros, related functions,
16867 fields in cgraph_node, ...).
16868 * builtin-types.def (BT_BND): Likewise.
16869 (BT_FN_BND_CONST_PTR): Likewise.
16870 (BT_FN_CONST_PTR_BND): Likewise.
16871 (BT_FN_VOID_PTR_BND): Likewise.
16872 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
16873 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
16874 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
16875 (expand_builtin_mempcpy_with_bounds): Likewise.
16876 (expand_builtin_memset_with_bounds): Likewise.
16877 (expand_builtin_memset_args): Likewise.
16878 (std_expand_builtin_va_start): Likewise.
16879 (expand_builtin): Likewise.
16880 (expand_builtin_with_bounds): Likewise.
16881 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
16882 (DEF_LIB_BUILTIN_CHKP): Likewise.
16883 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
16884 (DEF_CHKP_BUILTIN): Likewise.
16885 (BUILT_IN_MEMCPY): Likewise.
16886 (BUILT_IN_MEMMOVE): Likewise.
16887 (BUILT_IN_MEMPCPY): Likewise.
16888 (BUILT_IN_MEMSET): Likewise.
16889 (BUILT_IN_STPCPY): Likewise.
16890 (BUILT_IN_STRCAT): Likewise.
16891 (BUILT_IN_STRCHR): Likewise.
16892 (BUILT_IN_STRCPY): Likewise.
16893 (BUILT_IN_STRLEN): Likewise.
16894 (BUILT_IN_MEMCPY_CHK): Likewise.
16895 (BUILT_IN_MEMMOVE_CHK): Likewise.
16896 (BUILT_IN_MEMPCPY_CHK): Likewise.
16897 (BUILT_IN_MEMSET_CHK): Likewise.
16898 (BUILT_IN_STPCPY_CHK): Likewise.
16899 (BUILT_IN_STRCAT_CHK): Likewise.
16900 (BUILT_IN_STRCPY_CHK): Likewise.
16901 * calls.c (store_bounds): Likewise.
16902 (emit_call_1): Likewise.
16903 (special_function_p): Likewise.
16904 (maybe_warn_nonstring_arg): Likewise.
16905 (initialize_argument_information): Likewise.
16906 (finalize_must_preallocate): Likewise.
16907 (compute_argument_addresses): Likewise.
16908 (expand_call): Likewise.
16909 * cfgexpand.c (expand_call_stmt): Likewise.
16910 (expand_return): Likewise.
16911 (expand_gimple_stmt_1): Likewise.
16912 (pass_expand::execute): Likewise.
16913 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
16914 (cgraph_node::remove): Likewise.
16915 (cgraph_node::dump): Likewise.
16916 (cgraph_node::verify_node): Likewise.
16917 * cgraph.h (chkp_function_instrumented_p): Likewise.
16918 (symtab_node::get_alias_target): Likewise.
16919 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
16920 (cgraph_local_p): Likewise.
16921 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
16922 (cgraph_edge::rebuild_references): Likewise.
16923 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
16924 (walk_polymorphic_call_targets): Likewise.
16925 (cgraph_node::expand_thunk): Likewise.
16926 (symbol_table::output_weakrefs): Likewise.
16927 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
16928 (ix86_handle_option): Likewise.
16929 * config/i386/constraints.md: Likewise.
16930 * config/i386/i386-builtin-types.def (BND): Likewise.
16934 * config/i386/i386-builtin.def (BDESC_END): Likewise.
16935 (BDESC_FIRST): Likewise.
16937 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
16938 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
16939 * config/i386/i386.c (enum reg_class): Likewise.
16940 (ix86_target_string): Likewise.
16941 (ix86_option_override_internal): Likewise.
16942 (ix86_conditional_register_usage): Likewise.
16943 (ix86_valid_target_attribute_inner_p): Likewise.
16944 (ix86_set_indirect_branch_type): Likewise.
16945 (ix86_set_current_function): Likewise.
16946 (ix86_function_arg_regno_p): Likewise.
16947 (init_cumulative_args): Likewise.
16948 (ix86_function_arg_advance): Likewise.
16949 (ix86_function_arg): Likewise.
16950 (ix86_pass_by_reference): Likewise.
16951 (ix86_function_value_regno_p): Likewise.
16952 (ix86_function_value_1): Likewise.
16953 (ix86_function_value_bounds): Likewise.
16954 (ix86_return_in_memory): Likewise.
16955 (ix86_setup_incoming_vararg_bounds): Likewise.
16956 (ix86_va_start): Likewise.
16957 (indirect_thunk_need_prefix): Likewise.
16958 (print_reg): Likewise.
16959 (ix86_print_operand): Likewise.
16960 (ix86_expand_call): Likewise.
16961 (ix86_output_function_return): Likewise.
16962 (reg_encoded_number): Likewise.
16963 (BDESC_VERIFYS): Likewise.
16964 (ix86_init_mpx_builtins): Likewise.
16965 (ix86_init_builtins): Likewise.
16966 (ix86_emit_cmove): Likewise.
16967 (ix86_emit_move_max): Likewise.
16968 (ix86_expand_builtin): Likewise.
16969 (ix86_builtin_mpx_function): Likewise.
16970 (ix86_get_arg_address_for_bt): Likewise.
16971 (ix86_load_bounds): Likewise.
16972 (ix86_store_bounds): Likewise.
16973 (ix86_load_returned_bounds): Likewise.
16974 (ix86_store_returned_bounds): Likewise.
16975 (ix86_class_likely_spilled_p): Likewise.
16976 (ix86_hard_regno_mode_ok): Likewise.
16977 (x86_order_regs_for_local_alloc): Likewise.
16978 (ix86_mitigate_rop): Likewise.
16979 (ix86_bnd_prefixed_insn_p): Likewise.
16980 (ix86_mpx_bound_mode): Likewise.
16981 (ix86_make_bounds_constant): Likewise.
16982 (ix86_initialize_bounds): Likewise.
16983 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
16984 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
16985 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
16986 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
16987 (TARGET_CHKP_BOUND_MODE): Likewise.
16988 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
16989 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
16990 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
16991 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
16992 * config/i386/i386.h (TARGET_MPX): Likewise.
16993 (TARGET_MPX_P): Likewise.
16994 (VALID_BND_REG_MODE): Likewise.
16995 (FIRST_BND_REG): Likewise.
16996 (LAST_BND_REG): Likewise.
16997 (enum reg_class): Likewise.
16998 (BND_REG_P): Likewise.
16999 (BND_REGNO_P): Likewise.
17000 (BNDmode): Likewise.
17001 (ADJUST_INSN_LENGTH): Likewise.
17002 * config/i386/i386.md: Likewise.
17003 * config/i386/i386.opt: Likewise.
17004 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
17005 (defined): Likewise.
17006 (LINK_MPX): Likewise.
17007 (MPX_SPEC): Likewise.
17008 (LIBMPX_SPEC): Likewise.
17009 (LIBMPXWRAPPERS_SPEC): Likewise.
17010 (CHKP_SPEC): Likewise.
17011 * config/i386/predicates.md: Likewise.
17012 * dbxout.c (dbxout_type): Likewise.
17013 * doc/extend.texi: Likewise.
17014 * doc/invoke.texi: Likewise.
17015 * doc/md.texi: Likewise.
17016 * doc/tm.texi: Likewise.
17017 * doc/tm.texi.in: Likewise.
17018 * dwarf2out.c (is_base_type): Likewise.
17019 (gen_formal_types_die): Likewise.
17020 (gen_subprogram_die): Likewise.
17021 (gen_type_die_with_usage): Likewise.
17022 (gen_decl_die): Likewise.
17023 (dwarf2out_late_global_decl): Likewise.
17024 * expr.c (expand_assignment): Likewise.
17025 (emit_storent_insn): Likewise.
17026 (store_expr_with_bounds): Likewise.
17027 (store_expr): Likewise.
17028 (expand_expr_real_1): Likewise.
17029 * expr.h (store_expr_with_bounds): Likewise.
17030 * function.c (use_register_for_decl): Likewise.
17031 (struct bounds_parm_data): Likewise.
17032 (assign_parms_augmented_arg_list): Likewise.
17033 (assign_parm_find_entry_rtl): Likewise.
17034 (assign_parm_is_stack_parm): Likewise.
17035 (assign_parm_load_bounds): Likewise.
17036 (assign_bounds): Likewise.
17037 (assign_parms): Likewise.
17038 (expand_function_start): Likewise.
17039 * gcc.c (CHKP_SPEC): Likewise.
17040 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
17041 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
17042 (wrestrict_dom_walker::check_call): Likewise.
17043 * gimple.c (gimple_build_call_from_tree): Likewise.
17044 * gimple.h (enum gf_mask): Likewise.
17045 (gimple_call_with_bounds_p): Likewise.
17046 (gimple_call_set_with_bounds): Likewise.
17047 * gimplify.c (gimplify_init_constructor): Likewise.
17048 * ipa-cp.c (initialize_node_lattices): Likewise.
17049 (propagate_constants_across_call): Likewise.
17050 (find_more_scalar_values_for_callers_subset): Likewise.
17051 * ipa-hsa.c (process_hsa_functions): Likewise.
17052 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
17053 * ipa-icf.c (sem_function::merge): Likewise.
17054 * ipa-inline.c (early_inliner): Likewise.
17055 * ipa-pure-const.c (warn_function_noreturn): Likewise.
17056 (warn_function_cold): Likewise.
17057 (propagate_pure_const): Likewise.
17058 * ipa-ref.h (enum GTY): Likewise.
17059 * ipa-split.c (find_retbnd): Likewise.
17060 (consider_split): Likewise.
17061 (split_function): Likewise.
17062 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
17063 * ipa.c (walk_polymorphic_call_targets): Likewise.
17064 (symbol_table::remove_unreachable_nodes): Likewise.
17065 (process_references): Likewise.
17066 (cgraph_build_static_cdtor_1): Likewise.
17067 * lto-cgraph.c (lto_output_node): Likewise.
17068 (output_refs): Likewise.
17069 (compute_ltrans_boundary): Likewise.
17070 (input_overwrite_node): Likewise.
17071 (input_node): Likewise.
17072 (input_cgraph_1): Likewise.
17073 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
17074 * passes.c (pass_manager::execute_early_local_passes): Likewise.
17075 (class pass_chkp_instrumentation_passes): Likewise.
17076 (make_pass_chkp_instrumentation_passes): Likewise.
17077 * passes.def: Likewise.
17078 * rtl.h (struct GTY): Likewise.
17079 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
17080 * stor-layout.c (layout_type): Likewise.
17081 * symtab.c: Likewise.
17082 * target.def: Likewise.
17083 * targhooks.c (default_chkp_bound_type): Likewise.
17084 (default_chkp_bound_mode): Likewise.
17085 (default_builtin_chkp_function): Likewise.
17086 (default_chkp_function_value_bounds): Likewise.
17087 (default_chkp_make_bounds_constant): Likewise.
17088 (default_chkp_initialize_bounds): Likewise.
17089 * targhooks.h (default_chkp_bound_type): Likewise.
17090 (default_chkp_bound_mode): Likewise.
17091 (default_builtin_chkp_function): Likewise.
17092 (default_chkp_function_value_bounds): Likewise.
17093 (default_chkp_make_bounds_constant): Likewise.
17094 (default_chkp_initialize_bounds): Likewise.
17095 * toplev.c (compile_file): Likewise.
17096 (process_options): Likewise.
17097 * tree-core.h (DEF_BUILTIN): Likewise.
17098 (DEF_BUILTIN_CHKP): Likewise.
17099 * tree-inline.c (declare_return_variable): Likewise.
17100 (remap_gimple_stmt): Likewise.
17101 (copy_bb): Likewise.
17102 (initialize_inlined_parameters): Likewise.
17103 (expand_call_inline): Likewise.
17104 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
17105 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
17106 (make_pass_ipa_chkp_produce_thunks): Likewise.
17107 (make_pass_chkp): Likewise.
17108 (make_pass_chkp_opt): Likewise.
17109 (make_pass_chkp_instrumentation_passes): Likewise.
17110 * tree-pretty-print.c (dump_generic_node): Likewise.
17111 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
17112 * tree-ssa-dce.c (propagate_necessity): Likewise.
17113 (eliminate_unnecessary_stmts): Likewise.
17114 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
17115 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
17116 * tree-ssa-sccvn.h: Likewise.
17117 * tree-ssa-strlen.c (get_string_length): Likewise.
17118 (valid_builtin_call): Likewise.
17119 (adjust_last_stmt): Likewise.
17120 (handle_builtin_strchr): Likewise.
17121 (handle_builtin_strcpy): Likewise.
17122 (handle_builtin_stxncpy): Likewise.
17123 (handle_builtin_memcpy): Likewise.
17124 (handle_builtin_strcat): Likewise.
17125 (strlen_check_and_optimize_stmt): Likewise.
17126 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
17127 * tree-streamer-in.c: Likewise.
17128 * tree-streamer.c (record_common_node): Likewise.
17129 * tree.c (tree_code_size): Likewise.
17130 (wide_int_to_tree_1): Likewise.
17131 (type_contains_placeholder_1): Likewise.
17132 (build_common_tree_nodes): Likewise.
17133 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
17134 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
17135 (POINTER_BOUNDS_P): Likewise.
17136 (BOUNDED_TYPE_P): Likewise.
17137 (BOUNDED_P): Likewise.
17138 (CALL_WITH_BOUNDS_P): Likewise.
17139 (pointer_bounds_type_node): Likewise.
17140 * value-prof.c (gimple_ic): Likewise.
17141 * var-tracking.c (vt_add_function_parameters): Likewise.
17142 * varasm.c (make_decl_rtl): Likewise.
17143 (assemble_start_function): Likewise.
17144 (output_constant): Likewise.
17145 (maybe_assemble_visibility): Likewise.
17146 * varpool.c (ctor_for_folding): Likewise.
17147 * chkp-builtins.def: Remove.
17148 * ipa-chkp.c: Remove.
17149 * ipa-chkp.h: Remove.
17150 * rtl-chkp.c: Remove.
17151 * rtl-chkp.h: Remove.
17152 * tree-chkp-opt.c: Remove.
17153 * tree-chkp.c: Remove.
17154 * tree-chkp.h: Remove.
17156 2018-06-07 Carl Love <cel@us.ibm.com>
17158 * config/rs6000/vsx.md (vextract_fp_from_shorth,
17159 vextract_fp_from_shortl): Add BE support.
17161 2018-06-07 Paul Koning <ni1d@arrl.net>
17163 * compare-elim.c (try_merge_compare): Don't merge compare if
17164 address contains a side effect.
17165 (try_eliminate_compare): Likewise.
17167 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
17169 * config.gcc: Support "tremont".
17170 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
17171 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17173 * config/i386/i386.c (m_TREMONT): Define.
17174 (processor_target_table): Add "tremont".
17175 (PTA_TREMONT): Define.
17176 (ix86_lea_outperforms): Add TARGET_TREMONT.
17177 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
17178 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
17179 and M_INTEL_GOLDMONT_PLUS.
17180 (fold_builtin_cpu): Add "tremont".
17181 (ix86_add_stmt_cost): Add TARGET_TREMONT.
17182 (ix86_option_override_internal): Add "tremont".
17183 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
17184 (processor_type): Add PROCESSOR_TREMONT.
17185 * config/i386/x86-tune.def: Add m_TREMONT.
17186 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
17188 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
17190 * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
17191 symbol defined for msp430i* devices to be lower case.
17193 2018-06-07 Richard Biener <rguenther@suse.de>
17195 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
17196 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
17197 Properly wrap signed arithmetic if overflow wraps.
17199 2018-06-07 Jakub Jelinek <jakub@redhat.com>
17201 PR tree-optimization/69615
17202 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
17203 of a cast from a same precision integral SSA_NAME in a bb dominated
17204 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
17205 cast to utype if rhs2 has already a compatible type.
17207 2018-06-07 Richard Biener <rguenther@suse.de>
17209 PR tree-optimization/85935
17210 * graphite-scop-detection.c (find_params_in_bb): Analyze
17211 condition operands with respect to the correct loop. Assert
17212 the analysis doesn't fail.
17214 2018-06-04 Carl Love <cel@us.ibm.com>
17216 * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
17217 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
17218 as it is slightly cheaper.
17219 (first_match_or_eos_index_<mode>):
17220 Calculate index using natural element order.
17221 (first_match_index_<mode>):
17222 Calculate index using natural element order.
17223 (first_match_or_eos_index_<mode>):
17224 Calculate index using natural order.
17225 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
17226 for BE and LE modes.
17227 * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
17228 P9V_BUILTIN_VCLZLSBB_V16QI.
17229 * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
17232 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
17234 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
17235 indentation and line wrap for many prototypes. Add missing
17236 @smallexample directives around block of prototypes for vec_xl and
17239 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
17241 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
17242 track if we pass or return IEEE 128-bit floating point.
17243 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
17244 C++ mangling that is compatible with GCC 8.1.
17245 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
17246 (init_cumulative_args): Note if we pass or return IEEE 128-bit
17247 floating point types.
17248 (rs6000_function_arg_advance_1): Likewise.
17249 (rs6000_mangle_type): Optionally generate mangled names that match
17250 what GCC 8.1 generated for IEEE 128-bit floating point types.
17251 (rs6000_globalize_decl_name): If we have an external function that
17252 passes or returns IEEE 128-bit types, generate a weak reference
17253 from the mangled name used in GCC 8.1 to the current mangled
17255 (rs6000_init_builtins): Make __ibm128 use the long double type if
17256 long double is IBM extended double. Make __float128 use the long
17257 double type if long double is IEEE 128-bit.
17260 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
17261 macro for __ibm128 built-in functions.
17262 (PACK_IF): Add __ibm128 pack/unpack functions.
17263 (UNPACK_IF): Likewise.
17264 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
17265 enable long double built-in functions if long double is IEEE
17266 128-bit floating point.
17267 (rs6000_invalid_builtin): Update long double built-in function
17269 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
17270 functions, adjust the built-in function to use the long double
17271 built-in function if __ibm128 and long double are the same type.
17272 * doc/extend.texi (PowerPC builtins): Update documention for
17273 __builtin_{,un}pack_longdouble. Add documentation for
17274 __builtin_{,un}pack_ibm128.
17276 2018-06-06 Jim Wilson <jimw@sifive.com>
17278 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
17279 (struct machine_function): New field interrupt_mode.
17280 (riscv_handle_type_attribute): New function. Add forward declaration.
17281 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
17282 (riscv_expand_epilogue): Check interrupt_mode field.
17283 (riscv_set_current_function): Check interrupt attribute args and
17284 set interrupt_mode field.
17285 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
17286 (riscv_sret, riscv_uret): New.
17287 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
17288 new arguments to interrupt attribute.
17290 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
17293 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
17294 Don't handle -mcpu=power8 if -mpower9-vector is also used.
17296 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
17298 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
17299 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
17300 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
17301 several redundant entries.
17303 2018-06-06 David Malcolm <dmalcolm@redhat.com>
17305 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
17306 type from "rtx" to "rtx_insn *".
17307 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
17308 for local "call_insn", removing cast.
17309 (ix86_expand_call): Likewise, introducing a "call_insn" local.
17311 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
17313 PR tree-optimization/86066
17314 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
17315 for BIT_INSERT_EXPR stores.
17317 2018-06-06 Richard Biener <rguenther@suse.de>
17319 PR tree-optimization/86062
17320 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
17321 component refs ontop
17322 of to be offsetted base.
17324 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
17326 * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
17327 to be static and remove check on interrupt attribute name.
17329 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
17331 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17332 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
17334 2018-06-05 Steve Ellcey <sellcey@cavium.com>
17337 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
17339 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
17340 Remove second argument, change how error is called.
17341 (aarch64_layout_arg): Remove second argument from
17342 aarch64_err_no_fpadvsimd call.
17343 (aarch64_init_cumulative_args): Ditto.
17344 (aarch64_gimplify_va_arg_expr): Ditto.
17345 * config/aarch64/aarch64.md (mov<mode>): Ditto.
17347 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
17349 * config/i386/i386.md (simple_return_indirect_internal): New expander.
17350 (*simple_return_indirect_internal<mode>): Rename from
17351 simple_return_indirect_internal. Use W mode iterator.
17352 (rstorssp): New expander.
17353 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
17354 (clrssbsy): New expander.
17355 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
17357 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
17359 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
17361 (cmse_check_pointed_object): Likewise.
17363 2018-06-05 Martin Liska <mliska@suse.cz>
17365 PR gcov-profile/47618
17366 * doc/invoke.texi: Document how -fprofile-dir format
17369 2018-06-05 Richard Biener <rguenther@suse.de>
17371 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
17372 removal pretend DOM info isn't available so we do not update
17373 it and only remove edges, not dominated blocks. Actually free
17374 DOM info in case we removed something. Remove unreachable blocks.
17375 (mfb_keep_latches): Work with either DOM info or marked backedges.
17376 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
17377 first. Mark backedges if DOM info isn't available.
17378 (Re-)compute DOM info after cleanup_control_flow_pre.
17380 2018-06-05 Richard Biener <rguenther@suse.de>
17382 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
17383 (locus_discrim_hasher::hash): Adjust.
17384 (locus_discrim_hasher::equal): Likewise.
17385 (next_discriminator_for_locus): Work on line directly.
17386 (same_line_p): Pass in expanded locus1 as well.
17387 (assign_discriminators): Avoid redundant location expansions.
17389 2018-06-05 Richard Biener <rguenther@suse.de>
17391 PR tree-optimization/86046
17392 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
17393 if required after clearing TREE_ADDRESSABLE.
17395 2018-06-05 Richard Biener <rguenther@suse.de>
17397 PR tree-optimization/86047
17398 * tree-ssa-loop.c (for_each_index): Glob handling of all
17399 decls and constants and really handle all of them.
17401 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17404 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
17405 qualifier_void_pointer and qualifier_const_void_pointer.
17406 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
17407 (arm_init_builtins): Handle the above.
17408 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
17409 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
17410 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
17413 2018-06-05 Martin Liska <mliska@suse.cz>
17415 * auto-profile.c (read_autofdo_file): Do not use
17416 gcov_ctr_summary struct.
17417 (afdo_callsite_hot_enough_for_early_inline): Likewise.
17418 * coverage.c (struct counts_entry): Likewise.
17419 (read_counts_file): Read just single summary entry.
17420 (get_coverage_counts): Use gcov_summary struct.
17421 * coverage.h (get_coverage_counts): Likewise.
17422 * gcov-dump.c (dump_working_sets): Likewise.
17423 (tag_summary): Dump just single summary.
17424 * gcov-io.c (gcov_write_summary): Write just histogram
17426 (gcov_read_summary): Read just single summary.
17427 (compute_working_sets): Use gcov_summary struct.
17428 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
17429 of GCOV_COUNTERS_SUMMABLE.
17430 (GCOV_COUNTERS_SUMMABLE): Remove.
17431 (GCOV_FIRST_VALUE_COUNTER): Replace with
17432 GCOV_COUNTER_V_INTERVAL.
17433 (struct gcov_ctr_summary): Remove.
17434 (struct gcov_summary): Directly use fields of former
17436 (compute_working_sets): Use gcov_summary struct.
17437 * gcov.c (read_count_file): Do not use ctrs fields.
17438 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
17440 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
17442 * profile.c: Likewise.
17443 * profile.h: Likewise.
17445 2018-06-05 Martin Liska <mliska@suse.cz>
17447 PR gcov-profile/84846
17448 * gcov.c (output_lines): Print working directory only
17449 in intermediate format.
17451 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
17453 * config/s390/s390-builtin-types.def: Add void function type.
17454 * config/s390/s390-builtins.def: Use the function type for the
17457 2018-06-04 Jim Wilson <jimw@sifive.com>
17459 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
17461 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
17462 and maybe_eh_return. Change regno to unsigned int. Use new args to
17463 handle EH_RETURN_DATA_REGNO registers properly.
17464 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
17465 (riscv_expand_epilogue): Update comment. Change argument name and
17466 type. Update code to use new name and type. Pass new args to
17467 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
17469 * config/riscv/riscv.md (NORMAL_RETURN): New.
17470 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
17471 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
17472 (eh_return): Call gen_eh_return_internal and emit barrier.
17473 (eh_return_internal): Call riscv_expand_epilogue.
17475 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
17477 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
17478 bit_insertion field and declare can_be_merged_into method.
17479 (merged_store_group::can_be_merged_into): New method.
17480 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
17481 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
17482 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
17484 2018-06-04 Richard Biener <rguenther@suse.de>
17486 PR tree-optimization/85955
17487 * builtins.c (fold_builtin_sincos): Convert pointers to
17488 destination to appropriate type before dereferencing.
17490 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
17492 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
17494 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
17496 * expr.c (expand_expr_real_1): Force the operand into memory if
17497 its TYPE_MODE is BLKmode and if there is no integer mode for
17498 the number of bits being extracted.
17500 2018-06-04 Jakub Jelinek <jakub@redhat.com>
17504 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
17505 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
17507 2018-06-04 Richard Biener <rguenther@suse.de>
17509 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
17510 (cleanup_tree_cfg_noloop): ... single caller. Do
17511 start_recording_case_labels later.
17513 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
17515 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
17516 to _IMMINTRIN_H_INCLUDED.
17517 * config/i386/pconfigintrin.h: Ditto.
17518 * config/i386/waitpkgintrin.h: Ditto.
17519 * config/i386/immintrin.h: Add includes for sgxintrin.h,
17520 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
17521 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
17522 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
17523 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
17524 waitpkgintrin.h and cldemoteintrin.h.
17526 2018-06-04 Richard Biener <rguenther@suse.de>
17528 PR tree-optimization/86038
17529 * tracer.c (find_best_successor): Check probability for
17530 being initialized, bail out if not.
17532 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
17535 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
17536 of bits to ignore when comparing architectures.
17538 2018-06-04 Jakub Jelinek <jakub@redhat.com>
17540 PR tree-optimization/69615
17541 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
17542 maximum or minimum of the type, try to merge it also as if
17543 range1 is + [-, x - 1] or + [x + 1, -].
17546 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
17548 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
17550 PR tree-optimization/86034
17551 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
17552 the unsigned bitfield type in a bit insertion sequence if it does not
17553 have a larger precision than the bitfield size.
17554 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
17556 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
17558 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
17560 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
17562 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
17563 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
17564 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
17565 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
17567 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
17569 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
17570 Disable -fdelete-null-pointer-checks for ELF toolchain.
17572 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
17573 Kito Cheng <kito.cheng@gmail.com>
17575 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
17576 (nds32le-*-*, nds32be-*-*): Integrate checking process.
17577 (nds32*-*-*): Add glibc and uclibc conditions.
17578 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
17579 (TARGET_EXCEPT_UNWIND_INFO): Define.
17580 * config/nds32/elf.h: New file.
17581 * config/nds32/linux.h: New file.
17582 * config/nds32/nds32-elf.opt: New file.
17583 * config/nds32/nds32-linux.opt: New file.
17584 * config/nds32/nds32-fp-as-gp.c
17585 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
17586 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
17588 (nds32_asm_file_end): Ditto.
17589 (nds32_print_operand): Ditto.
17590 (nds32_insert_attributes): Ditto.
17591 (nds32_init_libfuncs): New function.
17592 (TARGET_HAVE_TLS): Define.
17593 (TARGET_INIT_LIBFUNCS): Define.
17594 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
17596 (TARGET_ELF): Apply different mcmodel setting.
17597 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
17598 been migrated into elf.h and linux.h files.
17599 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
17600 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
17601 (mcmodel): The content has been migrated into nds32-elf.opt and
17602 nds32-linux.opt files.
17603 * config/nds32/t-elf: New file.
17604 * config/nds32/t-linux: New file.
17606 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
17607 Shiva Chen <shiva0217@gmail.com>
17609 * config/nds32/constants.md (unspec_volatile_element): Add
17610 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
17611 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
17613 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
17614 (make_pass_nds32_fp_as_gp): Declare.
17615 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
17617 (nds32_asm_function_end_prologue): Remove unused asm output.
17618 (nds32_asm_function_begin_epilogue): Remove unused asm output.
17619 (nds32_asm_file_start): Output necessary fp_as_gp information.
17620 (nds32_option_override): Adjust register usage.
17621 (nds32_expand_prologue): Consider fp_as_gp situation.
17622 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
17623 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
17626 (simple_return): Ditto.
17627 (omit_fp_begin): Output special directive for fp_as_gp.
17628 (omit_fp_end): Output special directive for fp_as_gp.
17629 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
17630 mforbid-fp-as-gp): New options.
17632 2018-06-01 Mark Wielaard <mark@klomp.org>
17634 * dwarf2out.c (dwarf2out_finish): Remove generation of
17635 DW_AT_loclists_base.
17637 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
17639 * gimple-ssa-store-merging.c: Include gimple-fold.h.
17640 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
17641 (struct merged_store_group): Add bit_insertion field.
17642 (dump_char_array): Use standard hexadecimal format.
17643 (merged_store_group::merged_store_group): Set bit_insertion to false.
17644 (merged_store_group::apply_stores): Use optimal buffer size. Deal
17645 with BIT_INSERT_EXPR stores. Move up code updating the mask and
17646 also print the mask in the dump file.
17647 (pass_store_merging::gate): Minor tweak.
17648 (imm_store_chain_info::coalesce_immediate): Fix wrong association
17649 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
17650 stores with INTEGER_CST stores.
17651 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
17652 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
17653 and use it throughout. Generate bit insertion sequences if need be.
17654 (pass_store_merging::process_store): Remove redundant condition.
17655 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
17657 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
17659 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
17660 the 128-bit floating point types. Fix function comment.
17662 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
17664 * config/aarch64/aarch64-simd.md
17665 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
17667 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
17670 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
17672 PR tree-optimization/85989
17673 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
17675 (backprop::intersect_uses): Check it when deciding whether this
17676 is a backedge reference.
17677 (backprop::process_block): Add each phi to m_visited_phis
17678 after visiting it, then clear it at the end.
17680 2018-06-01 Richard Biener <rguenther@suse.de>
17682 * tree-vectorizer.h (vect_dr_stmt): New function.
17683 (vect_get_load_cost): Adjust.
17684 (vect_get_store_cost): Likewise.
17685 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
17686 Use vect_dr_stmt instead of DR_SMTT.
17687 (vect_record_base_alignments): Likewise.
17688 (vect_calculate_target_alignment): Likewise.
17689 (vect_compute_data_ref_alignment): Likewise and make static.
17690 (vect_update_misalignment_for_peel): Likewise.
17691 (vect_verify_datarefs_alignment): Likewise.
17692 (vector_alignment_reachable_p): Likewise.
17693 (vect_get_data_access_cost): Likewise. Pass down
17694 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
17695 (vect_get_peeling_costs_all_drs): Likewise.
17696 (vect_peeling_hash_get_lowest_cost): Likewise.
17697 (vect_enhance_data_refs_alignment): Likewise.
17698 (vect_find_same_alignment_drs): Likewise.
17699 (vect_analyze_data_refs_alignment): Likewise.
17700 (vect_analyze_group_access_1): Likewise.
17701 (vect_analyze_group_access): Likewise.
17702 (vect_analyze_data_ref_access): Likewise.
17703 (vect_analyze_data_ref_accesses): Likewise.
17704 (vect_vfa_segment_size): Likewise.
17705 (vect_small_gap_p): Likewise.
17706 (vectorizable_with_step_bound_p): Likewise.
17707 (vect_prune_runtime_alias_test_list): Likewise.
17708 (vect_analyze_data_refs): Likewise.
17709 (vect_supportable_dr_alignment): Likewise.
17710 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
17711 (vect_gen_prolog_loop_niters): Likewise.
17712 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
17713 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
17715 (vect_recog_mask_conversion_pattern): Likewise.
17716 (vect_try_gather_scatter_pattern): Likewise.
17717 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
17718 to vect_get_store_cost.
17719 (vect_get_store_cost): Get stmt_info instead of DR.
17720 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
17721 (vect_get_load_cost): Get stmt_info instead of DR.
17723 2018-06-01 Richard Biener <rguenther@suse.de>
17725 PR middle-end/86017
17726 * gimple-fold.c (var_decl_component_p): Also allow offsetted
17727 vars wrapped in MEM_REFs.
17729 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
17731 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17732 Fix subreg tests so that we only return a choice between
17733 GENERAL_REGS and FP_REGS if the original classes included both.
17735 2018-06-01 Richard Biener <rguenther@suse.de>
17738 * tree-ssa-structalias.c (get_function_part_constraint):
17739 Handle NULL fi->decl.
17740 (find_func_aliases_for_call): Properly handle indirect
17741 fi from direct call.
17742 (find_func_clobbers): Likewise.
17743 (ipa_pta_execute): Likewise.
17744 (create_variable_info_for): For functions that are ifunc_resolver
17745 resolve to a varinfo that contains the result of the resolver call.
17746 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
17749 2018-05-31 Michael Collison <michael.collison@arm.com>
17751 * config/aarch64/aarch64.md:
17752 (*fix_to_zero_extenddfdi2): New pattern.
17753 * gcc.target/aarch64/fix_extend1.c: New testcase.
17755 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
17757 PR middle-end/78809
17758 PR middle-end/83026
17759 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
17760 and BUILT_IN_STRNCMP_EQ.
17761 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
17762 BUILT_IN_STRNCMP_EQ.
17763 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
17764 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17765 (gimple_fold_builtin): Likewise.
17766 * tree-ssa-strlen.c (compute_string_length): New function.
17767 (determine_min_obsize): New function.
17768 (handle_builtin_string_cmp): New function to handle calls to
17769 string compare functions.
17770 (strlen_optimize_stmt): Add handling to builtin string compare
17772 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
17773 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17774 * tree.c (build_common_builtin_nodes): Add new defines of
17775 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
17777 2018-05-31 Jakub Jelinek <jakub@redhat.com>
17780 * bb-reorder.c (pass_partition_blocks::gate): Return false for
17781 functions with naked attribute.
17783 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
17785 * config/i386/sse.md (avx_vec_concat<mode>):
17786 Substitute concat_tg_mode mode attribute with xtg_mode.
17787 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
17788 (concat_tg_mode): Remove mode attribute.
17790 2018-05-31 Martin Sebor <msebor@redhat.com>
17793 * calls.c (alloc_max_size): Correct a logic error/typo.
17794 Treat excessive arguments as infinite. Warn for invalid arguments.
17795 * doc/invoke.texi (-Walloc-size-larger-than): Update.
17797 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
17800 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
17801 and movx for Haswell.
17803 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
17804 Cesar Philippidis <cesar@codesourcery.com>
17806 PR middle-end/85879
17807 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
17808 when emitting error on private/firstprivate reductions.
17809 * omp-low.c (lower_omp_target): Avoid reference-type processing
17810 on pointers for firstprivate clause.
17812 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
17814 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
17817 * config/aarch64/aarch64-simd.md
17818 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
17819 (aarch64_ld1_x3_<mode>): Likewise
17820 (aarch64_st1x2<VALLDIF:mode>): Likewise
17821 (aarch64_st1_x2_<mode>): Likewise
17822 (aarch64_st1x3<VALLDIF:mode>): Likewise
17823 (aarch64_st1_x3_<mode>): Likewise
17824 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
17825 (vld1_s8_x3): Likewise.
17826 (vld1_u16_x3): Likewise.
17827 (vld1_s16_x3): Likewise.
17828 (vld1_u32_x3): Likewise.
17829 (vld1_s32_x3): Likewise.
17830 (vld1_u64_x3): Likewise.
17831 (vld1_s64_x3): Likewise.
17832 (vld1_f16_x3): Likewise.
17833 (vld1_f32_x3): Likewise.
17834 (vld1_f64_x3): Likewise.
17835 (vld1_p8_x3): Likewise.
17836 (vld1_p16_x3): Likewise.
17837 (vld1_p64_x3): Likewise.
17838 (vld1q_u8_x3): Likewise.
17839 (vld1q_s8_x3): Likewise.
17840 (vld1q_u16_x3): Likewise.
17841 (vld1q_s16_x3): Likewise.
17842 (vld1q_u32_x3): Likewise.
17843 (vld1q_s32_x3): Likewise.
17844 (vld1q_u64_x3): Likewise.
17845 (vld1q_s64_x3): Likewise.
17846 (vld1q_f16_x3): Likewise.
17847 (vld1q_f32_x3): Likewise.
17848 (vld1q_f64_x3): Likewise.
17849 (vld1q_p8_x3): Likewise.
17850 (vld1q_p16_x3): Likewise.
17851 (vld1q_p64_x3): Likewise.
17852 (vst1_s64_x2): Likewise.
17853 (vst1_u64_x2): Likewise.
17854 (vst1_f64_x2): Likewise.
17855 (vst1_s8_x2): Likewise.
17856 (vst1_p8_x2): Likewise.
17857 (vst1_s16_x2): Likewise.
17858 (vst1_p16_x2): Likewise.
17859 (vst1_s32_x2): Likewise.
17860 (vst1_u8_x2): Likewise.
17861 (vst1_u16_x2): Likewise.
17862 (vst1_u32_x2): Likewise.
17863 (vst1_f16_x2): Likewise.
17864 (vst1_f32_x2): Likewise.
17865 (vst1_p64_x2): Likewise.
17866 (vst1q_s8_x2): Likewise.
17867 (vst1q_p8_x2): Likewise.
17868 (vst1q_s16_x2): Likewise.
17869 (vst1q_p16_x2): Likewise.
17870 (vst1q_s32_x2): Likewise.
17871 (vst1q_s64_x2): Likewise.
17872 (vst1q_u8_x2): Likewise.
17873 (vst1q_u16_x2): Likewise.
17874 (vst1q_u32_x2): Likewise.
17875 (vst1q_u64_x2): Likewise.
17876 (vst1q_f16_x2): Likewise.
17877 (vst1q_f32_x2): Likewise.
17878 (vst1q_f64_x2): Likewise.
17879 (vst1q_p64_x2): Likewise.
17880 (vst1_s64_x3): Likewise.
17881 (vst1_u64_x3): Likewise.
17882 (vst1_f64_x3): Likewise.
17883 (vst1_s8_x3): Likewise.
17884 (vst1_p8_x3): Likewise.
17885 (vst1_s16_x3): Likewise.
17886 (vst1_p16_x3): Likewise.
17887 (vst1_s32_x3): Likewise.
17888 (vst1_u8_x3): Likewise.
17889 (vst1_u16_x3): Likewise.
17890 (vst1_u32_x3): Likewise.
17891 (vst1_f16_x3): Likewise.
17892 (vst1_f32_x3): Likewise.
17893 (vst1_p64_x3): Likewise.
17894 (vst1q_s8_x3): Likewise.
17895 (vst1q_p8_x3): Likewise.
17896 (vst1q_s16_x3): Likewise.
17897 (vst1q_p16_x3): Likewise.
17898 (vst1q_s32_x3): Likewise.
17899 (vst1q_s64_x3): Likewise.
17900 (vst1q_u8_x3): Likewise.
17901 (vst1q_u16_x3): Likewise.
17902 (vst1q_u32_x3): Likewise.
17903 (vst1q_u64_x3): Likewise.
17904 (vst1q_f16_x3): Likewise.
17905 (vst1q_f32_x3): Likewise.
17906 (vst1q_f64_x3): Likewise.
17907 (vst1q_p64_x3): Likewise.
17909 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
17911 * config/msp430/msp430.c (msp430_output_labelref): Prepend
17912 user_label_prefix to name.
17914 * tree-core.h: Update comment about the format of NAME string
17915 passed to handler in attribute_spec.
17917 * config/msp430/msp430.md: Remove erroneous subreg expression from
17918 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
17919 zero_extend{q,h}isi2.
17921 2018-05-30 Borislav Petkov <bp@suse.de>
17923 * doc/extend.texi: Document some architecture specific
17924 constraints and sort entries.
17926 2018-05-30 Martin Sebor <msebor@redhat.com>
17928 PR middle-end/85369
17929 * builtins.c (expand_builtin_stpcpy_1): New function.
17930 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
17931 only if the former succeeds.
17933 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
17935 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
17938 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
17940 * doc/invoke.texi (-flinker-output): Document
17942 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
17944 * passes.c (ipa_write_summaries): Only modify statements if body
17946 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
17947 incrementally linking.
17948 (ipa_passes): Likewise.
17949 * lto-cgraph.c (lto_output_node): When incrementally linking do not
17950 pass down resolution info.
17951 * common.opt (flag_incremental_link): Update info.
17952 * gcc.c (plugin specs): Turn flinker-output=* to
17953 -plugin-opt=-linker-output-known
17954 * toplev.c (compile_file): Also cut compilation when doing incremental
17956 * flag-types. (enum lto_partition_model): Add
17957 LTO_LINKER_OUTPUT_NOLTOREL.
17958 (invoke.texi): Add -flinker-output docs.
17959 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
17960 link same way as WPA; do not stream in dead initializers.
17962 * dwarf2out.c (dwarf2out_die_ref_for_decl,
17963 darf2out_register_external_decl): Support incremental link.
17965 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
17967 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
17969 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
17971 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
17972 it down to simple_object_copy_lto_debug_sections.
17973 (run_gcc): Determine incremental LTO link time and configure
17974 lto1 into non-wpa mode, disable renaming of debug sections.
17976 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
17978 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17979 descriptions of various incorrectly documented functions.
17981 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
17984 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
17985 address check not strict.
17987 2018-05-30 Richard Biener <rguenther@suse.de>
17989 PR tree-optimization/85964
17990 * tracer.c (better_p): Drop initialized count check, we only
17991 call the function with initialized counts now.
17992 (find_best_successor): Do find a best edge if one
17993 has uninitialized count.
17994 (find_best_predecessor): Likewise. Do BB frequency check only
17995 if count is initialized.
17997 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
17999 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
18000 (aarch64_ldrstr_offset_compare): New.
18001 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
18002 load/store orderings.
18003 (aarch64_gen_adjusted_ldpstp): Likewise.
18005 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
18007 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
18008 Check for subset of GENERAL_REGS and FP_REGS.
18009 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
18012 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
18014 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
18015 and wi::to_poly_offset. Add the current offset and then check
18016 whether the sum fits, rather than using an unchecked addition of
18017 a checked term. Check for a shwi rather than a uhwi.
18018 * expr.c (get_bit_range): Use tree_to_poly_uint64.
18019 (store_constructor): Use poly_int_tree_p.
18020 (expand_expr_real_1): Likewise.
18021 * function.c (assign_temp): Likewise.
18022 * fold-const.c (const_binop): Use poly_int_tree_p and
18023 wi::to_poly_offset.
18024 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
18026 * ipa-icf-gimple.c (func_checker::compare_operand): Use
18027 to_poly_offset for MEM offsets.
18028 * ipa-icf.c (sem_variable::equals): Likewise.
18029 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
18030 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
18031 wi::to_poly_offset for BIT_FIELD_REF offsets.
18032 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
18033 wi::to_poly_offset.
18034 * var-tracking.c (emit_note_insn_var_location): Use
18035 tree_to_poly_uint64.
18037 2018-05-29 Jim Wilson <jimw@sifive.com>
18039 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
18041 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
18044 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
18045 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
18047 (sse4_1_round<mode>2): Use nonimmediate_operand
18048 for operand 1 predicate.
18050 2018-05-29 Martin Sebor <msebor@redhat.com>
18051 Richard Biener <rguenther@suse.de>
18054 * calls.c (get_size_range): Call determine_value_range instead
18055 of get_value_range..
18056 * tree-vrp.h (determine_value_range): Declared new function.
18057 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
18059 2018-05-29 Richard Biener <rguenther@suse.de>
18061 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
18062 sure to use non-pattern stmts for get_earlier_stmt arguments.
18063 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
18064 called on pattern stmts.
18065 (get_later_stmt): Likewise.
18067 2018-05-29 Martin Liska <mliska@suse.cz>
18069 PR gcov-profile/85759
18070 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
18073 2018-05-29 Jakub Jelinek <jakub@redhat.com>
18075 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
18077 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
18081 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
18082 VEC_PACK_FLOAT_EXPR): New tree codes.
18083 * tree-pretty-print.c (op_code_prio): Handle
18084 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
18085 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
18086 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
18087 * tree-inline.c (estimate_operator_cost): Likewise.
18088 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
18089 * fold-const.c (const_binop): Likewise.
18090 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
18091 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
18092 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
18093 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
18094 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
18095 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
18096 * expr.c (expand_expr_real_2): Likewise.
18097 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
18098 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
18099 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
18101 * optabs.c (expand_widen_pattern_expr): For
18102 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
18103 sign from result type rather than operand's type.
18104 (expand_binop_directly): For vec_packu_float_optab and
18105 vec_packs_float_optab allow result type to be different from operand's
18107 * optabs-tree.c (optab_for_tree_code): Handle
18108 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
18109 VEC_PACK_FLOAT_EXPR. Formatting fixes.
18110 * tree-vect-generic.c (expand_vector_operations_1): Handle
18111 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
18112 VEC_PACK_FLOAT_EXPR.
18113 * tree-vect-stmts.c (supportable_widening_operation): Handle
18115 (supportable_narrowing_operation): Handle FLOAT_EXPR.
18116 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
18117 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
18118 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
18119 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
18121 (vec_pack<floatprefix>_float_<mode>): New expander.
18122 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
18124 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
18125 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
18126 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
18127 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
18128 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
18130 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
18131 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
18132 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
18133 VEC_PACK_FLOAT_EXPR): Document.
18135 2018-05-29 Richard Biener <rguenther@suse.de>
18137 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
18139 (stmt_vec_info_vec): Make pointer.
18140 (init_stmt_vec_info_vec): Remove.
18141 (free_stmt_vec_info_vec): Likewise.
18142 (set_stmt_vec_info_vec): New function.
18143 (free_stmt_vec_infos): Likewise.
18144 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
18145 (set_vinfo_for_stmt): Likewise.
18146 (get_earlier_stmt): Likewise.
18147 (get_later_stmt): Likewise.
18148 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
18149 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
18150 (vec_info::~vec_info): Free stmt_vec_infos.
18151 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
18152 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
18153 (pass_slp_vectorize::execute): Likewise.
18154 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
18155 (free_stmt_vec_info_vec): Likewise.
18156 (set_stmt_vec_info_vec): New function.
18157 (free_stmt_vec_infos): Likewise.
18158 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
18159 the global stmt_vec_info_vec.
18160 * tree-parloops.c (gather_scalar_reductions): Use
18161 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
18164 2018-05-29 Richard Biener <rguenther@suse.de>
18166 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
18168 2018-05-29 Martin Liska <mliska@suse.cz>
18169 David Malcolm <dmalcolm@redhat.com>
18171 * vec.c (test_reverse): New.
18172 (vec_c_tests): Add new test.
18173 * vec.h (vl_ptr>::reverse): New function.
18175 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
18177 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
18179 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
18182 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
18184 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
18186 2018-05-28 Richard Biener <rguenther@suse.de>
18188 PR tree-optimization/85933
18189 * tree-vect-data-refs.c (vect_record_base_alignments): Only
18190 look at stmts marked as vectorizable.
18192 2018-05-28 Richard Biener <rguenther@suse.de>
18194 PR tree-optimization/85934
18195 * tree-vect-generic.c (expand_vector_operations_1): Hoist
18196 vector boolean check before scalar optimization.
18198 2018-05-28 Jakub Jelinek <jakub@redhat.com>
18200 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
18203 2018-05-28 Mark Wielaard <mark@klomp.org>
18205 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
18206 if it is an expression containing a minus sign.
18208 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
18210 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
18212 2018-05-27 Paul Koning <ni1d@arrl.net>
18214 * config/pdp11/pdp11.md (truncsihi2): Remove.
18216 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
18217 Chung-Ju Wu <jasonwucj@gmail.com>
18219 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
18221 (unaligned_store_dw): Ditto.
18222 * config/nds32/nds32-memory-manipulation.c
18223 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
18224 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
18225 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
18226 (emit_setmem_word_loop): Rename to ...
18227 (emit_setmem_doubleword_loop): ... this.
18228 (nds32_gen_dup_4_byte_to_word_value): New function.
18229 (nds32_gen_dup_8_byte_to_double_word_value): New function.
18230 (nds32_expand_setmem_loop): Refine implementation.
18231 (nds32_expand_setmem_loop_v3m): Ditto.
18232 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
18235 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
18237 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
18239 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
18241 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
18242 (nds32_init_machine_status): Initialize machine->attr_naked_p and
18243 machine->attr_no_prologue_p.
18244 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
18245 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
18246 (nds32_expand_epilogue): Consider attr_naked_p.
18247 (nds32_expand_epilogue_v3pop): Likewise.
18248 (nds32_can_use_return_insn): Likewise.
18249 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
18250 attr_no_prologue_p fields.
18251 * config/nds32/nds32.opt (mret-in-naked-func): New option.
18253 2018-05-27 Jakub Jelinek <jakub@redhat.com>
18256 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
18258 * config/i386/sse.md
18259 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
18261 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
18263 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
18265 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
18267 (*<floatsuffix>floatv2div2sf2): Rename to ...
18268 (*float<floatunssuffix>v2div2sf2): ... this.
18269 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
18270 (float<floatunssuffix>v2div2sf2_mask): ... this.
18271 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
18272 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
18273 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
18275 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
18277 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
18279 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
18281 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
18283 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
18285 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
18286 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
18287 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
18288 gen_ufix_truncv8dfv8si2.
18289 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
18290 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
18291 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
18292 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
18293 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
18294 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
18295 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
18296 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
18298 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
18302 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
18304 2018-05-25 Jim Wilson <jimw@sifive.com>
18306 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
18307 * config/riscv/riscv.c (struct machine_function): Add
18308 interrupt_handler_p and attribute_checked_p fields.
18309 (riscv_attribute_table): Add interrupt.
18310 (riscv_interrupt_type_p): New.
18311 (riscv_save_reg_p): Save extra regs for interrupt handler.
18312 (riscv_use_save_libcall): Return false for interrupt handler.
18313 (riscv_first_stack_step): Add forward declaration.
18314 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
18315 for interrupt handler with large frame. Use it for saved reg list.
18316 (riscv_expand_prologue): Move flag_stack_usage_info support to
18317 eliminate duplication.
18318 (riscv_expand_epilogue): Generate mret for interrupt handler.
18319 (riscv_epilogue_uses): New.
18320 (riscv_can_use_return_insn): Return false for interrupt handler.
18321 (riscv_function_ok_for_sibcall): Likewise.
18322 (riscv_set_current_function): Add interrupt handler support.
18323 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
18324 * config/riscv/riscv.md (UNSPECV_MRET): New.
18326 (riscv_frflags, riscv_fsflags): Use tab after opcode.
18328 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
18330 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
18332 PR tree-optimization/85712
18333 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
18334 this candidate has already been replaced in-situ by a copy.
18336 2018-05-25 Jason Merrill <jason@redhat.com>
18338 PR c++/80485 - inline function non-zero address.
18339 * symtab.c (nonzero_address): Check DECL_COMDAT.
18341 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
18344 * config/alpha/alpha.md (ashlsi3): New insn pattern.
18345 (*ashlsi_se): Rename from *ashldi_se. Define as sign
18346 extension of SImode operation. Use const123_operand predicate.
18347 (*saddsi_1): Remove.
18348 (*saddl_se_1): Ditto.
18349 (*ssubsi_1): Ditto.
18350 (*ssubl_se_1): Ditto.
18351 * config/alpha/predicates.md (const123_operand): New predicate.
18352 * config/alpha/constraints.md (P): Use IN_RANGE.
18354 2018-05-25 Richard Biener <rguenther@suse.de>
18356 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
18358 (ref_maybe_used_by_stmt_p): Likewise.
18359 (stmt_may_clobber_ref_p): Likewise.
18360 (stmt_may_clobber_ref_p_1): Likewise.
18361 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
18363 (ref_maybe_used_by_stmt_p): Likewise.
18364 (stmt_may_clobber_ref_p): Likewise.
18365 (stmt_may_clobber_ref_p_1): Likewise.
18366 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
18367 the alias oracle to disambiguate DRs with stmts DR analysis
18369 (vect_analyze_data_refs): Do not give up on not analyzable
18370 DRs for BB vectorization. Remove code truncating the dataref
18373 2018-05-25 Jakub Jelinek <jakub@redhat.com>
18376 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
18377 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
18378 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
18380 2018-05-25 Richard Biener <rguenther@suse.de>
18382 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
18383 function, combining stmt data ref gathering and fatal analysis
18385 (vect_analyze_data_refs): Remove now redudnant code and simplify.
18386 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
18387 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
18388 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
18389 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
18391 2018-05-25 Bin Cheng <bin.cheng@arm.com>
18393 PR tree-optimization/85720
18394 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
18395 SCC if all partitions are builtins.
18396 (version_loop_by_alias_check): New parameter. Generate cancelable
18397 runtime alias check if all partitions are builtins.
18398 (distribute_loop): Update call to above function.
18400 2018-05-25 Bin Cheng <bin.cheng@arm.com>
18402 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
18403 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
18404 (parm_default_def_partition_arg): Ditto.
18405 (set_parm_default_def_partition): Ditto.
18406 (get_parm_default_def_partitions): Ditto and make it static.
18407 (get_undefined_value_partitions): Ditto and make it static.
18408 (remove_ssa_form): Refactor call to init_var_map here.
18409 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
18410 computation for loop region.
18411 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
18412 (register_default_def): Delete.
18413 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
18414 (parm_default_def_partition_arg): Ditto.
18415 (set_parm_default_def_partition): Ditto.
18416 (get_parm_default_def_partitions): Ditto and make it static.
18417 (get_undefined_value_partitions): Ditto and make it static.
18418 (coalesce_with_default, coalesce_with_default): Update comment.
18419 (create_coalesce_list_for_region): New func factored out from
18420 create_outofssa_var_map.
18421 (populate_coalesce_list_for_outofssa): New func factored out from
18422 create_outofssa_var_map and coalesce_ssa_name.
18423 (create_outofssa_var_map): Delete.
18424 (coalesce_ssa_name): Refactor to support live range computation.
18425 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
18426 (get_parm_default_def_partitions): Delete.
18427 (get_undefined_value_partitions): Ditto.
18428 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
18429 computation for loop region.
18430 (new_tree_live_info, loe_visit_block): Ditto.
18431 (live_worklist, set_var_live_on_entry): Ditto.
18432 (calculate_live_on_exit, verify_live_on_entry): Ditto.
18433 * tree-ssa-live.h (struct _var_map): New fields.
18434 (init_var_map): Change decl.
18435 (region_contains_p): New.
18437 2018-05-25 Bin Cheng <bin.cheng@arm.com>
18439 * tree-ssa-live.h (live_merge_and_clear): Delete.
18441 2018-05-25 Richard Biener <rguenther@suse.de>
18444 * tree-dump.c (dequeue_and_dump): Remove access to removed
18445 operand 2 of a SWITCH_EXPR.
18447 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
18449 * doc/sourcebuild.texi (vect_double_cond_arith): Include
18450 multiplication and division.
18451 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
18452 (cond_udiv@var{m}, cond_umod@var{m}): Document.
18453 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
18454 (cond_udiv_optab, cond_umod_optab): New optabs.
18455 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
18456 (IFN_COND_RDIV): New internal functions.
18457 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
18458 TRUNC_MOD_EXPR and RDIV_EXPR.
18459 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
18460 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
18462 (SVE_INT_BINARY): Include mult.
18463 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
18464 (optab, sve_int_op): Handle mult.
18465 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
18467 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
18468 for SVE_INT_BINARY_SD.
18470 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
18472 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
18473 (optab, sve_int_op): Handle div and udiv.
18474 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
18475 for SVE_INT_BINARY_SD.
18476 (*<optab><mode>3): New insn for the same.
18478 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
18480 * tree-vect-patterns.c: Include predict.h.
18481 (vect_recog_divmod_pattern): Restrict check for division support
18482 to when optimizing for size.
18484 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
18486 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
18487 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
18488 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
18489 (gimple_match_op::set_op): Likewise.
18490 (gimple_resimplify4): Declare.
18491 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
18492 (expr::gen_transform): Likewise.
18493 (decision_tree::gen): Generate a simplification routine for 4 operands.
18494 * gimple-match-head.c (gimple_simplify): Add an overload for
18495 4 operands. In the top-level function, handle up to 4 call
18496 arguments and call gimple_resimplify4.
18497 (gimple_resimplify4): New function.
18498 (build_call_internal): Pass a fourth operand.
18499 (maybe_push_to_seq): Likewise.
18500 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
18501 Fold VEC_COND_EXPRs of an operation and a default value into
18502 an IFN_COND_* function if possible.
18503 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
18505 (SVE_COND_FP_BINARY): Include them.
18506 (optab, sve_fp_op): Handle them.
18507 (SVE_INT_BINARY_REV): New code iterator.
18508 (SVE_COND_FP_BINARY_REV): New int iterator.
18509 (commutative): New int attribute.
18510 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
18512 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
18514 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
18515 (*cond_<optab><mode>): New patterns for reversed operands.
18517 2018-05-25 Richard Biener <rguenther@suse.de>
18519 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
18520 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
18521 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
18522 (STMT_VINFO_GROUPED_ACCESS): Adjust.
18523 * tree-vect-data-refs.c (everywhere): Adjust users.
18524 * tree-vect-loop.c (everywhere): Likewise.
18525 * tree-vect-slp.c (everywhere): Likewise.
18526 * tree-vect-stmts.c (everywhere): Likewise.
18527 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
18529 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
18531 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
18533 (gcc_cv_as_section_exclude): ... this.
18534 Try Solaris as #exclude syntax.
18535 * configure: Regenerate.
18536 * config.in: Regenerate.
18537 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
18539 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
18540 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
18542 * varasm.c (default_elf_asm_named_section): Don't check if
18543 HAVE_GAS_SECTION_EXCLUDE is defined.
18545 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
18547 * doc/md.texi: Update the documentation of the cond_* optabs
18548 to mention the new final operand. Fix GET_MODE_NUNITS call.
18549 Describe the scalar case too.
18550 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
18551 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
18553 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
18554 (get_conditional_internal_fn): Update comment.
18555 * tree-vect-loop.c (vectorizable_reduction): Pass the original
18556 accumulator value as a final argument to conditional functions.
18557 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
18558 a define_expand and add an "else" operand. Assert for now that
18559 the else operand is equal to operand 2. Use SVE_INT_BINARY and
18560 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
18561 (*cond_<optab><mode>): New patterns.
18562 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
18563 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
18564 (UNSPEC_COND_EOR): Delete.
18565 (optab): Remove associated mappings.
18566 (SVE_INT_BINARY): New code iterator.
18567 (sve_int_op): Remove int attribute and add "minus" to the code
18569 (SVE_COND_INT_OP): Delete.
18570 (SVE_COND_FP_OP): Rename to...
18571 (SVE_COND_FP_BINARY): ...this.
18573 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
18575 * optabs.c (can_reuse_operands_p): New function.
18576 (maybe_legitimize_operands): Try to reuse the results for
18579 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
18581 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
18582 Add {q} suffix to insn mnemonic.
18584 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
18586 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
18587 (msp430_warn_func_return): New.
18589 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
18591 * fold-const.c (tree_nonzero_bits): New function.
18592 * fold-const.h (tree_nonzero_bits): Likewise.
18593 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
18594 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
18596 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
18600 * varasm.c (assemble_alias): Check ifunc_resolver only on
18603 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
18606 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
18607 when memory input operand is handled.
18609 2018-05-24 Luis Machado <luis.machado@linaro.org>
18611 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
18613 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
18615 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
18617 * match.pd: Delay FMA folds until after vectorization.
18619 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
18622 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
18623 address check not strict.
18625 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
18627 * gimple-match.h (gimple_match_op): New class.
18628 (mprts_hook): Replace parameters with a gimple_match_op *.
18629 (maybe_build_generic_op): Likewise.
18630 (gimple_simplified_result_is_gimple_val): Replace parameters with
18631 a const gimple_match_op *.
18632 (gimple_simplify): Replace code_helper * and tree * parameters with
18633 a gimple_match_op * parameter.
18634 (gimple_resimplify1): Replace code_helper *, tree and tree *
18635 parameters with a gimple_match_op * parameter.
18636 (gimple_resimplify2): Likewise.
18637 (gimple_resimplify3): Likewise.
18638 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
18639 parameters with a gimple_match_op * parameter.
18640 * gimple-match-head.c (gimple_simplify): Change prototypes of
18641 auto-generated functions to take a gimple_match_op * instead of
18642 separate code_helper * and tree * parameters. Make the same
18643 change in the top-level overload and update calls to the
18644 gimple_resimplify routines. Update calls to the auto-generated
18645 functions and to maybe_push_res_to_seq in the publicly-facing
18646 operation-specific gimple_simplify overloads.
18647 (gimple_match_op::MAX_NUM_OPS): Define.
18648 (gimple_resimplify1): Replace rcode and ops with a single res_op
18649 parameter. Update call to gimple_simplify.
18650 (gimple_resimplify2): Likewise.
18651 (gimple_resimplify3): Likewise.
18652 (mprts_hook): Replace parameters with a gimple_match_op *.
18653 (maybe_build_generic_op): Likewise.
18654 (build_call_internal): Replace type, nargs and ops with
18655 a gimple_match_op *.
18656 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
18657 with a single gimple_match_op *. Update calls to mprts_hook,
18658 build_call_internal and gimple_simplified_result_is_gimple_val.
18659 Factor out code that is common to the tree_code and combined_fn cases.
18660 * genmatch.c (expr::gen_transform): Replace tem_code and
18661 tem_ops with a gimple_match_op called tem_op. Update calls
18662 to the gimple_resimplify functions and maybe_push_res_to_seq.
18663 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
18664 res_ops. Update call to the gimple_resimplify functions.
18665 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
18666 (decision_tree::gen): Make the functions take a gimple_match_op *
18667 called res_op instead of separate res_code and res_ops parameters.
18668 Update call accordingly.
18669 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
18670 and ops with a single res_op parameter. Update calls to
18671 maybe_build_generic_op and maybe_push_res_to_seq.
18672 (fold_stmt_1): Update calls to gimple_simplify and
18673 replace_stmt_with_simplification.
18674 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
18675 and gimple_simplified_result_is_gimple_val.
18676 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
18678 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
18679 with a gimple_match_op *.
18680 (vn_nary_build_or_lookup): Likewise. Update call to
18681 vn_nary_build_or_lookup_1.
18682 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
18683 gimple_match_op *. Update calls to the gimple_resimplify routines
18684 and to gimple_simplified_result_is_gimple_val.
18685 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
18686 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
18687 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
18688 (visit_nary_op): Likewise.
18689 (visit_reference_op_load): Likewise.
18691 2018-05-23 Luis Machado <luis.machado@linaro.org>
18693 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
18694 modifier for printing the step amount.
18696 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
18699 * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
18702 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
18704 * doc/sourcebuild.texi (Endianness): New subsubsection.
18706 2018-05-23 Luis Machado <luis.machado@linaro.org>
18708 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18709 <prefetch_dynamic_strides>: New const bool field.
18710 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18711 prefetch_dynamic_strides.
18712 (exynosm1_prefetch_tune): Likewise.
18713 (thunderxt88_prefetch_tune): Likewise.
18714 (thunderx_prefetch_tune): Likewise.
18715 (thunderx2t99_prefetch_tune): Likewise.
18716 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
18718 (aarch64_override_options_internal): Update to set
18719 PARAM_PREFETCH_DYNAMIC_STRIDES.
18720 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18721 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18722 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18723 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18724 prefetch-dynamic-strides setting.
18726 2018-05-23 Luis Machado <luis.machado@linaro.org>
18728 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18729 <minimum_stride>: New const int field.
18730 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18731 minimum_stride field defaulting to -1.
18732 (exynosm1_prefetch_tune): Likewise.
18733 (thunderxt88_prefetch_tune): Likewise.
18734 (thunderx_prefetch_tune): Likewise.
18735 (thunderx2t99_prefetch_tune): Likewise.
18736 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
18737 <default_opt_level>: Set to 3.
18738 (aarch64_override_options_internal): Update to set
18739 PARAM_PREFETCH_MINIMUM_STRIDE.
18740 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18741 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18742 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18743 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18744 stride is constant and is below the minimum stride threshold.
18746 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18748 * config/arm/arm-cpus.in (mode26): Delete.
18749 (armv4): Delete mode26 reference.
18750 * config/arm/arm.c (arm_configure_build_target): Delete use of
18753 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
18755 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
18757 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
18758 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
18760 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
18761 (floatunsdidf2): Ditto.
18763 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
18765 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
18766 (fixuns_trunc<mode>si2_avx512f): Ditto.
18767 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
18768 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
18769 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
18771 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
18773 PR rtl-optimization/79985
18774 * df-scan.c (df_insn_refs_collect): Remove special case for
18775 global registers and asm statements.
18777 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
18779 * extend.texi (Global Register Variables): Rewrite the bullet list.
18780 Note that the register is available for allocation. Note that access
18781 via inline asm must use constraints. Add note about async-signal
18782 handlers. Remove paragraph about automagic register selection.
18784 2018-05-23 Richard Biener <rguenther@suse.de>
18786 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
18787 of fixed offset from memset VN.
18789 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
18791 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
18792 first_interp field.
18793 (alloc_cand_and_find_basis): Initialize first_interp field.
18794 (slsr_process_mul): Modify first_interp field.
18795 (slsr_process_add): Likewise.
18796 (slsr_process_cast): Modify first_interp field for each new
18798 (slsr_process_copy): Likewise.
18799 (dump_candidate): Dump first_interp field.
18800 (replace_mult_candidate): Process all interpretations, not just
18802 (replace_rhs_if_not_dup): Likewise.
18803 (replace_one_candidate): Likewise.
18805 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
18807 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
18809 (aarch64_needs_frame_chain): New function.
18810 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
18812 2018-05-23 Sudakshina Das <sudi.das@arm.com>
18815 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
18816 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
18817 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
18818 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
18819 as true for strict-align.
18820 (aarch64_can_inline_p): Perform checks even when callee has no
18821 attributes to check for strict alignment.
18822 * doc/extend.texi (AArch64 Function Attributes): Document
18824 * doc/invoke.texi: (AArch64 Options): Likewise.
18826 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
18828 PR tree-optimization/85853
18829 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
18830 the handling of the root of the node to...
18831 (vect_slp_analyze_node_operations_1): ...this new function,
18832 and run the whole thing with the child nodes' def types
18833 set according to their SLP node's def type.
18835 2018-05-23 Richard Biener <rguenther@suse.de>
18837 PR middle-end/85874
18838 * tree-data-ref.c (create_runtime_alias_checks): Defer
18839 and ignore overflow warnings.
18841 2018-05-23 Yury Gribov <tetra2005@gmail.com>
18843 PR tree-optimization/85822
18844 * tree-vrp.c (is_masked_range_test): Fix handling of negative
18847 2018-05-23 Richard Biener <rguenther@suse.de>
18849 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
18850 memset constants via native_interpret_expr.
18852 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
18855 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
18857 (cgraph_node::create_alias): Likewise.
18858 (cgraph_node::get_availability): Check ifunc_resolver instead
18859 of looking up ifunc attribute.
18860 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
18861 * varasm.c (do_assemble_alias): Likewise.
18862 (assemble_alias): Likewise.
18863 (default_binds_local_p_3): Likewise.
18864 * cgraph.h (cgraph_node): Add ifunc_resolver.
18865 (cgraph_node::only_called_directly_or_aliased_p): Return false
18866 for IFUNC resolver.
18867 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
18869 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
18870 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
18871 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
18872 instead of looking up ifunc attribute.
18874 2018-05-22 Luis Machado <luis.machado@linaro.org>
18876 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
18878 2018-05-22 Martin Sebor <msebor@redhat.com>
18880 PR middle-end/85359
18881 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
18882 only when expasion succeeds.
18883 (expand_builtin_strcmp): Same.
18884 (expand_builtin_strncmp): Same.
18886 2018-05-22 Martin Sebor <msebor@redhat.com>
18888 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
18890 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
18891 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18893 * config/aarch64/aarch64-ldpstp.md: Replace uses of
18894 aarch64_mem_pair_operand with memory_operand and delete operand swapping
18896 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18897 Add check for legitimate_address.
18898 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
18899 (aarch64_swap_ldrstr_operands): New.
18900 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
18903 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
18904 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
18906 * config/aarch64/aarch64.md: New patterns to generate stp
18908 (store_pair_sw, store_pair_dw): New patterns to generate stp for
18909 single words and double words.
18910 (load_pair_sw, load_pair_dw): Likewise.
18911 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
18913 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
18915 * config/aarch64/aarch64-ldpstp.md: Modify peephole
18916 for different mode ldpstp and add peephole for merged zero stores.
18917 Likewise for loads.
18918 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18920 (aarch64_gen_store_pair): Rename calls to match new patterns.
18921 (aarch64_gen_load_pair): Rename calls to match new patterns.
18922 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
18923 (load_pair<DREG:mode><DREG2:mode>): ... This.
18924 (store_pair<mode>): Rename to...
18925 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
18926 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
18927 New mode iterators.
18928 (V_INT_EQUIV): Handle SImode.
18929 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
18932 2018-05-22 Martin Sebor <msebor@redhat.com>
18935 * calls.c (maybe_warn_nonstring_arg): Use string length to set
18936 or ajust the presumed bound on an operation to avoid unnecessary
18939 2018-05-22 Martin Sebor <msebor@redhat.com>
18941 PR tree-optimization/85826
18942 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
18943 assuming that a DECL necesarily has a constant size.
18945 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
18947 PR middle-end/85862
18948 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
18950 2018-05-22 Richard Biener <rguenther@suse.de>
18952 PR tree-optimization/85834
18953 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
18954 non-constant and non-zero memset arguments.
18956 2018-05-22 Martin Liska <mliska@suse.cz>
18959 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
18961 2018-05-22 Richard Biener <rguenther@suse.de>
18963 PR tree-optimization/85863
18964 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
18965 comparisons when vectype is specified.
18966 (vectorizable_condition): Do not specify vectype for
18967 vect_is_simple_cond when SLP vectorizing.
18969 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
18972 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
18973 define __ibm128 as long double.
18974 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
18975 as a distinct type when IEEE 128-bit support is enabled.
18976 (init_float128_ieee): Fix up conversions between IFmode and IEEE
18977 128-bit types to use the correct functions.
18978 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
18979 convert between 128-bit floating point types that have different
18980 modes but the same representation, instead of using gen_lowpart to
18982 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
18984 (IFKF_reg): New attributes to give the register constraints for
18986 (extend<mode>tf2_internal): New insns to mark an explicit
18987 conversion between 128-bit floating point types that have a
18988 different mode but share the same representation.
18990 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
18992 PR tree-optimization/85814
18993 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
18994 a null return from get_strinfo when unsharing the next
18995 strinfo in the chain.
18997 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
19000 * varasm.c (weak_finish): Clean up weak_decls.
19002 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19004 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
19005 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
19006 UNSPEC_UADALP values.
19007 * config/aarch64/iterators.md (ABAL): New int iterator.
19010 (sur): Add mappings for the above.
19011 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
19013 (aarch64_<sur>abal<mode>_4): Likewise.
19014 (aarch64_<sur>adalp<mode>_3): Likewise.
19015 (<sur>sadv16qi): New define_expand.
19017 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
19019 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
19020 (*movdf_internal): Ditto.
19021 (*rcpsf2_sse): Ditto.
19022 (*rsqrtsf2_sse): Ditto.
19023 (*sqrt<mode>2_sse): Ditto.
19025 2018-05-21 Tamar Christina <tamar.christina@arm.com>
19027 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
19029 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
19030 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
19031 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
19032 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
19034 * config/aarch64/arm_neon.h: Likewise.
19035 * config/aarch64/iterators.md (VQ_I): New.
19037 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
19039 * config.gcc: Add arc/t-multilib-linux to tmake_file for
19041 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
19044 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
19046 * config/nds32/constraints.md (S): New constraint.
19047 * config/nds32/nds32.md (call_internal): Use constraint S.
19048 (call_value_internal): Likewise.
19049 (sibcall_internal): Likewise.
19050 (sibcall_value_internal): Likewise.
19052 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
19053 Chung-Ju Wu <jasonwucj@gmail.com>
19055 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
19056 into consideration.
19058 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
19059 Chung-Ju Wu <jasonwucj@gmail.com>
19061 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
19062 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
19063 (nds32_rtx_costs_impl): Simplify.
19064 (nds32_address_cost_impl): Simplify.
19065 (nds32_init_rtx_costs): New function.
19066 (nds32_rtx_costs_speed_prefer): Likewise.
19067 (nds32_rtx_costs_size_prefer): Likewise.
19068 (nds32_address_cost_speed_prefer): Likewise.
19069 (nds32_address_cost_speed_fwprop): Likewise.
19070 (nds32_address_cost_size_prefer): Likewise.
19071 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
19072 * config/nds32/nds32.c (nds32_option_override): Use
19073 nds32_init_rtx_costs function.
19075 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
19077 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
19078 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
19079 (TARGET_PIPELINE_N8): Likewise.
19080 (TARGET_PIPELINE_N10): Likewise.
19081 (TARGET_PIPELINE_N13): Likewise.
19082 (TARGET_PIPELINE_GRAYWOLF): Likewise.
19084 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
19086 * config/nds32/nds32-fpu.md: Update copyright year.
19088 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
19090 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
19092 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
19094 * config/nds32/nds32.c
19095 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
19096 * config/nds32/nds32.opt (minline-asm-r15): New option.
19098 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
19100 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
19102 * config/nds32/nds32.md (abssi2): New pattern.
19104 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
19106 * config/i386/i386.md (rex64namesuffix): New mode attribute.
19107 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
19108 Merge insn pattern from sse_cvtsi2ss<round_name> and
19109 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
19110 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
19111 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
19112 using SWI48 mode iterator.
19113 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
19114 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
19115 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
19116 pattern from sse_cvttss2si<round_saeonly_name>
19117 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
19118 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
19119 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
19120 using SWI48 mode iterator.
19121 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
19122 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
19123 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
19124 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
19125 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
19126 using SWI48 mode iterator.
19127 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
19128 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
19129 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
19130 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
19131 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
19132 SWI48 mode iterator.
19133 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
19134 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
19135 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
19136 pattern from sse_cvttsd2si<round_saeonly_name>
19137 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
19139 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
19141 * config/nds32/nds32-md-auxiliary.c
19142 (nds32_valid_smw_lwm_base_p): Refine.
19143 (nds32_output_smw_single_word): Refine.
19144 (nds32_output_smw_double_word): New.
19145 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
19147 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
19149 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
19150 (nds32_output_stack_pop): Refine.
19151 (nds32_expand_unaligned_load): Refine.
19152 (nds32_expand_unaligned_store): Refine.
19154 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
19155 Chung-Ju Wu <jasonwucj@gmail.com>
19157 * config/nds32/constants.md: Add TP_REGNUM constant.
19158 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
19159 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
19161 * config/nds32/nds32-doubleword.md: Consider flag_pic.
19162 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
19163 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
19164 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
19165 and PIC code generation.
19166 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
19168 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
19170 * config/nds32/nds32.md: Support TLS and PIC.
19171 * config/nds32/nds32.c: Support TLS and PIC.
19172 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
19173 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
19176 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
19178 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
19179 mode with E_ prefix.
19181 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
19182 Chung-Ju Wu <jasonwucj@gmail.com>
19184 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
19185 * config/nds32/nds32-md-auxiliary.c
19186 (symbolic_reference_mentioned_p): New.
19187 (nds32_legitimize_ict_address): New.
19188 (nds32_expand_ict_move): New.
19189 (nds32_indirect_call_referenced_p): New.
19190 (nds32_symbol_binds_local_p): Delete.
19191 (nds32_long_call_p): Modify.
19192 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
19193 * config/nds32/nds32-protos.h
19194 (symbolic_reference_mentioned_p): Declare.
19195 (nds32_legitimize_ict_address): Declare.
19196 (nds32_expand_ict_move): Declare.
19197 (nds32_indirect_call_referenced_p): Declare.
19198 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
19199 (nds32_relax_group): Use nds32_ict_const_p as condition.
19200 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
19201 (nds32_asm_file_start): Output ict_model directive in asm code.
19202 (nds32_legitimate_address_p): Consider indirect call.
19203 (nds32_print_operand): Consider indirect call.
19204 (nds32_print_operand_address): Consider indirect call.
19205 (nds32_insert_attributes): Handle "indirect_call" attribute.
19206 (TARGET_LEGITIMATE_ADDRESS_P): Define.
19207 (TARGET_LEGITIMATE_CONSTANT_P): Define.
19208 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
19209 (TARGET_DELEGITIMIZE_ADDRESS): Define.
19210 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
19211 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
19212 (TARGET_ICT_MODEL_SMALL): Define.
19213 (TARGET_ICT_MODEL_LARGE): Define.
19214 * config/nds32/nds32.md (movsi): Consider ict model.
19215 (call, call_value): Consider ict model.
19216 (sibcall, sibcall_value): Consider ict model.
19217 * config/nds32/nds32.opt (mict-model): New option.
19218 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
19221 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
19222 Monk Chiang <sh.chiang04@gmail.com>
19223 Jim Wilson <jimw@sifive.com>
19225 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
19226 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
19227 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
19228 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
19229 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
19230 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
19231 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
19232 compute save_libcall_adjustment properly.
19233 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
19234 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
19235 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
19236 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
19237 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
19238 (ABI_SPEC): Handle mabi=ilp32e.
19239 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
19240 (RVE): Add RVE mask.
19241 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
19242 <-march>: Add rv32e as an example.
19244 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
19247 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
19248 (intra_create_variable_infos): Handle C++ constructors.
19250 2018-05-18 Martin Liska <mliska@suse.cz>
19252 * passes.def: Remove a redundant pass.
19254 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
19257 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
19259 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19261 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
19263 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
19265 (armv2, armv2a, armv3, armv3m): Delete architectures.
19266 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
19267 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
19268 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
19270 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
19271 (*mulsidi3adddi): Likewise.
19272 (mulsidi3): Likewise.
19273 (*mulsidi3_nov6): Likewise.
19274 (umulsidi3): Likewise.
19275 (umulsidi3_nov6): Likewise.
19276 (umaddsidi4): Likewise.
19277 (*umulsidi3adddi): Likewise.
19278 (smulsi3_highpart): Likewise.
19279 (*smulsi3_highpart_nov6): Likewise.
19280 (umulsi3_highpart): Likewise.
19281 (*umulsi3_highpart_nov6): Likewise.
19282 * config/arm/arm.h (arm_arch3m): Delete.
19283 * config/arm/arm.c (arm_arch3m): Delete.
19284 (arm_option_override_internal): Update armv3-related comment.
19285 (arm_configure_build_target): Delete use of isa_bit_mode32.
19286 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
19287 (arm_rtx_costs_internal): Delete check of arm_arch3m.
19288 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
19289 (mulsa3): Likewise.
19290 (mulusa3): Likewise.
19291 * config/arm/arm-protos.h (arm_arch3m): Delete.
19292 * config/arm/arm-tables.opt: Regenerate.
19293 * config/arm/arm-tune.md: Likewise.
19294 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
19295 deleted architectures.
19297 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19299 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
19300 (armv5t, armv5te): New features.
19301 (ARMv5, ARMv5e): Delete fgroups.
19302 (ARMv5t, ARMv5te): Adjust for above changes.
19303 (ARMv6m): Likewise.
19304 (armv5, armv5e): Delete arches.
19305 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
19307 (*call_reg_arm): Likewise.
19308 (*call_value_reg_armv5): Likewise.
19309 (*call_value_reg_arm): Likewise.
19310 (*call_symbol): Likewise.
19311 (*call_value_symbol): Likewise.
19312 (*sibcall_insn): Likewise.
19313 (*sibcall_value_insn): Likewise.
19314 (clzsi2): Likewise.
19315 (prefetch): Likewise.
19316 (define_split and define_peephole2 dependent on arm_arch5):
19318 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
19320 (TARGET_ARM_QBIT): Likewise.
19321 (TARGET_DSP_MULTIPLY): Likewise.
19322 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
19323 (arm_arch5, arm_arch5e): Delete.
19324 (arm_arch5t, arm_arch5te): Declare.
19325 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
19326 (arm_arch5t): Declare.
19327 (arm_option_reconfigure_globals): Update for the above.
19328 (arm_options_perform_arch_sanity_checks): Update comment, replace
19329 use of arm_arch5 with arm_arch5t.
19330 (use_return_insn): Likewise.
19331 (arm_emit_call_insn): Likewise.
19332 (output_return_instruction): Likewise.
19333 (arm_final_prescan_insn): Likewise.
19334 (arm_coproc_builtin_available): Likewise.
19335 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
19336 arm_arch5e with arm_arch5t and arm_arch5te.
19337 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
19338 (arm_arch5t, arm_arch5te): Declare.
19339 * config/arm/arm-tables.opt: Regenerate.
19340 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
19341 * config/arm/t-multilib: Likewise.
19342 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
19343 instead of arm_arch5.
19344 (*call_reg_thumb1): Likewise.
19345 (*call_value_reg_thumb1_v5): Likewise.
19346 (*call_value_reg_thumb1): Likewise.
19347 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
19349 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
19351 2018-05-18 Martin Liska <mliska@suse.cz>
19353 PR gcov-profile/84846
19354 * doc/gcov.texi: Document -t option of gcov tool.
19356 2018-05-18 Martin Liska <mliska@suse.cz>
19358 PR gcov-profile/84846
19359 * gcov.c (print_usage): Add new -t option.
19360 (process_args): Handle the option.
19361 (generate_results): Use stdout as output when requested by
19364 2018-05-18 Martin Liska <mliska@suse.cz>
19366 PR gcov-profile/84846
19367 * coverage.c (coverage_init): Write PWD to .gcno file.
19368 * doc/gcov.texi: Document how working directory is printed.
19369 * gcov-dump.c (dump_gcov_file): Print PWD.
19370 * gcov.c (output_intermediate_file): Likewise.
19371 (read_graph_file): Read PWD string.
19372 (output_lines): Print PWD.
19374 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19376 PR middle-end/85817
19377 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
19378 for retval and return false if all args to phi are zero.
19380 2018-05-18 Richard Biener <rguenther@suse.de>
19382 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
19384 (evrp_dom_walker::before_dom_children): Call it.
19386 2018-05-18 Richard Biener <rguenther@suse.de>
19388 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
19389 results when processing array refs with variable index.
19391 2018-05-18 Toon Moene <toon@moene.org>
19393 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
19394 directly after that of -floop-interchange. Indicate that both
19395 options are enabled by default when specifying -O3.
19397 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19399 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
19400 iterator. Delete separate integer-mode vec_set<mode> expander.
19401 (aarch64_simd_vec_setv2di): Delete.
19402 (vec_setv2di): Delete.
19403 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
19404 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
19405 the "w, r" alternative.
19407 2018-05-18 Martin Liska <mliska@suse.cz>
19409 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
19410 * tree-pass.h (make_pass_lower_switch_O0): New function.
19411 * tree-switch-conversion.c (node_has_low_bound): Remove.
19412 (node_has_high_bound): Likewise.
19413 (node_is_bounded): Likewise.
19414 (class pass_lower_switch): Make it a template type and create
19416 (pass_lower_switch::execute): Add template argument.
19417 (make_pass_lower_switch): New function.
19418 (make_pass_lower_switch_O0): New function.
19419 (do_jump_if_equal): Remove.
19420 (emit_case_nodes): Simplify to just handle all 3 cases and leave
19421 all the hard work to tree optimization passes.
19423 2018-05-18 Martin Liska <mliska@suse.cz>
19425 * dbgcnt.c (limit_low): Renamed from limit.
19426 (limit_high): New variable.
19427 (dbg_cnt_is_enabled): Check for upper limit.
19428 (dbg_cnt): Adjust dumping.
19429 (dbg_cnt_set_limit_by_index): Add new argument for high
19431 (dbg_cnt_set_limit_by_name): Likewise.
19432 (dbg_cnt_process_single_pair): Parse new format.
19433 (dbg_cnt_process_opt): Use strtok.
19434 (dbg_cnt_list_all_counters): Remove 'value' and add
19436 * doc/invoke.texi: Document changes.
19438 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
19440 * doc/sourcebuild.texi (scalar_all_fma): Document.
19441 * tree.def (FMA_EXPR): Delete.
19442 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
19443 * internal-fn.c (ternary_direct): New macro.
19444 (expand_ternary_optab_fn): Likewise.
19445 (direct_ternary_optab_supported_p): Likewise.
19446 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
19447 * builtins.c (fold_builtin_fma): Delete.
19448 (fold_builtin_3): Don't call it.
19449 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
19450 * expr.c (expand_expr_real_2): Likewise.
19451 * fold-const.c (operand_equal_p): Likewise.
19452 (fold_ternary_loc): Likewise.
19453 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
19454 * gimple.c (DEFTREECODE): Likewise.
19455 * gimplify.c (gimplify_expr): Likewise.
19456 * optabs-tree.c (optab_for_tree_code): Likewise.
19457 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19458 * tree-eh.c (operation_could_trap_p): Likewise.
19459 (stmt_could_throw_1_p): Likewise.
19460 * tree-inline.c (estimate_operator_cost): Likewise.
19461 * tree-pretty-print.c (dump_generic_node): Likewise.
19462 (op_code_prio): Likewise.
19463 * tree-ssa-loop-im.c (stmt_cost): Likewise.
19464 * tree-ssa-operands.c (get_expr_operands): Likewise.
19465 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
19466 * fold-const-call.h (fold_fma): Delete.
19467 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
19468 CFN_FNMA and CFN_FNMS.
19469 (fold_fma): Delete.
19470 * genmatch.c (combined_fn): New enum.
19471 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
19472 (commutative_op): New function.
19473 (commutate): Use it. Handle more than 2 operands.
19474 (dt_operand::gen_gimple_expr): Use commutative_op.
19475 (parser::parse_expr): Allow :c to be used with non-binary
19476 operators if the commutative operand is known.
19477 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
19478 CFN_FMS, CFN_FNMA and CFN_FNMS.
19479 (backprop::process_assign_use): Remove FMA_EXPR handling.
19480 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
19481 (gen_hsa_fma): New function.
19482 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
19483 IFN_FNMA and IFN_FNMS.
19484 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
19485 * gimple-fold.h (follow_all_ssa_edges): Declare.
19486 * gimple-fold.c (follow_all_ssa_edges): New function.
19487 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
19488 gimple_build interface and use follow_all_ssa_edges to fold the result.
19489 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
19490 instead of checking for optabs directly.
19491 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
19492 rather than FMA_EXPRs.
19493 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
19494 call to IFN_FMA instead of an FMA_EXPR.
19496 2018-05-17 Jim Wilson <jimw@sifive.com>
19498 * expr.c (do_tablejump): When converting index to Pmode, if we have a
19499 sign extended promoted subreg, and the range does not have the sign bit
19500 set, then do a sign extend.
19502 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
19503 test, check for sign extended subreg and/or constant operands, and
19504 do a sign extend in that case.
19506 2018-05-17 Steve Ellcey <sellcey@cavium.com>
19508 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
19509 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
19511 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
19512 Change logics_shift_reg to logics_shift_imm.
19513 (thunderx2t99_fp_loadpair_basic): Delete.
19514 (thunderx2t99_fp_storepair_basic): Delete.
19515 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
19516 (thunderx2t99_asimd_polynomial): Delete.
19517 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
19518 and neon_fp_mul_d_scalar_q.
19519 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
19520 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
19521 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
19522 (thunderx2t99_asimd_lut): Add missing tbl types.
19523 (thunderx2t99_asimd_ext): Delete.
19524 (thunderx2t99_asimd_load1_1_mult): Delete.
19525 (thunderx2t99_asimd_load1_2_mult): Delete.
19526 (thunderx2t99_asimd_load1_ldp): New.
19527 (thunderx2t99_asimd_load1): New.
19528 (thunderx2t99_asimd_load2): Add missing *load2* types.
19529 (thunderx2t99_asimd_load3): New.
19530 (thunderx2t99_asimd_load4): New.
19531 (thunderx2t99_asimd_store1_1_mult): Delete.
19532 (thunderx2t99_asimd_store1_2_mult): Delete.
19533 (thunderx2t99_asimd_store2_mult): Delete.
19534 (thunderx2t99_asimd_store2_onelane): Delete.
19535 (thunderx2t99_asimd_store_stp): New.
19536 (thunderx2t99_asimd_store1): New.
19537 (thunderx2t99_asimd_store2): New.
19538 (thunderx2t99_asimd_store3): New.
19539 (thunderx2t99_asimd_store4): New.
19541 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
19543 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
19544 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
19546 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
19547 Segher Boessenkool <segher@kernel.crashing.org>
19550 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
19553 2018-05-17 Richard Biener <rguenther@suse.de>
19555 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
19556 for pruning loop and prune defs feeding only already visited PHIs.
19558 2018-05-17 Richard Biener <rguenther@suse.de>
19560 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
19562 2018-05-17 Bin Cheng <bin.cheng@arm.com>
19563 Richard Biener <rguenther@suse.de>
19565 PR tree-optimization/85793
19566 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
19567 for VMAT_ELEMENTWISE.
19569 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
19571 * internal-fn.h (lookup_internal_fn): Declare
19572 * internal-fn.c (lookup_internal_fn): New function.
19573 * gimple.c (gimple_build_call_from_tree): Handle calls to
19574 internal functions.
19575 * gimple-pretty-print.c (dump_gimple_call): Print "." before
19576 internal function names.
19577 * tree-pretty-print.c (dump_generic_node): Likewise.
19578 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
19580 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
19582 * gimple-fold.h (gimple_build): Make the function forms take
19583 combined_fn rather than built_in_function.
19584 (gimple_simplify): Likewise.
19585 * gimple-match-head.c (gimple_simplify): Likewise.
19586 * gimple-fold.c (gimple_build): Likewise.
19587 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
19588 rather than gimple_build_call_internal.
19589 (get_initial_defs_for_reduction): Likewise.
19590 (vect_create_epilog_for_reduction): Likewise.
19591 (vectorizable_live_operation): Likewise.
19593 2018-05-17 Martin Liska <mliska@suse.cz>
19595 * gimple-ssa-sprintf.c (format_directive): Do not use
19596 space in between 'G_' and '('.
19598 2018-05-17 Jakub Jelinek <jakub@redhat.com>
19601 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
19602 even if the mask is not all ones.
19605 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
19607 (ix86_gimple_fold_builtin): Likewise.
19610 * config/i386/i386.c: Include tree-vector-builder.h.
19611 (ix86_vector_shift_count): New function.
19612 (ix86_fold_builtin): Fold shift builtins by scalar count.
19613 (ix86_gimple_fold_builtin): Likewise.
19615 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
19616 _mm512_setzero): New intrinsics.
19618 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
19619 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19621 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
19622 code generation for cases where splatting a value is not useful.
19623 * simplify-rtx.c (simplify_ternary_operation): Simplify
19624 vec_merge across a vec_duplicate and a paradoxical subreg forming
19625 a vector mode to a vec_concat.
19627 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
19629 * config.gcc: Support "goldmont-plus".
19630 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19632 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19633 PROCESSOR_GOLDMONT_PLUS.
19634 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
19635 (processor_target_table): Add "goldmont-plus".
19636 (PTA_GOLDMONT_PLUS): Define.
19637 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
19638 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
19639 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
19640 (fold_builtin_cpu): Add "goldmont-plus".
19641 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
19642 (ix86_option_override_internal): Add "goldmont-plus".
19643 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
19644 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
19645 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
19646 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
19648 2018-05-17 Richard Biener <rguenther@suse.de>
19650 PR tree-optimization/85757
19651 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
19652 remove defs that only feed that PHI from further processing.
19654 2018-05-16 Jim Wilson <jimw@sifive.com>
19656 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
19658 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
19660 2018-05-16 Mark Wielaard <mark@klomp.org>
19662 * dwarf2out.c (count_index_strings): New function.
19663 (output_indirect_strings): Call count_index_strings and generate
19664 header for dwarf_version >= 5.
19666 2018-05-16 Mark Wielaard <mark@klomp.org>
19668 * dwarf2out.c (dwarf_FORM): New function.
19669 (set_indirect_string): Use dwarf_FORM.
19670 (reset_indirect_string): Likewise.
19671 (size_of_die): Likewise.
19672 (value_format): Likewise.
19673 (output_die): Likewise.
19674 (add_skeleton_AT_string): Likewise.
19675 (output_macinfo_op): Likewise.
19676 (index_string): Likewise.
19677 (output_index_string_offset): Likewise.
19678 (output_index_string): Likewise.
19679 (count_index_strings): Likewise.
19681 2018-05-16 Carl Love <cel@us.ibm.com>
19683 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19684 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
19686 2018-05-16 Martin Jambor <mjambor@suse.cz>
19688 * ipa-prop.c (ipa_free_all_edge_args): Remove.
19689 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
19691 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
19693 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
19694 (fnma<mode>4): Likewise.
19695 (fms<mode>4): Likewise.
19696 (fnms<mode>4): Likewise.
19697 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
19698 (aarch64_fnma<mode>4): Likewise.
19699 (aarch64_fms<mode>4): Likewise.
19700 (aarch64_fnms<mode>4): Likewise.
19701 (aarch64_fnmadd<mode>4): Likewise.
19703 2018-05-16 Jason Merrill <jason@redhat.com>
19705 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
19707 2018-05-16 Richard Biener <rguenther@suse.de>
19709 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
19710 (dump_stmt_cost): Declare.
19711 (add_stmt_cost): Dump cost we add.
19712 (add_stmt_costs): New function.
19713 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
19714 No longer exported.
19715 (vect_analyze_stmt): Adjust prototype.
19716 (vectorizable_condition): Likewise.
19717 (vectorizable_live_operation): Likewise.
19718 (vectorizable_reduction): Likewise.
19719 (vectorizable_induction): Likewise.
19720 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
19721 cost vector to pass to vectorizable_ and record afterwards.
19722 (vect_model_reduction_cost): Take cost vector argument and adjust.
19723 (vect_model_induction_cost): Likewise.
19724 (vectorizable_reduction): Likewise.
19725 (vectorizable_induction): Likewise.
19726 (vectorizable_live_operation): Likewise.
19727 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
19728 SLP_TREE_NUMBER_OF_VEC_STMTS.
19729 (vect_analyze_slp_cost_1): Remove.
19730 (vect_analyze_slp_cost): Likewise.
19731 (vect_slp_analyze_node_operations): Take visited args and
19732 a target cost vector. Avoid processing already visited stmt sets.
19733 (vect_slp_analyze_operations): Use a local cost vector to gather
19734 costs and register those of non-discarded instances.
19735 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
19736 (vect_schedule_slp_instance): Remove copying of
19737 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
19739 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
19740 adding cost. Record cost entry location.
19741 (vect_prologue_cost_for_slp_op): Function to compute cost of
19742 a constant or invariant generated for SLP vect in the prologue,
19743 split out from vect_analyze_slp_cost_1.
19744 (vect_model_simple_cost): Make static. Adjust for SLP costing.
19745 (vect_model_promotion_demotion_cost): Likewise.
19746 (vect_model_store_cost): Likewise, make static.
19747 (vect_model_load_cost): Likewise.
19748 (vectorizable_bswap): Add cost vector arg and adjust.
19749 (vectorizable_call): Likewise.
19750 (vectorizable_simd_clone_call): Likewise.
19751 (vectorizable_conversion): Likewise.
19752 (vectorizable_assignment): Likewise.
19753 (vectorizable_shift): Likewise.
19754 (vectorizable_operation): Likewise.
19755 (vectorizable_store): Likewise.
19756 (vectorizable_load): Likewise.
19757 (vectorizable_condition): Likewise.
19758 (vectorizable_comparison): Likewise.
19759 (can_vectorize_live_stmts): Likewise.
19760 (vect_analyze_stmt): Likewise.
19761 (vect_transform_stmt): Adjust calls to vectorizable_*.
19762 * tree-vectorizer.c: Include gimple-pretty-print.h.
19763 (dump_stmt_cost): New function.
19765 2018-05-16 Richard Biener <rguenther@suse.de>
19767 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
19768 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
19769 * tree-ssa-dse.c: Include tree-ssa-loop.h.
19770 (check_name): New callback.
19771 (dse_classify_store): Track cycles via a visited bitmap of PHI
19772 defs and simplify handling of in-loop and across loop dead stores
19773 and properly fail for loop-variant refs. Handle byte-tracking with
19774 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
19777 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
19779 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
19780 (vect_get_mask_type_for_stmt): Likewise.
19781 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
19783 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
19784 to determine the statement's vector type and the vector type that
19785 should be used for calculating nunits. Deal with cases in which
19786 the type has to be deferred.
19787 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
19788 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
19789 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
19790 (vect_determine_vf_for_stmt): New functions, split out from...
19791 (vect_determine_vectorization_factor): ...here.
19792 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
19793 (vect_get_mask_type_for_stmt): New functions, split out from
19794 vect_determine_vectorization_factor.
19796 2018-05-16 Richard Biener <rguenther@suse.de>
19798 * tree-cfg.c (verify_gimple_assign_ternary): Properly
19799 verify the [VEC_]COND_EXPR embedded comparison.
19801 2018-05-15 Martin Sebor <msebor@redhat.com>
19803 PR tree-optimization/85753
19804 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
19805 RECORD_TYPE in addition to ARRAY_TYPE.
19807 2018-05-15 Martin Sebor <msebor@redhat.com>
19809 PR middle-end/85643
19810 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
19812 2018-05-15 Richard Biener <rguenther@suse.de>
19814 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
19815 add by_clobber_p one. Change algorithm to collect all defs
19816 representing uses we need to walk and try reducing them to
19817 a single one before failing.
19818 (dse_dom_walker::dse_optimize_stmt): Adjust.
19820 2018-05-13 Mark Wielaard <mark@klomp.org>
19822 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
19823 (size_of_loc_descr): Likewise.
19824 (output_loc_operands): Likewise.
19825 (output_loc_operands_raw): Likewise.
19826 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
19827 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
19828 (hash_loc_operands): Likewise.
19829 (compare_loc_operands): Likewise.
19831 2018-05-14 Mark Wielaard <mark@klomp.org>
19833 * dwarf2out.c (count_index_addrs): New function.
19834 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
19836 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19838 PR tree-optimization/83648
19839 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
19840 return value as malloc candidate.
19842 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
19845 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
19846 param as true in call to suggest_attribute.
19848 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
19850 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
19851 -mreadonly-in-sdata.
19853 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
19855 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
19857 (aarch64_crypto_aesd_fused): Likewise.
19859 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
19861 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
19862 (movsi_aarch64): Likewise.
19863 (load_pairsi): Likewise.
19864 (load_pairdi): Likewise.
19865 (store_pairsi): Likewise.
19866 (store_pairdi): Likewise.
19867 (load_pairsf): Likewise.
19868 (load_pairdf): Likewise.
19869 (store_pairsf): Likewise.
19870 (store_pairdf): Likewise.
19871 (zero_extend): Likewise.
19872 (trunc): Swap alternatives.
19873 (fcvt_target): Add '?' to prefer w over r.
19875 2018-05-14 Jakub Jelinek <jakub@redhat.com>
19878 * config/i386/i386.md: Disallow non-commutative arithmetics in
19879 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
19880 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
19881 in the peephole2 before it.
19883 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
19885 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
19886 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
19887 (ix86_handle_option): Handle -mcldemote.
19888 * config.gcc: New header.
19889 * config/i386/cldemoteintrin.h: New file.
19890 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
19891 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19893 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19894 OPTION_MASK_ISA_CLDEMOTE.
19895 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
19896 (ix86_valid_target_attribute_inner_p): Ditto.
19897 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
19898 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
19899 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
19900 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
19901 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
19903 * config/i386/i386.opt: Add -mcldemote.
19904 * config/i386/x86intrin.h: New header.
19905 * doc/invoke.texi: Add -mcldemote.
19907 2018-05-14 Richard Biener <rguenther@suse.de>
19909 * doc/match-and-simplify.texi: Adjust :s documentation.
19911 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
19913 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
19914 intended memcpy size.
19915 (REORDER_45): Likewise.
19917 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
19919 * sort.cc: New file.
19920 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
19921 * vec.c (qsort_chk): Use gcc_qsort.
19922 * Makefile.in (OBJS-libcommon): Add sort.o.
19923 (build/sort.o): New target. Use it...
19924 (BUILD_RTL): ... here, and...
19925 (build/gencfn-macros): ... here, and...
19926 (build/genmatch): ... here.
19928 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
19929 Chung-Ju Wu <jasonwucj@gmail.com>
19931 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
19932 * config/nds32/nds32-graywolf.md: New file.
19933 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
19934 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
19936 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
19937 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
19938 * config/nds32/nds32.md (pipeline_model): Add graywolf.
19939 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
19940 * config/nds32/pipelines.md: Include n15 settings.
19942 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
19943 Chung-Ju Wu <jasonwucj@gmail.com>
19945 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
19946 * config/nds32/nds32-n13.md: New file.
19947 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
19948 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
19950 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
19951 * config/nds32/nds32.md (pipeline_model): Add n13.
19952 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
19953 * config/nds32/pipelines.md: Include n13 settings.
19955 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
19956 Chung-Ju Wu <jasonwucj@gmail.com>
19958 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
19959 * config/nds32/nds32-n10.md: New file.
19960 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
19961 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
19963 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
19964 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
19965 * config/nds32/nds32.md (pipeline_model): Add n10.
19966 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
19967 * config/nds32/pipelines.md: Include n10 settings.
19969 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
19970 Kito Cheng <kito.cheng@gmail.com>
19971 Chung-Ju Wu <jasonwucj@gmail.com>
19973 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
19974 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19975 Add enum values for DSP extension instructions.
19976 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
19978 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
19979 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
19980 New code iterators.
19981 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
19982 * config/nds32/nds32-dspext.md: New file for DSP implementation.
19983 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
19984 * config/nds32/nds32-intrinsic.md: Likewise.
19985 * config/nds32/nds32_intrinsic.h: Likewise.
19986 * config/nds32/nds32-md-auxiliary.c: Likewise.
19987 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
19988 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
19989 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
19990 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
19991 * config/nds32/nds32-protos.h: New declarations for DSP extension.
19992 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
19993 TYPE_DMAC in switch statement.
19994 * config/nds32/nds32.c: New checking and implementation for DSP
19995 extension instructions.
19996 * config/nds32/nds32.h: Likewise.
19997 * config/nds32/nds32.md: Likewise.
19998 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
19999 * config/nds32/predicates.md: Implement new predicates for DSP
20002 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
20004 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
20005 Reformat alternatives and attributes so it is easier to identify
20006 which constraints/attributes go with which instruction.
20007 (mov<mode>_hardfloat32, FMOVE64): Likewise.
20008 (mov<mode>_softfloat32, FMOVE64): Likewise.
20009 (mov<mode>_hardfloat64, FMOVE64): Likewise.
20010 (mov<mode>_softfloat64, FMOVE64): Likewise.
20012 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
20014 * doc/extend.texi (PowerPC Built-in Functions): Rename this
20016 (Basic PowerPC Built-in Functions): The new name of the
20017 subsection previously known as "PowerPC Built-in Functions".
20018 (Basic PowerPC Built-in Functions Available on all Configurations):
20020 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
20021 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
20022 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
20023 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
20025 2018-05-11 Martin Jambor <mjambor@suse.cz>
20028 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
20031 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
20034 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
20036 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
20038 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
20039 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
20040 (ix86_handle_option): Handle -mwaitpkg.
20041 * config.gcc: New header.
20042 * config/i386/cpuid.h (bit_WAITPKG): New bit.
20043 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
20044 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
20046 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20047 OPTION_MASK_ISA_WAITPKG.
20048 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
20049 (ix86_option_override_internal): Add PTA_WAITPKG.
20050 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
20051 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
20052 IX86_BUILTIN_TPAUSE.
20053 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
20054 __builtin_ia32_umwait and __builtin_ia32_tpause.
20055 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
20056 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
20057 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
20058 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
20059 UNSPECV_TPAUSE): New.
20060 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
20061 * config/i386/i386.opt: Add -mwaitpkg.
20062 * config/i386/waitpkgintrin.h: New file.
20063 * config/i386/x86intrin.h: New header.
20064 * doc/invoke.texi: Add -mwaitpkg.
20066 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
20069 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
20071 (cortex-m0): Use armv6s-m isa.
20072 (cortex-m0plus): Likewise.
20073 (cortex-m1): Likewise.
20074 (cortex-m0.small-multiply): Likewise.
20075 (cortex-m0plus.small-multiply): Likewise.
20076 (cortex-m1.small-multiply): Likewise.
20078 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
20079 Jakub Jelinek <jakub@redhat.com>
20081 PR tree-optimization/85692
20082 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
20083 source permute as well.
20085 2018-05-11 Martin Liska <mliska@suse.cz>
20088 * doc/extend.texi: Document LLVM style format for no_sanitize
20091 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
20093 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
20094 mode_supports_vsx_dform_quad to mode_supports_dq_form.
20095 (mode_supports_vsx_dform_quad): Likewise.
20096 (mode_supports_vmx_dform): Move these functions to be next to the
20097 other mode_supports functions.
20098 (mode_supports_dq_form): Likewise.
20099 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
20100 mode_supports_dq_form.
20101 (reg_offset_addressing_ok_p): Likewise.
20102 (offsettable_ok_by_alignment): Likewise.
20103 (rs6000_legitimate_offset_address_p): Likewise.
20104 (legitimate_lo_sum_address_p): Likewise.
20105 (rs6000_legitimize_address): Likewise.
20106 (rs6000_legitimize_reload_address): Likewise.
20107 (rs6000_secondary_reload_inner): Likewise.
20108 (rs6000_preferred_reload_class): Likewise.
20109 (rs6000_output_move_128bit): Likewise.
20111 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
20113 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
20114 Generate SImode target register for null target.
20115 <case IX86_BUILTIN_XGETBV>: Ditto.
20116 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
20117 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
20119 2018-05-10 Carl Love <cel@us.ibm.com>
20121 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
20122 dcbtt and dcbtstt if operands[2] is 0.
20124 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
20127 * config/i386/sse.md (usadv64qi): New expander.
20129 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
20131 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
20132 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
20133 -maltivec=be support.
20134 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
20135 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
20136 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
20137 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
20138 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
20139 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
20140 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
20141 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
20142 altivec_vsumsws): Adjust.
20143 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
20144 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
20145 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
20147 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
20148 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
20149 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
20150 (altivec_lve<VI_char>x): Delete expand.
20151 (*altivec_lve<VI_char>x_internal): Rename to...
20152 (altivec_lve<VI_char>x): ... this.
20153 (altivec_lvxl_<mode>): Delete expand.
20154 (*altivec_lvxl_<mode>_internal): Rename to ...
20155 (altivec_lvxl_<mode>): ... this.
20156 (altivec_stvxl_<mode>): Delete expand.
20157 (*altivec_stvxl_<mode>_internal): Rename to ...
20158 (altivec_stvxl_<mode>): ... this.
20159 (altivec_stve<VI_char>x): Delete expand.
20160 (*altivec_stve<VI_char>x_internal): Rename to ...
20161 (altivec_stve<VI_char>x): ... this.
20162 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
20163 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
20164 reduc_plus_scal_<mode>): Adjust.
20165 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
20167 (rs6000_cpu_cpp_builtins): Adjust.
20168 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
20169 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
20170 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
20171 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
20172 -maltivec=be support.
20173 (rs6000_split_vec_extract_var): Adjust.
20174 (rs6000_split_v4si_init): Adjust.
20175 (swap_selector_for_mode): Delete.
20176 (altivec_expand_lvx_be, altivec_expand_stvx_be,
20177 altivec_expand_stvex_be): Delete.
20178 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
20179 -maltivec=be support.
20180 (rs6000_gimple_fold_builtin): Ditto.
20181 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
20183 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
20184 (TARGET_DIRECT_MOVE_64BIT): Adjust.
20185 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
20186 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
20187 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
20188 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
20189 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
20190 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
20191 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
20192 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
20193 anonymous split): Adjust.
20194 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
20195 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
20197 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
20199 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
20200 when --with-gxx-include-dir is also specified.
20201 * configure: Regenerate.
20203 2018-05-09 Jim Wilson <jimw@sifive.com>
20206 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
20207 * config/riscv/t-withmultilib: New.
20208 * config/riscv/withmultilib.h: New.
20209 * doc/install.texi: Document RISC-V --with-multilib-list support.
20211 2018-05-09 Richard Biener <rguenther@suse.de>
20213 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
20215 (vect_bb_vectorization_profitable_p): Adjust. Compute
20216 actual scalar cost using the cost vector and the add_stmt_cost
20219 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
20221 PR rtl-optimization/85645
20222 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
20223 in the REG_CFA_REGISTER note for LR, don't leave it empty.
20225 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
20227 PR rtl-optimization/85645
20228 * shrink-wrap.c (spread_components): Return a boolean saying if
20229 anything was changed.
20230 (try_shrink_wrapping_separate): Iterate spread_components until
20231 nothing changes anymore.
20233 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
20235 PR rtl-optimization/85645
20236 * regrename.c (build_def_use): Also kill the chains that include the
20237 destination of a REG_CFA_REGISTER note.
20239 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
20241 PR rtl-optimization/85645
20242 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
20243 insn that has a REG_CFA_REGISTER note.
20245 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
20247 * cfgexpand.c (expand_clobber): New function.
20248 (expand_gimple_stmt_1): Use it.
20249 * tree-vect-stmts.c (vect_clobber_variable): New function,
20251 (vectorizable_simd_clone_call): ...here.
20252 (vectorizable_store): Emit a clobber either side of an
20253 IFN_STORE_LANES sequence.
20254 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
20256 2018-05-09 Tom de Vries <tom@codesourcery.com>
20259 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
20260 (define_insn "trap_if_false"): Add exit after trap.
20262 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
20264 PR rtl-optimization/85638
20265 * bb-reorder.c: Include common/common-target.h.
20266 (create_forwarder_block): New function extracted from...
20267 (fix_up_crossing_landing_pad): ...here. Rename into...
20268 (dw2_fix_up_crossing_landing_pad): ...this.
20269 (sjlj_fix_up_crossing_landing_pad): New function.
20270 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
20271 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
20272 from both partitions and exit the loop after one iteration.
20274 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
20277 * doc/extend.texi (PowerPC Built-in Functions): Rename this
20279 (Basic PowerPC Built-in Functions): The new name of the
20280 subsection previously known as "PowerPC Built-in Functions".
20281 (Basic PowerPC Built-in Functions Available on all Configurations):
20283 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
20285 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
20287 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
20289 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
20292 2018-05-08 Jim Wilson <jimw@sifive.com>
20294 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
20295 (LD_EMUL_SUFFIX): New.
20296 (LINK_SPEC): Use it.
20298 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
20300 * doc/extend.texi (PowerPC Built-in Functions): Rename this
20302 (Basic PowerPC Built-in Functions): The new name of the
20303 subsection previously known as "PowerPC Built-in Functions".
20304 (Basic PowerPC Built-in Functions Available on all Configurations):
20306 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
20308 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
20310 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
20312 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
20315 2018-05-08 Jakub Jelinek <jakub@redhat.com>
20318 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
20319 after cmpelim optimization.
20321 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
20323 * config.gcc: Support "goldmont".
20324 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
20325 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20326 PROCESSOR_GOLDMONT.
20327 * config/i386/i386.c (m_GOLDMONT): Define.
20328 (processor_target_table): Add "goldmont".
20329 (PTA_GOLDMONT): Define.
20330 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
20331 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
20332 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
20333 (fold_builtin_cpu): Add "goldmont".
20334 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
20335 (ix86_option_override_internal): Add "goldmont".
20336 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
20337 (processor_type): Add PROCESSOR_GOLDMONT.
20338 * config/i386/i386.md: Add CPU "glm".
20339 * config/i386/glm.md: New file.
20340 * config/i386/x86-tune.def: Add m_GOLDMONT.
20341 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
20343 2018-05-08 Jakub Jelinek <jakub@redhat.com>
20346 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
20348 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
20349 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
20350 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
20353 * config/i386/i386.c (ix86_fold_builtin): Handle
20354 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
20357 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
20358 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
20360 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
20363 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
20364 (check_arch): Likewise.
20365 (check_fpu): Return the result rather than printing it.
20366 (end arch): Fix operator precedence.
20367 (end cpu): Likewise.
20368 (END): Print the result from check_fpu.
20370 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
20371 Alan Hayward <alan.hayward@arm.com>
20372 David Sherwood <david.sherwood@arm.com>
20374 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
20375 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
20376 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
20377 (*fcmuo<mode>_and): New patterns.
20379 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
20381 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
20382 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
20383 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
20384 (cmp_op, sve_imm_con): New code attributes.
20385 (SVE_COND_INT_CMP, imm_con): Delete.
20386 (cmp_op): Remove above unspecs from int attribute.
20387 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
20389 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
20390 comparison-specific unspecs.
20391 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
20392 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
20393 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
20394 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
20395 (*vec_fcm<cmp_op><mode>): Rename to...
20396 (*fcm<cmp_op><mode>): ...this and adjust likewise.
20397 (*vec_fcmuo<mode>): Rename to...
20398 (*fcmuo<mode>): ...this and adjust likewise.
20399 (*pred_fcm<cmp_op><mode>): New pattern.
20400 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
20401 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
20403 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
20405 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
20406 (aarch64_emit_sve_predicated_cond): New function.
20407 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
20408 (aarch64_emit_unspec_cond_or): Replace with...
20409 (aarch64_emit_sve_or_conds): ...this new function. Use
20410 aarch64_emit_sve_ptrue_op for the individual comparisons and
20411 aarch64_emit_binop to OR them together.
20412 (aarch64_emit_inverted_unspec_cond): Replace with...
20413 (aarch64_emit_sve_inverted_cond): ...this new function. Use
20414 aarch64_emit_sve_ptrue_op for the comparison and
20415 aarch64_emit_unop to invert the result.
20416 (aarch64_expand_sve_vec_cmp_float): Update after the above
20417 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
20419 2018-05-07 Nathan Sidwell <nathan@acm.org>
20421 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
20422 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
20423 (Backwards Compatibility): Likewise.
20425 2018-05-07 Luis Machado <luis.machado@linaro.org>
20429 2018-05-07 Luis Machado <luis.machado@linaro.org>
20431 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20432 <prefetch_dynamic_strides>: New const bool field.
20433 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20434 prefetch_dynamic_strides.
20435 (exynosm1_prefetch_tune): Likewise.
20436 (thunderxt88_prefetch_tune): Likewise.
20437 (thunderx_prefetch_tune): Likewise.
20438 (thunderx2t99_prefetch_tune): Likewise.
20439 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
20441 (aarch64_override_options_internal): Update to set
20442 PARAM_PREFETCH_DYNAMIC_STRIDES.
20443 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20444 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20445 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20446 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20447 prefetch-dynamic-strides setting.
20449 2018-05-07 Luis Machado <luis.machado@linaro.org>
20451 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20452 <minimum_stride>: New const int field.
20453 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20454 minimum_stride field.
20455 (exynosm1_prefetch_tune): Likewise.
20456 (thunderxt88_prefetch_tune): Likewise.
20457 (thunderx_prefetch_tune): Likewise.
20458 (thunderx2t99_prefetch_tune): Likewise.
20459 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
20460 (aarch64_override_options_internal): Update to set
20461 PARAM_PREFETCH_MINIMUM_STRIDE.
20462 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20463 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20464 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20465 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20466 stride is constant and is below the minimum stride threshold.
20468 2018-05-07 Luis Machado <luis.machado@linaro.org>
20470 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
20473 2018-05-07 Luis Machado <luis.machado@linaro.org>
20475 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20476 <prefetch_dynamic_strides>: New const bool field.
20477 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20478 prefetch_dynamic_strides.
20479 (exynosm1_prefetch_tune): Likewise.
20480 (thunderxt88_prefetch_tune): Likewise.
20481 (thunderx_prefetch_tune): Likewise.
20482 (thunderx2t99_prefetch_tune): Likewise.
20483 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
20485 (aarch64_override_options_internal): Update to set
20486 PARAM_PREFETCH_DYNAMIC_STRIDES.
20487 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20488 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20489 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20490 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20491 prefetch-dynamic-strides setting.
20493 2018-05-07 Luis Machado <luis.machado@linaro.org>
20495 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20496 <minimum_stride>: New const int field.
20497 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20498 minimum_stride field.
20499 (exynosm1_prefetch_tune): Likewise.
20500 (thunderxt88_prefetch_tune): Likewise.
20501 (thunderx_prefetch_tune): Likewise.
20502 (thunderx2t99_prefetch_tune): Likewise.
20503 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
20504 (aarch64_override_options_internal): Update to set
20505 PARAM_PREFETCH_MINIMUM_STRIDE.
20506 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20507 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20508 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20509 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20510 stride is constant and is below the minimum stride threshold.
20512 2018-05-06 Jakub Jelinek <jakub@redhat.com>
20515 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
20516 the type is addressable. Don't force op into register if it has
20519 2018-05-05 Roland McGrath <mcgrathr@google.com>
20522 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
20523 any section for which we don't know a specific type it should have,
20524 regardless of name. Previously this was done only for the exact
20525 names ".init_array", ".fini_array", and ".preinit_array".
20526 (default_elf_asm_named_section): Add comment about
20527 relationship with default_section_type_flags and SECTION_NOTYPE.
20528 (get_section): Don't consider it a type conflict if one side has
20529 SECTION_NOTYPE and the other doesn't, as long as neither has the
20530 SECTION_BSS et al used in the default_section_type_flags logic.
20532 2018-05-05 Tom de Vries <tom@codesourcery.com>
20535 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
20536 (workaround_barsyncs): New function.
20537 (nvptx_reorg): Use workaround_barsyncs.
20538 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
20539 (define_expand "nvptx_membar_cta"): New define_expand.
20540 (define_insn "*nvptx_membar_cta"): New insn.
20542 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
20544 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
20545 To improve optimization opportunities.
20546 * builtin-types.def: The new needed builtin types for the above.
20548 2018-05-04 Richard Biener <rguenther@suse.de>
20550 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
20551 * gimple-ssa-store-merging.c
20552 (imm_store_chain_info::output_merged_store): Remove redundant create,
20553 release split_store vector contents on failure.
20554 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
20555 scalar stmt vector on cache hit.
20557 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
20559 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
20561 * config.gcc (powerpc-xilinx-eabi*): Remove.
20562 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
20564 (fusion_addis_mem_combo_load): Ditto.
20565 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
20567 (rs6000_cpu_cpp_builtins): Ditto.
20568 * config/rs6000/rs6000-linux.c
20569 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
20570 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
20571 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
20573 (rs6000_setup_reg_addr_masks): Ditto.
20574 (rs6000_init_hard_regno_mode_ok): Ditto.
20575 (rs6000_option_override_internal): Ditto.
20576 (legitimate_lo_sum_address_p): Ditto.
20577 (rs6000_legitimize_address): Ditto.
20578 (rs6000_legitimize_reload_address): Ditto.
20579 (rs6000_legitimate_address_p): Ditto.
20580 (abi_v4_pass_in_fpr): Ditto.
20581 (setup_incoming_varargs): Ditto.
20582 (rs6000_gimplify_va_arg): Ditto.
20583 (rs6000_split_multireg_move): Ditto.
20584 (rs6000_savres_strategy): Ditto.
20585 (rs6000_emit_prologue_components): Ditto.
20586 (rs6000_emit_epilogue_components): Ditto.
20587 (rs6000_emit_prologue): Ditto.
20588 (rs6000_emit_epilogue): Ditto.
20589 (rs6000_elf_file_end): Ditto.
20590 (rs6000_function_value): Ditto.
20591 (rs6000_libcall_value): Ditto.
20592 * config/rs6000/rs6000.h: Ditto.
20593 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
20594 (TARGET_MINMAX): ... this. New.
20595 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
20596 * config/rs6000/rs6000.md: Remove Xilinx FP support.
20597 (*movsi_internal1_single): Delete.
20598 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
20599 mfpu=, mxilinx-fpu): Delete.
20600 * config/rs6000/singlefp.h: Delete.
20601 * config/rs6000/sysv4.h: Remove Xilinx FP support.
20602 * config/rs6000/t-rs6000: Ditto.
20603 * config/rs6000/t-xilinx: Delete.
20604 * config/rs6000/titan.md: Adjust for fp_type removal.
20605 * config/rs6000/vsx.md: Remove Xilinx FP support.
20606 (VStype_simple): Delete.
20607 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
20608 * config/rs6000/xfpu.h: Delete.
20609 * config/rs6000/xfpu.md: Delete.
20610 * config/rs6000/xilinx.h: Delete.
20611 * config/rs6000/xilinx.opt: Delete.
20612 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
20613 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
20615 2018-05-04 Tom de Vries <tom@codesourcery.com>
20618 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
20621 2018-05-04 Richard Biener <rguenther@suse.de>
20623 PR middle-end/85627
20624 * tree-complex.c (update_complex_assignment): We are always in SSA form.
20625 (expand_complex_div_wide): Likewise.
20626 (expand_complex_operations_1): Likewise.
20627 (expand_complex_libcall): Preserve EH info of the original stmt.
20628 (tree_lower_complex): Handle removed blocks.
20629 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
20630 on complex multiplication and division libcall builtins.
20632 2018-05-04 Richard Biener <rguenther@suse.de>
20634 PR middle-end/85574
20635 * fold-const.c (negate_expr_p): Restrict negation of operand
20636 zero of a division to when we know that can happen without
20638 (fold_negate_expr_1): Likewise.
20640 2018-05-04 Jakub Jelinek <jakub@redhat.com>
20643 * real.h (real_nextafter): Declare.
20644 * real.c (real_nextafter): New function.
20645 * fold-const-call.c (fold_const_nextafter): New function.
20646 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
20647 CASE_CFN_NEXTTOWARD.
20648 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
20649 even when arg1_mode is different from arg0_mode.
20651 2018-05-03 Nathan Sidwell <nathan@acm.org>
20653 * doc/extend.texi (Deprecated Features): Remove
20654 -ffriend-injection.
20655 (Backwards Compatibility): Likewise.
20656 * doc/invoke.texi (C++ Language Options): Likewise.
20657 (C++ Dialect Options): Likewise.
20659 2018-05-03 Jakub Jelinek <jakub@redhat.com>
20662 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
20663 _mm512_mask_mullox_epi64): New intrinsics.
20665 2018-05-03 Tom de Vries <tom@codesourcery.com>
20668 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20669 dump files): Add offload-tree.
20671 2018-05-03 Richard Biener <rguenther@suse.de>
20673 PR tree-optimization/85615
20674 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
20675 to loops not nested in BBs loop father to avoid creating multi-entry
20678 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
20680 PR tree-optimization/70291
20681 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
20682 arguments. Change return type to tree. Emit libcall as a new
20683 statement rather than replacing existing one when inplace_p is true.
20684 (expand_complex_multiplication_components): New function.
20685 (expand_complex_multiplication): Expand floating-point complex
20686 multiplication using the above.
20687 (expand_complex_division): Rename inner_type parameter to type.
20688 Update expand_complex_libcall call-site.
20689 (expand_complex_operations_1): Update expand_complex_multiplication
20690 and expand_complex_division call-sites.
20692 2018-05-02 Jakub Jelinek <jakub@redhat.com>
20695 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20696 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20697 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
20698 the highest significant bit of the shift count mask is clear. In
20699 check whether and[sq]i3 is needed verify that all significant bits
20700 of the shift count other than the highest are set.
20702 2018-05-02 Tom de Vries <tom@codesourcery.com>
20705 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
20706 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
20707 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
20708 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
20709 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
20710 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
20711 __builtin_goacc_parlevel_size.
20713 2018-05-02 Richard Biener <rguenther@suse.de>
20715 PR tree-optimization/85597
20716 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
20717 do not use split vect_get_vec_defs call but call vect_get_slp_defs
20720 2018-05-02 Tom de Vries <tom@codesourcery.com>
20723 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20724 dump files): Add ltrans-tree.
20726 2018-05-02 Tom de Vries <tom@codesourcery.com>
20729 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20730 dump files): Add wpa-ipa.
20732 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
20734 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
20735 powerpc*-*-linux*paired* target.
20736 * config/rs6000/750cl.h: Delete.
20737 * config/rs6000/paired.h: Delete.
20738 * config/rs6000/paired.md: Delete.
20739 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
20741 * config/rs6000/rs6000-builtin.def: Remove paired float support.
20742 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20743 comment. Remove paired float support.
20744 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
20745 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
20747 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
20748 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
20750 * config/rs6000/rs6000.c: Remove paired float support.
20751 (paired_expand_vector_init, paired_expand_vector_move,
20752 paired_emit_vector_compare, paired_emit_vector_cond_expr,
20753 (paired_expand_lv_builtin, paired_expand_stv_builtin,
20754 paired_expand_builtin, paired_expand_predicate_builtin,
20755 paired_init_builtins): Delete.
20756 * config/rs6000/rs6000.h: Remove paired float support.
20757 * config/rs6000/rs6000.md: Remove paired float support.
20758 (move_from_CR_ov_bit): Delete.
20759 * config/rs6000/rs6000.opt (mpaired): Delete.
20760 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
20761 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
20763 2018-05-02 Richard Biener <rguenther@suse.de>
20765 PR middle-end/85567
20766 * gimplify.c (gimplify_save_expr): When in SSA form allow
20767 SAVE_EXPRs to compute to SSA vars.
20769 2018-05-02 Jakub Jelinek <jakub@redhat.com>
20772 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20773 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20774 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
20775 clobber operands[2], instead use a new pseudo. Formatting fixes.
20777 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
20779 PR tree-optimization/85586
20780 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
20781 exit early for statements in the same group if the accesses are
20784 2018-05-02 Tom de Vries <tom@codesourcery.com>
20787 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
20790 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
20792 PR tree-optimization/85143
20793 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
20795 2018-05-01 Tom de Vries <tom@codesourcery.com>
20798 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
20799 not found" error message.
20801 2018-05-01 Tom de Vries <tom@codesourcery.com>
20804 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
20805 * vec.c (test_ordered_remove_if): New function.
20806 (vec_c_tests): Call test_ordered_remove_if.
20807 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
20808 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
20809 * tree-vect-patterns.c (vect_pattern_recog_1): Use
20810 VEC_ORDERED_REMOVE_IF.
20812 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
20814 PR tree-optimization/82665
20815 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
20816 pointer subtraction where arguments come from a memchr call.
20818 2018-05-01 Jakub Jelinek <jakub@redhat.com>
20820 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
20821 --push-state --as-needed and --pop-state instead of --as-needed and
20822 --no-as-needed if ld supports it.
20823 * configure: Regenerated.
20826 * doc/install.texi2html: Replace _002d with - and _002a with * in
20827 generated html files using sed.
20829 2018-04-30 David Malcolm <dmalcolm@redhat.com>
20832 * gcc-rich-location.c (blank_line_before_p): New function.
20833 (use_new_line): New function.
20834 (gcc_rich_location::add_fixit_insert_formatted): New function.
20835 * gcc-rich-location.h
20836 (gcc_rich_location::add_fixit_insert_formatted): New function.
20838 2018-04-30 David Malcolm <dmalcolm@redhat.com>
20840 * selftest.c (assert_streq): Rename "expected" and "actual" to
20841 "val1" and "val2". Extend NULL-handling to cover both inputs
20842 symmetrically, while still requiring both to be non-NULL for a pass.
20843 * selftest.h (assert_streq): Rename "expected" and "actual" to
20845 (ASSERT_EQ): Likewise.
20846 (ASSERT_EQ_AT): Likewise.
20847 (ASSERT_KNOWN_EQ): Likewise.
20848 (ASSERT_KNOWN_EQ_AT): Likewise.
20849 (ASSERT_NE): Likewise.
20850 (ASSERT_MAYBE_NE): Likewise.
20851 (ASSERT_MAYBE_NE_AT): Likewise.
20852 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
20853 the assertion to pass.
20854 (ASSERT_STREQ_AT): Likewise.
20856 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
20858 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
20859 interaction with -pie.
20861 2018-04-30 David Malcolm <dmalcolm@redhat.com>
20863 * selftest.h: Fix alphabetization of per-source-file selftest
20866 2018-04-30 Jason Merrill <jason@redhat.com>
20868 PR c++/61982 - dead stores to destroyed objects.
20869 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
20872 2018-04-30 Jason Merrill <jason@redhat.com>
20874 * tree.c (build_clobber): New.
20875 * tree.h: Declare it.
20876 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
20878 2018-04-30 David Malcolm <dmalcolm@redhat.com>
20880 * diagnostic-show-locus.c (layout::layout): Update for
20881 location_get_source_line returning a char_span.
20882 (struct char_span): Move to input.h.
20883 (struct correction): Update for fields in char_span becoming
20885 (struct source_line): Update for location_get_source_line
20886 returning a char_span.
20887 (layout::print_line): Likewise.
20888 * edit-context.c (edited_file::print_content): Likewise.
20889 (edited_file::print_diff_hunk): Likewise.
20890 (edited_file::print_run_of_changed_lines): Likewise.
20891 (edited_file::get_num_lines): Likewise.
20892 (edited_line::edited_line): Likewise.
20893 * final.c (asm_show_source): Likewise.
20894 * input.c (location_get_source_line): Convert return type
20895 from const char * to char_span, losing the final "line_len"
20897 (dump_location_info): Update for the above.
20898 (get_substring_ranges_for_loc): Likewise. Use a char_span
20899 when handling the literal within the line.
20900 (test_reading_source_line): Update for location_get_source_line
20901 returning a char_span.
20902 * input.h (class char_span): Move here from
20903 diagnostic-show-locus.c, converting from a struct to a class.
20904 Make data members private.
20905 (char_span::operator bool): New.
20906 (char_span::length): New.
20907 (char_span::get_buffer): New.
20908 (char_span::operator[]): New.
20909 (char_span::subspan): Make const.
20910 (char_span::xstrdup): New.
20911 (location_get_source_line): Convert return type from const char *
20912 to char_span, losing the final "line_size" param.
20914 2018-04-30 Jan Hubicka <jh@suse.cz>
20916 * lto-wrapper.c (ltrans_priorities): New static var.
20917 (cmp_priority): New.
20918 (run_gcc): Read priorities and if doing parallel build order
20919 the Makefile by them.
20921 2018-04-30 David Malcolm <dmalcolm@redhat.com>
20923 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
20925 2018-04-30 Richard Biener <rguenther@suse.de>
20927 * tree-cfg.c (verify_address): Remove base argument, add
20928 flag whether to check TREE_ADDRESSABLE and do that.
20929 (verify_expr): Remove.
20930 (verify_types_in_gimple_reference): Add pieces from verify_expr.
20931 (verify_gimple_assign_single): Likewise.
20932 (verify_gimple_switch): Likewise.
20933 (verify_expr_location_1): Dereference tp once. Add (disabled)
20934 piece from verify_expr.
20935 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
20937 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
20939 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
20941 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
20943 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
20944 (small_data_pattern): Likewise.
20945 (arc_rewrite_small_data): Likewise.
20946 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
20947 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
20948 (get_symbol_alignment): New function.
20949 (legitimate_small_data_address_p): Likewise.
20950 (legitimate_scaled_address): Update, call
20951 legitimate_small_data_address_p.
20952 (output_sdata): New static variable.
20953 (arc_print_operand): Update how we handle small data operands.
20954 (arc_print_operand_address): Likewise.
20955 (arc_legitimate_address_p): Update, use
20956 legitimate_small_data_address_p.
20957 (arc_rewrite_small_data_p): Remove.
20958 (arc_rewrite_small_data_1): Likewise.
20959 (arc_rewrite_small_data): Likewise.
20960 (small_data_pattern): Likewise.
20961 (compact_sda_memory_operand): Update to use
20962 legitimate_small_data_address_p and get_symbol_alignment.
20963 (prepare_move_operands): Don't rewite sdata pattern.
20964 (prepare_extend_operands): Remove.
20965 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
20967 (zero_extendqisi2): Likewise.
20968 (zero_extendhisi2): Likewise.
20969 (extendqihi2): Likewise.
20970 (extendqisi2): Likewise.
20971 (extendhisi2): Likewise.
20972 (addsi3): Likewise.
20973 (subsi3): Likewise.
20974 (andsi3): Likewise.
20975 * config/arc/constraints.md (Usd): Change it to memory constraint.
20977 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
20979 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
20980 as source of std instructions.
20981 * config/arc/arc.md (movsi_insn): Update pattern predicate to
20982 allow 6-bit constants as source for store instructions.
20983 (movdi_insn): Update instruction pattern to allow 6-bit constants
20984 as source for store instructions.
20986 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
20988 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
20990 2018-04-30 Nathan Sidwell <nathan@acm.org>
20991 Sandra Loosemore <sandra@codesourcery.com>
20993 * dumpfile.c (dump_open): Allow '-' for stdout.
20994 * doc/invoke.texi (Developer Options): Document dump filename
20995 determination early. Document stdin/stdout selection.
20997 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
20999 Microblaze Target: PIC data text relative
21001 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
21002 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
21004 * config/microblaze/microblaze.h (microblaze_constant_address_p):
21005 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
21006 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
21007 New addressing mode for data-text relative position indepenedent code.
21008 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
21009 'ADDRESS_SYMBOLIC_TXT_REL'.
21010 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
21011 (microblaze_legitimate_pic_operand): Exclude function calls from
21012 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
21013 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
21015 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
21016 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
21017 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
21018 for 'address + offset'.
21019 (microblaze_expand_prologue): Add new function prologue call for
21021 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
21022 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
21023 table in case of TARGET_PIC_DATA_TEXT_REL.
21024 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
21025 * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
21026 Add new macros 'UNSPEC_TEXT',
21027 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
21028 + exclude function calls from 'UNSPEC_PLT' in case of data text
21030 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
21031 new target hook for generating address diff vector tables in case of
21033 * doc/tm.texi : Regenerate.
21034 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
21035 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
21036 of addr diff vector generation.
21037 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
21038 target hook definition.
21039 * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
21040 Add default function for generate_pic_addr_diff_vec -> flag_pic.
21041 * doc/invoke.texi (Add new pic option): Add new microblaze pic
21042 option for data text relative.
21044 2018-04-30 Richard Biener <rguenther@suse.de>
21046 * tree-chrec.h (evolution_function_is_constant_p): Remove
21048 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
21050 2018-04-30 Richard Biener <rguenther@suse.de>
21053 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
21055 2018-04-30 Richard Biener <rguenther@suse.de>
21057 PR tree-optimization/28364
21058 PR tree-optimization/85275
21059 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
21060 copying first exit test.
21062 2018-04-28 Mark Wielaard <mark@klomp.org>
21064 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
21065 dwarf_version >= 5.
21066 (dwarf_AT): Handle DW_AT_addr_base.
21067 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
21069 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
21072 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
21073 (*ashl<dwi>3_doubleword_mask_1): Ditto.
21074 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
21075 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
21077 2018-04-28 Richard Biener <rguenther@suse.de>
21079 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
21080 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
21081 to reflect use. Only add interesting stmts.
21083 2018-04-27 Martin Jambor <mjambor@suse.cz>
21086 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
21087 the jump function allows for passing through aggregate values.
21089 2018-04-27 David Malcolm <dmalcolm@redhat.com>
21091 * input.h (in_system_header_at): Convert from macro to inline
21093 (from_macro_expansion_at): Likewise.
21094 (from_macro_definition_at): Likewise.
21096 2018-04-27 Jeff Law <law@redhat.com>
21098 * config.gcc: Mark tile* targets as deprecated/obsolete.
21100 2018-04-27 Richard Biener <rguenther@suse.de>
21102 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
21105 2018-04-27 Richard Biener <rguenther@suse.de>
21107 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
21109 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
21111 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
21112 with Yd constraint. Set "preferred_for_speed" attribute from
21113 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
21114 with Yd constraint.
21115 (*movdi_internal): Ditto.
21116 (movti_interunit splitters): Remove
21117 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
21118 (movdi_interunit splitters): Ditto.
21119 * config/i386/constraints.md (Ye): Remove.
21120 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
21122 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21125 * config/aarch64/constraints.md (Usg): Limit to 31.
21126 (Usj): Limit to 63.
21128 2018-04-27 Jakub Jelinek <jakub@redhat.com>
21130 PR tree-optimization/85529
21131 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
21132 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
21133 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
21134 zero extension or masking of the MSB bit.
21135 (optimize_range_tests): Add FIRST_BB argument, pass it through
21136 to optimize_range_tests_var_bound.
21137 (maybe_optimize_range_tests, reassociate_bb): Adjust
21138 optimize_range_tests callers.
21140 2018-04-26 Richard Biener <rguenther@suse.de>
21141 Jakub Jelinek <jakub@redhat.com>
21143 * cgraph.h (symbol_table): Just declare debug method here.
21144 * symtab.c (symbol_table::debug): Define.
21146 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
21148 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
21150 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
21152 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
21153 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
21154 (*movdi_internal): Substitute Yi and Yj constraint with x
21155 and Ym and Yn constraint with y constraint. Update "isa"
21156 attribute and set "preferred_for_speed" attribute from
21157 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
21158 (*movsi_internal): Ditto.
21159 (*movdf_internal): Ditto.
21160 (*movsf_internal): Ditto.
21161 (*zero_extendsidi2): Ditto.
21162 * config/i386/sse.md (vec_set<mode>_0): Ditto.
21163 (sse2_loadld): Ditto.
21164 (*vec_extract<ssevecmodelower>_0): Ditto.
21165 (*vec_extractv4si_0_zext_sse4): Ditto.
21166 (vec_concatv2di): Ditto.
21167 (*vec_dup<mode>): Ditto.
21168 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
21169 * config/i386/constraints.md (Yi): Remove.
21174 2018-04-26 Nathan Sidwell <nathan@acm.org>
21176 * dumpfile.c (dump_open): New.
21177 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
21178 (dump_finish): Detect stdio/stderr by value not name.
21180 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
21182 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
21184 2018-04-26 Tom de Vries <tom@codesourcery.com>
21187 * config/nvptx/nvptx.c (verify_neutering_jumps)
21188 (verify_neutering_labels): New function
21189 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
21191 2018-04-26 Tom de Vries <tom@codesourcery.com>
21194 * config/nvptx/nvptx.c (needs_neutering_p): New function.
21195 (nvptx_single): Use needs_neutering_p to skip over insns that do not
21198 2018-04-26 Richard Biener <rguenther@suse.de>
21199 Tom de Vries <tom@codesourcery.com>
21202 * lto-streamer-out.c (output_function): Fixup loops if required to match
21203 discovery done in the reader.
21205 2018-04-26 Richard Biener <rguenther@suse.de>
21207 PR tree-optimization/85116
21208 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
21209 have a loop exit from the single latch predecessor. Remove
21210 case of header with just condition.
21211 (ch_base::copy_headers): Exclude infinite loops from any
21213 (pass_ch::execute): Record exits.
21215 2018-04-26 Richard Biener <rguenther@suse.de>
21217 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
21218 prologue cost vector and pass it to vect_get_load_cost.
21219 (vect_get_peeling_costs_all_drs): Likewise.
21220 (vect_peeling_hash_get_lowest_cost): Likewise.
21221 (vect_enhance_data_refs_alignment): Likewise.
21223 2018-04-26 Richard Biener <rguenther@suse.de>
21225 PR middle-end/85450
21226 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
21227 checking of integer<->pointer conversions.
21228 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
21229 sign-/zero-extending pointer types.
21230 (expand_omp_for_static_chunk): Likewise.
21232 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
21233 Jean Lee <xiaoyur347@gmail.com>
21235 * config/mips/mips.c (mips_asan_shadow_offset): New function.
21236 (TARGET_ASAN_SHADOW_OFFSET): Define.
21237 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
21238 true for -fsanitize=address.
21240 2018-04-25 Mark Wielaard <mark@klomp.org>
21242 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
21245 2018-04-25 Jakub Jelinek <jakub@redhat.com>
21247 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
21248 than "alu", remove explicit "memory" and "imm_disp" attributes.
21249 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
21251 PR middle-end/85414
21252 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
21253 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
21254 gen_lowpart_no_emit.
21256 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
21259 * config/i386/i386.c (ix86_expand_builtin): Change memory
21260 operand to XI, extend p0 to Pmode.
21261 * config/i386/i386.md: Change unspec volatile and operand
21262 1 mode to XI, change operand 0 mode to P.
21264 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
21266 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
21267 GET_MODE_MASK before any checking.
21268 (nds32_can_use_bset_p): Likewise.
21269 (nds32_can_use_btgl_p): Likewise.
21271 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
21273 * config/nds32/nds32-doubleword.md: New define_split pattern for
21274 illegal register number.
21276 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
21278 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
21280 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
21282 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
21284 2018-04-25 Richard Biener <rguenther@suse.de>
21286 * lto-streamer.h (LTO_major_version): Bump to 8.
21288 2018-04-25 Jakub Jelinek <jakub@redhat.com>
21290 * BASE-VER: Set to 9.0.0.
21292 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
21294 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
21295 in __abskf2 and __powikf2.
21297 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
21300 * config/aarch64/constraints.md (Usg, Usj): New constraints.
21301 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
21302 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
21303 Use the above on operand 2. Reindent.
21304 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
21306 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
21309 * common/config/i386/i386-common.c (ix86_handle_option): Don't
21311 * config/i386/i386.opt (mcet): Removed.
21312 * doc/install.texi: Remove -mcet documentation.
21313 * doc/invoke.texi: Likewise.
21315 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
21318 * doc/install.texi: Remove -mcet from bootstrap-cet.
21320 2018-04-24 Jakub Jelinek <jakub@redhat.com>
21323 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
21324 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
21328 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
21329 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
21330 containing a CONST_VECTOR.
21332 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
21334 * doc/install.texi: Update newlib dependency for nvptx.
21336 2018-04-24 Jakub Jelinek <jakub@redhat.com>
21339 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
21340 instead of INTVAL when shifting x left.
21342 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
21344 PR tree-optimization/85478
21345 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
21346 vect_grouped_store_supported for single element vectors.
21348 2018-04-24 Richard Biener <rguenther@suse.de>
21351 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
21352 load cost increase to the case of non-constant step.
21354 2018-04-24 Jakub Jelinek <jakub@redhat.com>
21357 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
21358 destination if any_malformed_asm.
21360 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
21362 PR middle-end/85496
21363 * expr.c (store_field): In the bitfield case, if the value comes from
21364 a function call and is returned in registers by means of a PARALLEL,
21365 do not change the mode of the temporary unless BLKmode and VOIDmode.
21367 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
21369 PR rtl-optimization/85423
21370 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
21371 dependencies to debug insns when the previous insn is non-debug.
21373 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
21375 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
21376 enums into a single definition.
21377 (fls): Fix predicates and printing.
21380 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
21382 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
21383 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
21384 and short u6 immediate.
21385 (check_if_valid_sleep_operand): Remove.
21386 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
21388 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
21390 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
21391 flag_always_save_lp condition.
21392 * config/nds32/nds32.opt (malways-save-lp): New option.
21394 2018-04-22 Shiva Chen <shiva0217@gmail.com>
21396 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
21397 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
21398 * config/nds32/nds32.h
21399 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
21400 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
21402 2018-04-22 Shiva Chen <shiva0217@gmail.com>
21404 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
21405 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
21407 2018-04-22 Shiva Chen <shiva0217@gmail.com>
21408 Chung-Ju Wu <jasonwucj@gmail.com>
21410 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
21412 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
21413 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
21415 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
21417 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
21419 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
21421 * config/nds32/nds32-protos.h (nds32_data_alignment,
21422 nds32_local_alignment): Declare.
21423 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
21424 nds32_local_alignment): New functions.
21425 (TARGET_CONSTANT_ALIGNMENT): Define.
21426 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
21428 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
21430 * config/nds32/nds32.c
21431 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
21432 (TARGET_MODES_TIEABLE_P): Likewise.
21434 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
21436 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
21437 level Ofast and Og.
21439 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
21440 Chung-Ju Wu <jasonwucj@gmail.com>
21442 * config/nds32/constants.md (unspec_volatile_element): Add enum values
21443 for unaligned access.
21444 * config/nds32/nds32-intrinsic.c: Implementation of expanding
21446 * config/nds32/nds32-intrinsic.md: Likewise.
21447 * config/nds32/nds32_intrinsic.h: Likewise.
21448 * config/nds32/nds32.h (nds32_builtins): Likewise.
21449 * config/nds32/nds32.opt (munaligned-access): New option.
21450 * config/nds32/nds32.c (nds32_asm_file_start): Display
21451 flag_unaligned_access status.
21453 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
21455 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
21456 -mno-relax is present.
21457 * config/riscv/linux.h (LINK_SPEC): Ditto.
21459 2018-04-20 Martin Sebor <msebor@redhat.com>
21462 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
21464 (gimple_fold_builtin_stxcpy_chk): Same.
21465 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
21467 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
21470 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
21471 __powikf2 when long double is IEEE 128-bit.
21473 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
21475 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
21476 step to make sure stack always aligned.
21478 2018-04-20 Carl Love <cel@us.ibm.com>
21481 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
21482 size check for arg0.
21484 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
21485 Tom de Vries <tom@codesourcery.com>
21488 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
21489 Emit insns for calls too.
21490 (nvptx_find_par): Always look for worker-level predecessor insn.
21491 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
21493 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
21494 (nvptx_process_pars): Propagate frames for calls.
21496 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
21499 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
21501 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
21502 (ix86_handle_option): Don't handle OPT_mibt.
21503 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
21505 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
21507 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
21508 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
21509 (ix86_target_macros): Define __CET__ with flag_cf_protection
21510 for -fcf-protection.
21511 * config/i386/i386.c (isa2_opts): Remove -mibt.
21512 * config/i386/i386.h (TARGET_IBT): Removed.
21513 (TARGET_IBT_P): Likewise.
21514 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
21515 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
21516 * config/i386/i386.opt (mcet): Update help message.
21517 (mshstk): Likewise.
21519 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
21520 -mcet as an alias for -mshstk.
21522 2018-04-20 Richard Biener <rguenther@suse.de>
21524 PR middle-end/85475
21525 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
21526 complexity by forcing a single use of the multiply operand.
21528 2018-04-20 Martin Jambor <mjambor@suse.cz>
21531 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
21532 recursion dependency to only apply to non-clones.
21534 2018-04-20 Martin Jambor <mjambor@suse.cz>
21537 * ipa-cp.c (create_specialized_node): Check that clones of
21538 self-recursive edges exist during IPA-CP.
21540 2018-04-19 Toon Moene <toon@moene.org>
21542 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
21545 2018-04-19 Jakub Jelinek <jakub@redhat.com>
21547 PR tree-optimization/85467
21548 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
21549 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
21550 VECTOR_CST element to type.
21552 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
21555 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
21556 * config/i386/i386.md (builtin_setjmp_setup): Removed.
21557 (builtin_longjmp): Likewise.
21558 (save_stack_nonlocal): New pattern.
21559 (restore_stack_nonlocal): Likewise.
21561 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
21564 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21565 Replace ASM_OUTPUT_LABEL with fprintf.
21567 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
21570 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21571 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
21572 * config/i386/i386-c.c (ix86_target_macros_internal): Also
21573 define __IBT__ and __SHSTK__ for -fcf-protection.
21574 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
21576 (ix86_trampoline_init): Likewise.
21577 (x86_output_mi_thunk): Likewise.
21578 (ix86_notrack_prefixed_insn_p): Likewise.
21579 (ix86_option_override_internal): Don't disallow -fcf-protection.
21580 * config/i386/i386.md (rdssp<mode>): Also enable for
21582 (incssp<mode>): Likewise.
21583 (nop_endbr): Likewise.
21584 * config/i386/i386.opt (mcet): Change help message to built-in
21587 (mshstk): Likewise.
21588 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
21589 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
21590 enable CET built-in functions.
21592 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
21594 * common/config/i386/i386-common.c
21595 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
21596 OPTION_MASK_ISA_MOVDIRI_UNSET,
21597 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
21598 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
21599 * config.gcc (movdirintrin.h): New header.
21600 * config/i386/cpuid.h (bit_MOVDIRI,
21601 bit_MOVDIR64B): New bits.
21602 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
21604 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
21605 (VOID, PVOID, PCVOID)): New function types.
21606 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
21607 __builtin_ia32_directstoreu_u64,
21608 __builtin_ia32_movdir64b): New builtins.
21609 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
21610 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
21612 (ix86_valid_target_attribute_inner_p): Ditto.
21613 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
21614 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
21615 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
21616 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
21617 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
21618 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
21619 (movdiri<mode>, movdir64b_<mode>): New.
21620 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
21621 * config/i386/immintrin.h: Include movdirintrin.h.
21622 * config/i386/movdirintrin.h: New file.
21623 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
21625 2018-04-19 Richard Biener <rguenther@suse.de>
21627 PR middle-end/85455
21628 * cfg.c (clear_bb_flags): When loop state says we have
21629 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
21631 2018-04-19 Richard Biener <rguenther@suse.de>
21633 PR tree-optimization/84737
21634 * tree-vect-data-refs.c (vect_copy_ref_info): New function
21635 copying restrict info.
21636 (vect_setup_realignment): Use it.
21637 * tree-vectorizer.h (vect_copy_ref_info): Declare.
21638 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
21639 the first DR to all generated stores.
21640 (vectorizable_load): Likewise for loads.
21642 2018-04-19 Jakub Jelinek <jakub@redhat.com>
21644 PR tree-optimization/85446
21645 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
21646 the integral and pointer types to have the same precision.
21648 * doc/install.texi: Document --disable-cet being the default and
21651 2018-04-18 Martin Liska <mliska@suse.cz>
21653 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
21656 2018-04-18 Martin Liska <mliska@suse.cz>
21659 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
21662 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21663 arguments if they are comparable.
21665 2018-04-18 Martin Liska <mliska@suse.cz>
21668 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
21671 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21674 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
21677 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
21678 ENDBR after calling __morestack.
21680 2018-04-18 David Malcolm <dmalcolm@redhat.com>
21683 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
21684 by using gcc_base_ver to generate a gcc_driver_version, and use
21685 it when generating GCC_DRIVER_NAME.
21686 * configure: Regenerate.
21688 2018-04-18 Jakub Jelinek <jakub@redhat.com>
21691 * config.gcc: Obsolete powerpc*-*-*spe*.
21693 2018-04-17 Jakub Jelinek <jakub@redhat.com>
21696 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
21697 (stabstr_D): Change type of unum from unsigned int to
21698 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
21701 2018-04-17 Jim Wilson <jimw@sifive.com>
21704 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
21705 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
21706 Set arg_pointer_offset after using pretend_args_size.
21708 2018-04-17 Jakub Jelinek <jakub@redhat.com>
21710 PR rtl-optimization/85431
21711 * dse.c (record_store): Ignore zero width stores.
21714 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
21715 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
21716 __builtin_stack_restore rather than after it.
21717 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
21718 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
21719 argument instead of virtual_dynamic_stack_rtx.
21721 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
21723 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
21725 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21726 Add note to error message to explain internal mapping of overloaded
21727 built-in function name to non-overloaded built-in function name.
21728 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
21731 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
21734 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
21735 where the inputs overlap with the output.
21737 2018-04-17 Jakub Jelinek <jakub@redhat.com>
21739 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
21740 (=v, v) alternative and explicit "memory" attribute.
21741 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
21742 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21744 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
21745 "sselog1" type instead of "sselog".
21746 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
21747 "sselog". Remove explicit "memory" attribute.
21748 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
21749 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21751 (vec_extract_hi_v32hi): Merge all alternatives into one, use
21752 "sselog1" type instead of "sselog". Remove explicit "memory"
21754 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
21755 use "sselog1" type instead of "sselog". Remove explicit "memory"
21757 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
21758 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21760 (vec_extract_hi_v64qi): Merge all alternatives into one, use
21761 "sselog1" type instead of "sselog". Remove explicit "memory"
21763 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
21764 use "sselog1" type instead of "sselog". Remove explicit "memory"
21768 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
21770 PR middle-end/85414
21771 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
21774 2018-04-17 Martin Jambor <mjambor@suse.cz>
21777 * ipa-cp.c (create_specialized_node): Call
21778 expand_all_artificial_thunks if necessary.
21780 2018-04-17 Martin Liska <mliska@suse.cz>
21783 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
21784 in message, remote space in between '_G' and '('.
21786 2018-04-17 Jakub Jelinek <jakub@redhat.com>
21789 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
21790 avx512f_vmcmp<mode>3<round_saeonly_name>,
21791 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
21792 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
21793 avx512f_rndscale<mode><round_saeonly_name>,
21794 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
21795 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
21796 Use %<iptr>2 instead of %2 for -masm=intel.
21797 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
21798 avx512f_vcvttss2usi<round_saeonly_name>,
21799 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
21801 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
21802 avx512f_vcvttsd2usi<round_saeonly_name>,
21803 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
21804 Use %q1 instead of %1 for -masm=intel.
21805 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
21806 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
21807 of %3 for -masm=intel.
21808 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
21810 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
21812 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
21814 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
21815 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
21817 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
21819 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
21820 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
21821 %g1 and one with %0 and %1.
21822 (avx512er_vmrcp28<mode><round_saeonly_name>,
21823 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
21824 %1 for -masm=intel.
21825 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
21826 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
21827 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
21828 of %0 and %{%4%} for -masm=intel.
21829 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
21830 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
21831 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
21832 order of %0 and %{%5%}%{z%} for -masm=intel.
21834 2018-04-17 Jan Hubicka <jh@suse.cz>
21837 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
21839 2018-04-17 Martin Liska <mliska@suse.cz>
21842 * multiple_target.c (create_dispatcher_calls): Set apostrophes
21843 for target_clone error message. Make default implementation
21844 clone to be a local declaration.
21845 (separate_attrs): Add new argument and check for an empty
21847 (expand_target_clones): Handle it.
21848 (ipa_target_clone): Make redirection just for target_clones
21851 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
21852 Tom de Vries <tom@codesourcery.com>
21854 PR middle-end/84955
21855 * omp-expand.c (expand_oacc_for): Add dummy false branch for
21856 tiled basic blocks without omp continue statements.
21858 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
21861 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
21862 vec_extract expression as having side effects to make sure it gets
21865 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
21868 * config/i386/i386.c (get_builtin_code_for_version): Check
21871 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
21874 * config.gcc: Support "skylake".
21875 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21877 * config/i386/i386.c (m_SKYLAKE): Define.
21878 (processor_target_table): Add "skylake".
21879 (ix86_option_override_internal): Add "skylake".
21880 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
21881 PROCESSOR_CANNONLAKE.
21882 (get_builtin_code_for_version): Fix priority for
21883 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
21884 PROCESSOR_SKYLAKE-AVX512.
21885 * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
21886 (processor_type): Add PROCESSOR_SKYLAKE.
21888 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
21889 Jason Merrill <jason@redhat.com>
21892 * convert.c (convert_to_integer_1): Use direct recursion for
21893 enumeral types and types with a precision less than the number
21894 of bits in their mode.
21896 2018-04-16 Julia Koval <julia.koval@intel.com>
21899 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
21900 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
21902 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
21905 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
21906 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
21907 and -mno-direct-move.
21909 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
21912 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
21913 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
21914 Ensure negative shifts result in {0}.
21916 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
21918 PR rtl-optimization/79916
21919 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
21920 regs (if any) to define how to gnerate SD moves when LRA is in
21923 2018-04-13 Jakub Jelinek <jakub@redhat.com>
21925 PR rtl-optimization/85393
21926 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
21927 * except.c (expand_dw2_landing_pad_for_region): Make static.
21928 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
21929 a label and unconditional jump to old_bb, rather than
21930 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
21933 PR rtl-optimization/85376
21934 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
21935 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
21936 instead of a specific value.
21938 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
21939 Bin Cheng <bin.cheng@arm.com>
21941 PR tree-optimization/82965
21942 PR tree-optimization/83991
21943 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
21944 by_profile_only parameter.
21945 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
21946 information if the loop was predicted to iterate too many times.
21947 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
21949 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
21952 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
21955 2018-04-13 Martin Liska <mliska@suse.cz>
21956 Jakub Jelinek <jakub@redhat.com>
21958 PR middle-end/81657
21959 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
21960 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
21961 * builtins.c (expand_builtin_memory_copy_args): Use
21962 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
21963 handle dest_addr == pc_rtx.
21965 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
21968 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
21969 asked to not generate direct moves.
21970 (fix_trunc<mode>si2_stfiwx): Similar.
21971 (fix_trunc<mode>si2_internal): Similar.
21973 2018-04-12 Jakub Jelinek <jakub@redhat.com>
21976 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
21977 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
21978 lookup if dest in some wider mode is known to be const0_rtx and
21979 if so, record permanent equivalence for it to be ZERO_EXTEND of
21980 the narrower mode destination.
21982 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
21984 * lto-streamer-out.c (output_function): Revert 259346.
21985 * omp-expand.c (expand_oacc_for): Likewise.
21987 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
21989 PR rtl-optimization/85354
21990 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
21991 * sel-sched.c (sel_global_init): ... here.
21993 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
21996 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
21997 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
21998 mode for PE-COFF targets.
21999 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
22000 (i386_pe_asm_lto_end): Likewise.
22001 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
22002 (TARGET_ASM_LTO_END): Likewise.
22003 * config/i386/winnt.c (saved_debug_info_level): New static variable.
22004 (i386_pe_asm_lto_start): New function.
22005 (i386_pe_asm_lto_end): Likewise.
22007 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
22008 Richard Biener <rguenther@suse.de>
22010 PR middle-end/84955
22011 * lto-streamer-out.c (output_function): Fix CFG loop state before
22013 * omp-expand.c (expand_oacc_for): Handle calls to internal
22014 functions like regular functions.
22016 2018-04-12 Richard Biener <rguenther@suse.de>
22019 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
22020 for the early LTO debug to properly generate references to it
22021 during DIE emission. Do not re-use that for the skeleton for
22023 (dwarf2out_early_finish): Likewise.
22025 2018-04-12 Jakub Jelinek <jakub@redhat.com>
22028 * config/i386/sse.md
22029 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
22030 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
22031 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
22032 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
22033 and output is a reg, avoid creating invalid lowpart subreg, but
22034 instead split into a 512-bit move. Don't split if not AVX512VL,
22035 input is xmm16+ reg and output is a mem.
22036 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
22037 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
22038 xmm16+ reg and output is a mem.
22040 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
22042 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
22043 also for flag_dwarf2_cfi_asm.
22045 2018-04-12 Jakub Jelinek <jakub@redhat.com>
22047 PR rtl-optimization/85342
22048 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
22049 a bool scalar var inside of the loop instead. Don't try to update
22050 recog_data.operand after failed apply_change_group.
22052 2018-04-12 Tom de Vries <tom@codesourcery.com>
22055 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
22056 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
22057 array with flexible array member as array without given dimension.
22058 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
22059 argument for undefined param to true.
22061 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
22064 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
22065 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
22066 from PowerPC section.
22067 * config/rs6000/sysv4.opt (mcall-): Improve help text.
22068 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
22069 help text that is too long.
22070 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
22071 help text that is too long.
22072 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
22073 help text that is too long.
22075 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
22077 * config/alpha/alpha.md (stack_probe_internal): Rename
22078 from "probe_stack". Update all callers.
22080 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
22082 PR rtl-optimization/84566
22083 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
22084 sched_macro_fuse_insns.
22086 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
22089 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
22090 (compute_block_dependences): ... from here.
22092 2018-04-11 Jakub Jelinek <jakub@redhat.com>
22094 PR tree-optimization/85331
22095 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
22096 from int to HOST_WIDE_INT.
22098 2018-04-11 Martin Jambor <mjambor@suse.cz>
22101 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
22102 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
22103 not the same as the source val.
22104 (cgraph_edge_brings_value_p): New parameter.
22105 (gather_edges_for_value): Pass destination value to
22106 cgraph_edge_brings_value_p.
22107 (perhaps_add_new_callers): Likewise.
22108 (get_info_about_necessary_edges): Likewise and exclude values brought
22109 only by self-recursive edges.
22110 (create_specialized_node): Redirect only clones of self-calling edges.
22111 (+self_recursive_pass_through_p): New function.
22112 (find_more_scalar_values_for_callers_subset): Use it.
22113 (find_aggregate_values_for_callers_subset): Likewise.
22114 (known_aggs_to_agg_replacement_list): Removed.
22115 (decide_whether_version_node): Re-calculate known constants for all
22116 remaining context clones.
22118 2018-04-11 Richard Biener <rguenther@suse.de>
22121 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
22122 from early DWARF output.
22123 (dwarf2out_early_finish): Output line info unconditionally into
22124 early DWARF and add reference to it.
22126 2018-04-11 Jakub Jelinek <jakub@redhat.com>
22129 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
22130 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
22131 other than V2DFmode using iptr mode attribute.
22132 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
22134 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
22136 PR rtl-optimization/84659
22137 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
22139 2018-04-11 Jakub Jelinek <jakub@redhat.com>
22142 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
22144 (output_loc_list): Pass address of a dummy size variable even in the
22145 locview handling loop.
22146 (index_location_lists): Add comment on why skip_loc_list_entry can't
22149 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
22152 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
22155 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
22158 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
22159 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
22160 and -mstring-compare-inline-limit.
22162 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
22165 * config/rs6000/rs6000.md (allocate_stack): Put the residual size
22166 for stack clash protection in a register whenever we need it to be in
22169 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
22171 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
22172 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
22174 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
22177 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
22179 (mlong-double-): Ditto.
22180 * config/rs6000/sysv4.opt (msdata=): Ditto.
22181 (mtls-size=): Ditto.
22183 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
22185 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
22186 erroneous entries for
22187 "vector int vec_ldl (int, long int *)", and
22188 "vector unsigned int vec_ldl (int, unsigned long int *)".
22189 Add comments and entries for
22190 "vector bool char vec_ldl (int, bool char *)",
22191 "vector bool short vec_ldl (int, bool short *)",
22192 "vector bool int vec_ldl (int, bool int *)",
22193 "vector bool long long vec_ldl (int, bool long long *)",
22194 "vector pixel vec_ldl (int, pixel *)",
22195 "vector long long vec_ldl (int, long long *)",
22196 "vector unsigned long long vec_ldl (int, unsigned long long *)".
22197 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
22198 type tree bool_long_long_type_node and correct definition of
22199 bool_V2DI_type_node to make reference to this new type tree.
22200 (rs6000_mangle_type): Replace erroneous reference to
22201 bool_long_type_node with bool_long_long_type_node.
22202 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
22203 comments to emphasize sign distinctions for char and int types and
22204 replace RS6000_BTI_bool_long constant with
22205 RS6000_BTI_bool_long_long constant. Also add comment to restrict
22206 use of RS6000_BTI_pixel.
22207 (bool_long_type_node): Remove this macro definition.
22208 (bool_long_long_type_node): New macro definition
22210 2018-04-10 Jakub Jelinek <jakub@redhat.com>
22212 PR rtl-optimization/85300
22213 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
22214 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
22215 simplify_unary_operation fails.
22217 2018-04-10 Martin Liska <mliska@suse.cz>
22219 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
22220 cgraph_edge and ipa_ref.
22222 2018-04-10 Jakub Jelinek <jakub@redhat.com>
22226 * config/i386/sse.md
22227 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
22228 computation of the VEC_MERGE selector from mask.
22229 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
22230 Fix decoding of the VEC_MERGE selector into mask.
22232 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
22234 PR tree-optimization/85286
22235 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
22237 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
22239 * final.c (final_1): Set insn_last_address as well as
22240 insn_current_address.
22242 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
22245 * explow.c (emit_stack_probe): Call validize_mem on memory location
22246 before passing it to gen_probe_stack. Create address operand and
22247 legitimize it for the probe_stack_address case.
22249 2018-04-09 Jan Hubicka <jh@suse.cz>
22252 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
22253 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
22254 * tree.c (free_lang_data_in_type): Fix handling of binfos;
22256 (free_lang_data): Rebuild type inheritance graph.
22258 2018-04-09 Martin Sebor <msebor@redhat.com>
22260 * invoke.texi (-finline-small-functions): Mention other optimization
22262 (-findirect-inlining, -fpartial-inlining): Same.
22263 (-finline-functions-called-once): Same.
22264 (-freorder-blocks-and-partition): Same.
22266 2018-04-09 Jan Hubicka <jh@suse.cz>
22269 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
22270 jumps; choose last target that matches the criteria (i.e.
22271 no partition changes for non-crossing jumps).
22272 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
22273 support for redirecting crossing jumps to non-crossing.
22275 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
22277 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
22278 also for naked functions.
22280 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
22282 * config/arc/arc.md (add_shift): New pattern.
22283 (add_shift2): Likewise.
22284 (sub_shift): Likewise.
22285 (sub_shift_cmp0_noout): Likewise.
22286 (compare_si_ashiftsi): Likewise.
22287 (xbfu_cmp0_noout): New combine pattern.
22288 (xbfu_cmp0"): Likewise.
22289 (movsi_set_cc_insn): Place the predicable variant first.
22290 (commutative_binary_cmp0_noout): Remove clobber.
22291 (commutative_binary_cmp0): New pattern.
22292 (noncommutative_binary_cmp0): Likewise.
22293 (noncommutative_binary_cmp0_noout): Likewise.
22294 (noncommutative_binary_comparison_result_used): Removed.
22295 (rsub_cmp0): New pattern.
22296 (rsub_cmp0_noout): Likewise.
22297 (extzvsi): Changed, keep only meaningful variants.
22298 (SQH, SEZ): New iterators.
22299 (SQH_postfix): New mode attribute.
22300 (SEZ_prefix): New code attribute.
22301 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
22302 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
22303 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
22304 of numerical value.
22305 (noncommutative_operator): Check the availability of barrel
22308 2018-04-09 Richard Biener <rguenther@suse.de>
22310 PR tree-optimization/85284
22311 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
22312 Only use the niter constraining form of simple_iv when the exit
22313 is always executed.
22315 2018-04-09 Tom de Vries <tom@codesourcery.com>
22318 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
22319 (define_expand "*memory_barrier"): New define_expand.
22320 (define_insn "memory_barrier"): New insn.
22322 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
22324 PR rtl-optimization/80463
22325 PR rtl-optimization/83972
22326 PR rtl-optimization/83480
22328 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
22329 correct producer for the insn.
22330 (tidy_control_flow): Fixup seqnos in case of debug insns.
22332 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
22334 PR rtl-optimization/83913
22336 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
22337 different sched-times when merging exprs.
22339 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
22341 PR rtl-optimization/83962
22343 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
22344 tidy_fallthru_edge and tidy_control_flow.
22346 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
22348 PR rtl-optimization/83530
22350 * sel-sched.c (force_next_insn): New global variable.
22351 (remove_insn_for_debug): When force_next_insn is true, also leave only
22352 next insn in the ready list.
22353 (sel_sched_region): When the region wasn't scheduled, make another pass
22354 over it with force_next_insn set to 1.
22356 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
22358 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
22360 * config/nds32/constants.md (unspec_volatile_element): Add enum values
22361 for interrupt control.
22362 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
22363 functions for interrupt control.
22364 * config/nds32/nds32-intrinsic.md: Likewise.
22365 * config/nds32/nds32_intrinsic.h: Likewise.
22366 * config/nds32/nds32.h (nds32_builtins): Likewise.
22368 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
22370 * config/nds32/nds32.c (nds32_init_machine_status,
22371 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
22372 strict_aligned_p field.
22373 (nds32_expand_to_rtl_hook): New function.
22374 (TARGET_EXPAND_TO_RTL_HOOK): Define.
22375 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
22377 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
22378 Chung-Ju Wu <jasonwucj@gmail.com>
22380 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
22381 * config/nds32/nds32-n7.md: New file.
22382 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
22383 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
22385 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
22386 * config/nds32/nds32.md (pipeline_model): Add n7.
22387 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
22388 * config/nds32/pipelines.md: Include n7 settings.
22390 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
22391 Chung-Ju Wu <jasonwucj@gmail.com>
22393 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
22394 * config/nds32/nds32-e8.md: New file.
22395 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
22396 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
22398 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
22399 * config/nds32/nds32.md (pipeline_model): Add e8.
22400 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
22401 * config/nds32/pipelines.md: Include e8 settings.
22403 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
22404 Chung-Ju Wu <jasonwucj@gmail.com>
22406 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
22407 * config/nds32/nds32-n8.md: New file.
22408 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
22409 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
22411 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
22412 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
22413 * config/nds32/nds32.md (pipeline_model): Add n8.
22414 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
22415 * config/nds32/pipelines.md: Include n8 settings.
22417 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
22418 Chung-Ju Wu <jasonwucj@gmail.com>
22420 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
22421 * config/nds32/nds32-n9-2r1w.md: New file.
22422 * config/nds32/nds32-n9-3r2w.md: New file.
22423 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
22424 nds32_register_ports): New or modify for cpu n9.
22425 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
22427 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
22428 * config/nds32/nds32-utils.c: New file.
22429 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
22430 TARGET_MUL_SLOW): Define.
22431 * config/nds32/nds32.md (pipeline_model): New attribute.
22432 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
22433 New options that support cpu n9.
22434 * config/nds32/pipelines.md: Include n9 settings.
22435 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
22437 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
22439 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
22440 information if necessary.
22441 (output_cond_branch_compare_zero): Likewise.
22442 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
22443 (nds32_target_alignment): Refine for alignment.
22444 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
22445 (FUNCTION_BOUNDARY): Modify.
22446 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
22448 * config/nds32/nds32.opt (malways-align, malign-functions): New.
22450 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
22452 * config/nds32/constants.md (unspec_volatile_element): Add values for
22453 TLB operation and data prefetch.
22454 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
22455 functions for TLB operation and data prefetch.
22456 * config/nds32/nds32-intrinsic.md: Likewise.
22457 * config/nds32/nds32_intrinsic.h: Likewise.
22458 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
22459 (nds32_print_operand): Likewise.
22460 * config/nds32/nds32.h (nds32_builtins): Likewise.
22462 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
22463 Andrew Pinski <pinsika@gcc.gnu.org>
22465 PR middle-end/82976
22466 * match.pd: Use constant_boolean_node of correct type instead of
22467 boolean_true_node or boolean_false_node for simplifying
22468 pointer comparisons to zero.
22470 2018-04-07 Jakub Jelinek <jakub@redhat.com>
22472 PR tree-optimization/80021
22473 * tree.c (verify_type_variant): Make error call in verify_variant_match
22474 translatable and remove final full stop.
22476 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
22478 * config/nds32/constants.md (unspec_volatile_element): Add
22479 UNSPEC_VOLATILE_EH_RETURN.
22480 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
22481 nds32_output_stack_pop): Support dwarf exception handling process.
22482 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
22483 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
22484 exception handling process.
22485 (nds32_compute_stack_frame): Likewise.
22486 (nds32_return_addr_rtx): Likewise.
22487 (nds32_initial_elimination_offset): Likewise.
22488 (nds32_expand_prologue): Likewise.
22489 (nds32_expand_epilogue): Likewise.
22490 (nds32_dynamic_chain_address): New function.
22491 * config/nds32/nds32.h (machine_function): Add fields for dwarf
22492 exception handling.
22493 (DYNAMIC_CHAIN_ADDRESS): Define.
22494 (EH_RETURN_DATA_REGNO): Define.
22495 (EH_RETURN_STACKADJ_RTX): Define.
22496 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
22497 patterns for dwarf exception handling.
22499 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
22501 * config/nds32/nds32.h: Clean up obsolete macros.
22503 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
22505 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22506 Add enum values for particular instructions.
22507 * config/nds32/nds32-intrinsic.c: Implementation of expanding
22508 particular intrinsic functions.
22509 * config/nds32/nds32-intrinsic.md: Likewise.
22510 * config/nds32/nds32_intrinsic.h: Likewise.
22511 * config/nds32/nds32.h (nds32_builtins): Likewise.
22512 * config/nds32/nds32.md (type): Add pbsad and pbsada.
22513 (btst, ave): New patterns for particular instructions.
22515 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
22517 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22518 Add enum values for atomic load/store and memory sync.
22519 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
22521 * config/nds32/nds32-intrinsic.md: Likewise.
22522 * config/nds32/nds32_intrinsic.h: Likewise.
22523 * config/nds32/nds32.h (nds32_builtins): Likewise.
22525 2018-04-07 Jakub Jelinek <jakub@redhat.com>
22527 PR tree-optimization/85257
22528 * fold-const.c (native_encode_vector): If not all elts could fit
22529 and off is -1, return 0 rather than offset.
22530 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
22531 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
22532 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
22533 adjust buffer in native_interpret_expr call.
22535 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
22537 * config/nds32/constants.md (unspec_volatile_element): Add cache
22538 control enum values.
22539 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
22540 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
22541 * config/nds32/nds32.c (nds32_cctl_names): New.
22542 (nds32_print_operand): Handle cache control register names.
22543 * config/nds32/nds32.h (nds32_builtins): New enum values.
22544 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
22546 * config/nds32/nds32.md (type): Add mmu.
22547 * config/nds32/pipelines.md (simple_insn): Add mmu.
22549 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
22551 * config/nds32/nds32.md (type): Remove call.
22552 * config/nds32/pipelines.md (simple_insn): Likewise.
22554 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
22556 * config/nds32/constants.md (unspec_volatile_element): Add
22557 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
22558 UNSPEC_VOLATILE_FMFCFG.
22559 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
22560 description for fmfcfg and fmfcsr.
22561 (bdesc_1arg): Add fmtcsr.
22562 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
22563 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
22564 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
22565 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
22566 unspec_fmfcfg): New patterns.
22567 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
22568 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
22569 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
22570 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
22571 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
22572 __nds32__fmfcfg): Define.
22574 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
22576 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
22577 intrinsic register names.
22578 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
22579 intrinsic register enum values and macros.
22581 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
22583 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
22584 for load/store addressing form.
22585 (nds32_print_operand_address): Likewise.
22587 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
22590 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
22591 based on LABEL_REF. Remove useless assertion.
22592 (pic_address_needs_scratch): Fix formatting.
22593 (sparc_legitimize_pic_address): Minor tweaks.
22594 (sparc_delegitimize_address): Adjust assertion accordingly.
22595 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
22596 into symbolic_operand.
22597 (movsi_high_pic_label_ref): Likewise.
22598 (movsi_lo_sum_pic_label_ref): Likewise.
22599 (movdi_pic_label_ref): Likewise.
22600 (movdi_high_pic_label_ref): Likewise.
22601 (movdi_lo_sum_pic_label_ref): Likewise.
22603 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
22605 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
22606 custom LIB_SPEC setup.
22608 2018-04-06 Ruslan Bukin <br@bsdpad.com>
22609 Kito Cheng <kito.cheng@gmail.com>
22611 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
22612 * config/riscv/freebsd.h: New.
22614 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
22616 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
22617 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
22620 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
22621 Kito Cheng <kito.cheng@gmail.com>
22623 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
22624 nds32_output_call, nds32_symbol_binds_local_p): New functions.
22625 * config/nds32/nds32-protos.h (nds32_output_call,
22626 nds32_output_return): Declare.
22627 * config/nds32/nds32.md: Refine all the call and return patterns.
22629 2018-04-06 Jakub Jelinek <jakub@redhat.com>
22632 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
22633 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
22635 PR rtl-optimization/84872
22636 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
22637 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
22638 EDGE_CROSSING edge.
22640 2018-04-06 Tamar Christina <tamar.christina@arm.com>
22642 * expr.c (copy_blkmode_to_reg): Revert 254862.
22643 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
22645 2018-04-06 Richard Biener <rguenther@suse.de>
22647 PR middle-end/85244
22648 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
22649 after seeing a component reference with an adjacent field. Treat
22650 refs to arrays at struct end of external decls similar to
22651 refs to unconstrained commons.
22653 2018-04-06 Jakub Jelinek <jakub@redhat.com>
22656 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
22657 look through SAVE_EXPRs with non-side-effects argument. Adjust
22659 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
22662 2018-04-06 Richard Biener <rguenther@suse.de>
22664 PR middle-end/85180
22665 * alias.c (find_base_term): New wrapper around find_base_term
22666 unwinding CSELIB_VAL_PTR changes.
22667 (find_base_term): Do not restore CSELIB_VAL_PTR during the
22670 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
22672 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
22674 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
22675 constant definitions.
22676 ("nop"): lr 0,0 -> nopr r0
22677 ("nop_lr0", "nop_lr1"): New insn definitions.
22679 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
22681 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
22682 NDS32_V3PUSH_AVAILABLE_P macro.
22684 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
22685 Chung-Ju Wu <jasonwucj@gmail.com>
22687 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
22688 (nds32*-*-*): Add float and fpu_config into supported_defaults.
22689 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
22690 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
22691 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
22692 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
22693 * config/nds32/constraints.md: New constraints and checking for hard
22694 float configuration.
22695 * config/nds32/iterators.md: New mode iterator and attribute for hard
22696 float configuration.
22697 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
22699 * config/nds32/nds32-fpu.md: New file.
22700 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
22701 deal with hard float code generation.
22702 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
22704 (abi_type, float_reg_number): New enum type.
22705 * config/nds32/nds32-predicates.c: New predicates for hard float.
22706 * config/nds32/nds32-protos.h: Declare functions for hard float.
22707 * config/nds32/nds32.c: Implementation for hard float configuration.
22708 * config/nds32/nds32.h: Definitions for hard float configuration.
22709 * config/nds32/nds32.md: Include hard float machine description and
22710 modify patterns for hard float configuration.
22711 * config/nds32/nds32.opt: New options for hard float configuration.
22712 * config/nds32/predicates.md: New predicates for hard float
22715 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
22717 * common/config/nds32/nds32-common.c
22718 (nds32_option_optimization_table): Enable -mreleax-hint by default.
22720 2018-04-05 Jakub Jelinek <jakub@redhat.com>
22722 PR middle-end/85195
22723 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
22724 CONSTRUCTOR_ELT (ctor, ...)->value.
22726 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
22729 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
22731 2018-04-05 Tom de Vries <tom@codesourcery.com>
22734 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
22737 2018-04-05 Shiva Chen <shiva0217@gmail.com>
22738 Kito Cheng <kito.cheng@gmail.com>
22740 * config/nds32/constraints.md (U33): Fine-tune checking condition.
22741 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
22742 * config/nds32/nds32.h (nds32_16bit_address_type): Add
22743 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
22745 2018-04-05 Shiva Chen <shiva0217@gmail.com>
22746 Kito Cheng <kito.cheng@gmail.com>
22748 * config/nds32/constraints.md (Ufe): New memory constraint.
22749 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
22750 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
22751 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
22753 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
22754 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
22756 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
22758 * config/nds32/nds32.md: Use optimize_size in the condition for
22759 alu-shift instructions.
22761 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
22763 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
22765 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
22767 * config/nds32/nds32.md (negsi2): Refine pattern.
22769 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
22770 Chung-Ju Wu <jasonwucj@gmail.com>
22772 * config/nds32/iterators.md (shift_rotate): New code iterator.
22773 (shift): New code attribute.
22774 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
22775 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
22776 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
22777 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
22778 bit-wise operations.
22779 (andsi3, *andsi3): Ditto.
22780 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
22781 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
22782 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
22783 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
22784 nds32_ior_operand, nds32_xor_operand): New predicates.
22786 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
22788 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
22789 (addsi3, subsi3): ... this.
22791 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
22793 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
22795 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
22797 * config/nds32/nds32.md: Adjust indention.
22799 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
22801 * config/nds32/nds32.md (feature): New attribute.
22803 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
22805 * config/nds32/nds32.md (subtype): New attribute.
22807 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
22810 * config/arm/arm-builtins.c (arm_expand_builtin): Change
22811 expansion to perform a bitwise AND of the argument followed by a
22812 boolean negation of the result.
22814 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
22816 PR rtl-optimization/84878
22817 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
22818 the basic block. Assert the use reference is not artificial and that
22819 it has an associated insn.
22821 2018-04-04 Michael Matz <matz@suse.de>
22823 * builtins.c (compute_objsize): Pass correct operand
22824 to array_at_struct_end_p.
22826 2018-04-04 Richard Biener <rguenther@suse.de>
22829 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
22830 from contexts for DINFO_LEVEL_TERSE and below.
22832 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
22834 * config/nds32/nds32-doubleword.md (move_<mode>): Require
22835 resiter_operand condition.
22836 * config/nds32/nds32.md (*move<mode>): Ditto.
22838 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
22839 Monk Chiang <sh.chiang04@gmail.com>
22841 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
22843 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
22845 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
22847 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
22848 Kito Cheng <kito.cheng@gmail.com>
22850 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
22851 nds32_cond_code_str, output_cond_branch,
22852 output_cond_branch_compare_zero, nds32_expand_cbranch,
22853 nds32_expand_cstore, nds32_expand_movcc,
22854 nds32_output_cbranchsi4_equality_zero,
22855 nds32_output_cbranchsi4_equality_reg,
22856 nds32_output_cbranchsi4_equality_reg_or_const_int,
22857 nds32_output_cbranchsi4_greater_less_zero: New functions.
22858 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
22859 nds32_expand_cstore, nds32_expand_movcc,
22860 nds32_output_cbranchsi4_equality_zero,
22861 nds32_output_cbranchsi4_equality_reg,
22862 nds32_output_cbranchsi4_equality_reg_or_const_int,
22863 nds32_output_cbranchsi4_greater_less_zero): Declare.
22864 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
22865 nds32_rimm11s_operand): New predicates.
22866 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
22867 * config/nds32/nds32.md: Rewrite all the branch and conditional move
22870 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
22872 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
22873 * config/nds32/nds32.md: Ditto.
22874 * config/nds32/pipelines.md: Ditto.
22876 2018-04-04 Richard Biener <rguenther@suse.de>
22878 PR tree-optimization/85168
22879 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
22880 propagating abnormals.
22882 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
22884 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
22886 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
22887 Kito Cheng <kito.cheng@gmail.com>
22889 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
22890 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
22891 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
22892 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
22893 * config/nds32/nds32.md (sibcall_internal): New.
22894 (sibcall_register): Remove.
22895 (sibcall_immediate): Remove.
22896 (sibcall_value_internal): New.
22897 (sibcall_value_register): Remove.
22898 (sibcall_value_immediate): Remove.
22899 * config/nds32/predicates.md (nds32_general_register_operand): New.
22900 (nds32_call_address_operand): New.
22902 2018-04-03 Jakub Jelinek <jakub@redhat.com>
22904 PR rtl-optimization/85167
22905 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
22906 bb_defs if *split_p, instead preinitialize it to NULL.
22908 PR tree-optimization/85156
22909 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
22910 evaluating the argument multiple times.
22912 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
22914 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
22916 (_mm_cvtpd_ps): Likewise.
22917 (_mm_cvttpd_epi32): Likewise.
22918 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
22919 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
22920 vector, pixel, and bool following altivec.h include.
22922 2018-04-03 Martin Sebor <msebor@redhat.com>
22924 * doc/extend.texi (Common Function Attributes): Clarify.
22925 (const attribute): Likewise.
22926 (pure attribute): Likewise.
22928 2018-04-03 Jakub Jelinek <jakub@redhat.com>
22931 * config/i386/i386.c (ix86_expand_vector_set): Use
22932 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
22934 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
22936 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
22937 instructions when changing rounding bits to preserve precision bits
22938 in the x87 control word.
22940 2018-04-03 Martin Liska <mliska@suse.cz>
22942 PR tree-optimization/82491
22943 * rtl.h (strip_offset_and_add): Replace += suboffset with
22944 poly_uint64 () + suboffset.
22946 2018-03-29 Martin Liska <mliska@suse.cz>
22947 Martin Jambor <mjambor@suse.cz>
22950 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
22951 param_type is not an integral or pointer type.
22953 2018-04-03 Richard Biener <rguenther@suse.de>
22955 * sese.h (recompute_all_dominators): Remove.
22957 2018-04-02 Martin Sebor <msebor@redhat.com>
22959 * doc/invoke.texi (-Wrestrict): Fix typos.
22961 2018-04-02 Jim Wilson <jimw@sifive.com>
22963 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
22964 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
22965 (<optab>di3, <optab>si3_extend): Likewise.
22966 (<optab>si3_mask, <optab>si3_mask_1): New.
22967 (<optab>di3_mask, <optab>di3_mask_1): New.
22968 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
22969 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
22970 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
22972 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
22974 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
22977 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
22979 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
22980 (nds32_canonicalize_comparison): New function.
22982 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
22983 Kito Cheng <kito.cheng@gmail.com>
22984 Kuan-Lin Chen <kuanlinchentw@gmail.com>
22986 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
22987 * config/nds32/constants.md (unspec_volatile_element): Add
22988 UNSPEC_VOLATILE_RELAX_GROUP.
22989 * config/nds32/nds32-relax-opt.c: New file.
22990 * config/nds32/nds32-predicates.c
22991 (nds32_symbol_load_store_p): New function.
22992 * config/nds32/nds32-protos.h
22993 (nds32_symbol_load_store_p): Declare function.
22994 (make_pass_nds32_relax_opt): Declare new rtl pass function.
22995 * config/nds32/nds32.c
22996 (nds32_register_pass): New function to register pass.
22997 (nds32_register_passes): New function to register passes.
22998 * config/nds32/nds32.md (relax_group): New pattern.
22999 * config/nds32/nds32.opt (mrelax-hint): New option.
23000 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
23002 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
23004 * config/nds32/t-nds32: Modify files dependency.
23006 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
23008 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
23009 (PROFILE_HOOK): Define its implementation.
23011 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
23013 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
23014 type and 32-bit size.
23016 2018-04-01 Jakub Jelinek <jakub@redhat.com>
23018 PR middle-end/85090
23019 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
23020 (V_128_256): New mode iterator.
23021 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
23022 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
23023 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
23025 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
23026 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
23028 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
23031 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
23032 NaN inputs correctly.
23034 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
23037 * config/rs6000/vsx.md (??r): New mode attribute.
23038 (*vsx_mov<mode>_64bit): Use it.
23039 (*vsx_mov<mode>_32bit): Likewise.
23041 2018-03-30 Martin Sebor <msebor@redhat.com>
23043 PR tree-optimization/84818
23044 * builtins.c (check_access): Use warning_n.
23046 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23049 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
23051 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
23054 2018-03-30 Julia Koval <julia.koval@intel.com>
23057 * x86-tune.def (movx, partial_reg_dependency): Enable for
23060 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
23062 PR inline-asm/84985
23063 * lra-constraints.c (process_alt_operands): Move setting
23064 this_alternative_matches below.
23066 2018-03-29 Martin Liska <mliska@suse.cz>
23069 * doc/invoke.texi: Document how LTO works with debug info.
23070 Describe auto-load support of binutils. Mention 'x86-64'
23071 as valid option value of -march option.
23073 2018-03-29 Jakub Jelinek <jakub@redhat.com>
23075 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
23078 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
23079 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
23080 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
23083 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
23086 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
23087 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
23088 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
23089 for RS6000_BTM_POWERPC64.
23090 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
23091 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
23092 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
23095 (DIVDEU): Likewise.
23097 2018-03-28 Carl Love <cel@us.ibm.com>
23100 2017-09-27 Carl Love <cel@us.ibm.com>
23102 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
23103 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
23104 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
23107 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
23109 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
23110 instead of __vector bool.
23111 (_mm_max_pu8): Likewise.
23112 (_mm_min_pi16): Likewise.
23114 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
23117 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
23118 (DIVWEUO): Likewise.
23119 (DIVDEO): Likewise.
23120 (DIVDEUO): Likewise.
23121 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
23122 DIVWEUO and DIVDEUO.
23123 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
23124 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
23125 (div_extend): Likewise.
23126 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
23128 (__builtin_divweuo): Likewise.
23129 (__builtin_divdeo): Likewise.
23130 (__builtin_divdeuo): Likewise.
23132 2018-03-28 Jakub Jelinek <jakub@redhat.com>
23135 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
23136 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
23138 PR tree-optimization/82004
23139 * gimple-match-head.c (optimize_pow_to_exp): New function.
23140 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
23141 Don't fold to exp if optimize_pow_to_exp is false.
23143 2018-03-28 Martin Liska <mliska@suse.cz>
23146 * calls.c (initialize_argument_information): Fix trailing space.
23147 * common.opt: Fix typo and provide better explanation for
23148 -fsanitize-coverage option.
23149 * config/i386/i386.opt: Fix typo.
23151 2018-03-28 Jakub Jelinek <jakub@redhat.com>
23152 Martin Liska <mliska@suse.cz>
23155 * gimplify.c (asan_poison_variable): Don't do the check for
23156 gimplify_omp_ctxp here.
23157 (gimplify_decl_expr): Do it here.
23158 (gimplify_target_expr): Likewise.
23160 2018-03-28 Martin Liska <mliska@suse.cz>
23163 * config/i386/i386.c (ix86_function_arg_advance): Do not call
23164 chkp_type_bounds_count if MPX is not enabled.
23166 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
23168 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
23170 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
23173 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
23174 function to create the function decl for complex long double
23175 multiply and divide for -mabi=ieeelongdouble.
23176 (init_float128_ieee): Call it.
23178 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
23181 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
23182 -fcf-protection=branch -mibt.
23183 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
23185 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
23188 * config/arm/arm.c (arm_valid_symbolic_address): Handle
23189 arm_word_relocations.
23191 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
23194 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
23195 extern array declarations.
23197 2018-03-27 Richard Biener <rguenther@suse.de>
23199 PR middle-end/84067
23200 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
23201 explicit single_use checks.
23203 2018-03-27 Richard Biener <rguenther@suse.de>
23205 PR tree-optimization/85082
23206 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
23209 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
23211 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
23212 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
23213 Turn on fasynchronous-unwind-tables and funwind-tables.
23215 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
23218 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
23219 (*bmi_blsr_<mode>_ccz): Ditto.
23221 2018-03-26 Tom de Vries <tom@codesourcery.com>
23223 PR tree-optimization/85063
23224 * omp-general.c (offloading_function_p): New function. Factor out
23226 * omp-offload.c (pass_omp_target_link::gate): ... here.
23227 * omp-general.h (offloading_function_p): Declare.
23228 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
23229 with attribute omp declare target for offloading functions.
23231 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
23233 PR tree-optimization/84005
23234 * tree-data-ref.h (get_base_for_alignment): Declare.
23235 * tree-data-ref.c (get_base_for_alignment_1): New function.
23236 (get_base_for_alignment): Likewise.
23237 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
23238 get_base_for_alignment to find a suitable base object, instead
23239 of always using drb->base_address.
23241 2018-03-23 Jakub Jelinek <jakub@redhat.com>
23243 PR inline-asm/85022
23244 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
23245 known size by default.
23247 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
23249 PR inline-asm/85030
23250 * lra-constraints.c (process_alt_operands): Don't match BLKmode
23251 and non BLKmode operands.
23253 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23256 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
23257 Clean up attributes.
23259 2018-03-23 Richard Biener <rguenther@suse.de>
23262 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
23263 we are going to emit early debug for LTO.
23265 2018-03-23 Jakub Jelinek <jakub@redhat.com>
23267 PR inline-asm/85034
23268 * function.c (match_asm_constraints_1): Don't optimize if input
23269 doesn't satisfy general_operand predicate for output's mode.
23271 PR inline-asm/85022
23272 * alias.c (write_dependence_p): Don't require for x_canonicalized
23273 non-VOIDmode if x has VOIDmode.
23276 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
23277 just don't try to optimize it rather than assert it never happens.
23279 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
23281 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
23282 macro expansions for definition of ST_INTERNAL_<mode> and
23283 LD_INTERNAL_<mode> builtins.
23284 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
23286 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
23288 (altivec_expand_st_builtin): Likewise.
23289 (altivec_expand_builtin): Remove calls to deleted functions.
23290 (rs6000_address_for_altivec): Delete this function.
23291 * config/rs6000/vector.md: Remove expands for
23292 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
23294 2018-03-22 Sudakshina Das <sudi.das@arm.com>
23297 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
23298 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
23299 re-computing once computed.
23300 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
23301 (arm_init_machine_status): Initialize
23302 machine->static_chain_stack_bytes.
23304 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
23307 * doc/extend.texi: Add four new prototypes for vec_ld.
23308 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
23309 definitions for more logical presentation.
23310 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
23311 entries for V1TI variants of __builtin_altivec_ld builtin.
23312 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
23313 handling of V1TI variant of LVX icode pattern.
23314 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
23315 (rs6000_gimple_fold_builtin): Likewise.
23316 (altivec_init_builtins): Add code to define
23317 __builtin_altivec_lvx_v1ti function.
23319 2018-03-22 Jakub Jelinek <jakub@redhat.com>
23321 PR inline-asm/84941
23322 * function.c (match_asm_constraints_1): Don't do the optimization
23323 if input isn't a REG, SUBREG, MEM or constant.
23325 2018-03-22 Tom de Vries <tom@codesourcery.com>
23327 PR tree-optimization/84956
23328 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
23329 bb_has_abnormal_pred.
23331 2018-03-22 Jakub Jelinek <jakub@redhat.com>
23334 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
23335 DECL_INITIAL (decl) to decl at the end.
23336 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
23337 adjust the comment.
23339 2018-03-21 Joseph Myers <joseph@codesourcery.com>
23341 * doc/extend.texi (__builtin_tgmath): Document when complex
23342 integer types are treated as _Complex _Float64.
23344 2018-03-21 Tom de Vries <tom@codesourcery.com>
23346 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
23348 2018-03-21 Jakub Jelinek <jakub@redhat.com>
23350 PR tree-optimization/84960
23351 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
23352 if it is ENTRY block, move them into single succ of ENTRY in that case.
23354 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
23356 PR tree-optimization/84811
23357 * poly-int.h (poly_span_traits): Remove the T3 parameter and
23358 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
23359 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
23360 (known_subrange_p): Update accordingly. Cast each value involved
23361 in the size comparison, rather than casting the result of the
23364 2018-03-21 Jakub Jelinek <jakub@redhat.com>
23366 PR tree-optimization/84982
23367 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
23368 by flipping the least significant bit rather than all bits from
23369 bitpos to bitpos + bitsize - 1.
23371 2018-03-21 Nathan Sidwell <nathan@acm.org>
23373 * doc/extend.texi (Deprecated Features): Remove mention of
23374 long-deleted deprecations.
23376 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
23379 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
23380 * configure: Regenerate.
23382 2018-03-21 Tom de Vries <tom@codesourcery.com>
23384 PR tree-optimization/83126
23385 * tree-parloops.c (num_phis): New function.
23386 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
23388 2018-03-21 Nathan Sidwell <nathan@acm.org>
23390 * doc/extend.texi (Deprecated Features): Update deprecated flags,
23391 mention anon-struct/union members and trailing attributes.
23393 2018-03-21 Bin Cheng <bin.cheng@arm.com>
23395 PR tree-optimization/84969
23396 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
23397 builtin memset partitions if they set different rhs values.
23399 2018-03-21 Jakub Jelinek <jakub@redhat.com>
23401 PR rtl-optimization/84989
23402 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
23403 VEC_DUPLICATE with scalar result mode.
23405 2018-03-21 Martin Liska <mliska@suse.cz>
23408 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
23409 not intended return statement.
23411 2018-03-21 Martin Liska <mliska@suse.cz>
23414 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
23415 (chkp_find_bound_slots_1): Limit number of iterations.
23417 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
23420 * Minor grammar fixes for x86 options.
23422 2018-03-20 Jakub Jelinek <jakub@redhat.com>
23425 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
23426 holding REG_CFA_RESTORE notes, instead turn them into a USE.
23428 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
23431 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
23432 (altivec_lvx_<mode>_1op): Likewise.
23433 (altivec_stvx_<mode>_2op): Likewise.
23434 (altivec_stvx_<mode>_1op): Likewise.
23435 (altivec_lvx_<VM2:mode>): New define_expand.
23436 (altivec_stvx_<VM2:mode>): Likewise.
23437 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
23438 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
23439 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
23440 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
23441 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
23442 (rs6000_gen_lvx): Likewise.
23443 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
23444 (altivec_expand_stv_builtin): Likewise.
23445 (altivec_expand_builtin): Likewise.
23446 * config/rs6000/vector.md: Likewise.
23448 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
23451 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
23454 2018-03-20 Richard Biener <rguenther@suse.de>
23457 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
23458 sign-conversions as zero, fall back to standard scalar_stmt
23461 2018-03-20 Martin Liska <mliska@suse.cz>
23464 * predict.c (rebuild_frequencies): Handle case when we have
23465 PROFILE_ABSENT, but flag_guess_branch_prob is false.
23467 2018-03-20 Jakub Jelinek <jakub@redhat.com>
23470 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
23471 flag_section_anchors.
23472 * varasm.c (use_blocks_for_decl_p): Remove hack for
23473 dw2_force_const_mem.
23476 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
23478 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
23479 be created, use lowpart_subreg of operands[0] rather than operands[0]
23481 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
23482 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
23483 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
23484 and n constraint instead of aarch64_shift_imm_di and Usd.
23485 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
23486 (*aarch64_<optab>_reg_minus<mode>3): ... this.
23488 2018-03-20 Sudakshina Das <sudi.das@arm.com>
23491 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
23492 to favor GPR over NEON registers.
23493 (<shift>di3_neon): Likewise.
23495 2018-03-20 Tom de Vries <tom@codesourcery.com>
23498 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
23499 (nvptx_process_pars): Emit bar.sync asap and alap.
23501 2018-03-20 Tom de Vries <tom@codesourcery.com>
23504 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
23505 seen_label if seen_label is already set.
23507 2018-03-20 Jakub Jelinek <jakub@redhat.com>
23510 * config/i386/i386.c (fold_builtin_cpu): For features above 31
23511 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
23512 Use 1U instead of 1. Formatting fixes.
23515 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
23516 instead of TREE_TYPE (s1) for the return value.
23518 2018-03-19 Jakub Jelinek <jakub@redhat.com>
23520 PR tree-optimization/84946
23521 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
23522 bitsize + bitsize in poly_uint64 rather than poly_int64.
23525 * dwarf2asm.c: Include fold-const.c.
23526 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
23527 of decl rather than decl itself.
23529 PR rtl-optimization/84643
23530 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
23532 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
23535 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
23536 calling assemble_variable.
23538 2018-03-19 Sudakshina Das <sudi.das@arm.com>
23541 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
23542 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
23544 2018-03-19 Jim Wilson <jimw@sifive.com>
23547 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
23548 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
23549 (riscv_first_stack_step): Likewise.
23550 (riscv_option_override): Use STACK_BOUNDARY instead of
23551 MIN_STACK_BOUNDARY.
23552 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
23553 MIN_STACK_BOUNDARY.
23554 (BIGGEST_ALIGNMENT): Set to 128.
23555 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
23556 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
23559 2018-03-19 Richard Biener <rguenther@suse.de>
23561 PR tree-optimization/84933
23562 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
23563 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
23565 2018-03-19 Richard Biener <rguenther@suse.de>
23567 PR tree-optimization/84859
23568 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
23569 (cond_if_else_store_replacement): Perform sinking operation on
23570 single-store BBs regardless of MAX_STORES_TO_SINK setting.
23571 Generalize what a BB with a single eligible store is.
23573 2018-03-19 Richard Biener <rguenther@suse.de>
23575 PR tree-optimization/84929
23576 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
23577 chrec_is_positive against non-chrec arg.
23579 2018-03-19 Tamar Christina <tamar.christina@arm.com>
23582 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
23584 2018-03-18 Martin Liska <mliska@suse.cz>
23586 PR rtl-optimization/84635
23587 * regrename.c (build_def_use): Use matches_mode only when
23590 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
23592 PR tree-optimization/84913
23593 * tree-vect-loop.c (vectorizable_reduction): Don't try to
23594 vectorize chains of COND_EXPRs.
23596 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
23598 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
23600 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
23602 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
23604 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
23606 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
23608 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
23609 Kito Cheng <kito.cheng@gmail.com>
23611 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
23612 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
23613 (nds32_adjust_reg_alloc_order): New function.
23614 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
23616 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
23618 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
23619 nds32_print_operand, nds32_print_operand_address): Use
23620 HOST_WIDE_INT_PRINT_DEC instead.
23622 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
23624 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
23626 2018-03-17 Jakub Jelinek <jakub@redhat.com>
23629 * config/i386/i386.c (initial_ix86_tune_features,
23630 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
23631 unsigned long long.
23632 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
23633 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
23634 rather than 1u << ix86_tune. Formatting fix.
23635 (ix86_option_override_internal): Change ix86_arch_mask from
23636 unsigned int to unsigned HOST_WIDE_INT, initialize to
23637 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
23638 (ix86_function_specific_restore): Likewise.
23640 2018-03-16 Jakub Jelinek <jakub@redhat.com>
23643 * postreload.c (reload_combine_recognize_pattern): Perform
23644 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
23645 truncate_int_for_mode the result for the destination's mode.
23648 * hsa-gen.c (mem_type_for_type): Fix comment typo.
23649 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
23651 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23654 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
23657 * lra-assigns.c (lra_split_hard_reg_for): Don't use
23658 regno_allocno_class_array and sorted_pseudos.
23659 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
23660 insns where regno is used.
23662 2018-03-16 Martin Liska <mliska@suse.cz>
23665 * multiple_target.c (create_dispatcher_calls): Redirect
23666 reference in the symbol table.
23668 2018-03-16 Martin Liska <mliska@suse.cz>
23671 * multiple_target.c (create_dispatcher_calls): Redirect also
23674 2018-03-16 Jakub Jelinek <jakub@redhat.com>
23678 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
23679 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
23680 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
23682 2018-03-16 Julia Koval <julia.koval@intel.com>
23684 * doc/invoke.texi (Skylake Server): Add CLWB.
23685 Cannonlake): Remove CLWB.
23687 2018-03-16 Jakub Jelinek <jakub@redhat.com>
23689 PR tree-optimization/84841
23690 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
23692 (FLOAT_ONE_CONST_TYPE): Define.
23693 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
23694 (sort_by_operand_rank): Put entries with higher constant_type last
23695 rather than first to match comments.
23697 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
23699 * config/nios2/nios2.md (movsi_internal): Fix thinko in
23702 2018-03-15 Jakub Jelinek <jakub@redhat.com>
23705 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
23706 check and use address of target always.
23708 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
23711 * config/i386/i386.c (indirect_thunk_needed): Update comments.
23712 (indirect_thunk_bnd_needed): Likewise.
23713 (indirect_thunks_used): Likewise.
23714 (indirect_thunks_bnd_used): Likewise.
23715 (indirect_return_needed): New.
23716 (indirect_return_bnd_needed): Likewise.
23717 (output_indirect_thunk_function): Add a bool argument for
23719 (output_indirect_thunk_function): Don't generate alias for
23720 function return thunk.
23721 (ix86_code_end): Call output_indirect_thunk_function to generate
23722 function return thunks.
23723 (ix86_output_function_return): Set indirect_return_bnd_needed
23724 and indirect_return_needed instead of indirect_thunk_bnd_needed
23725 and indirect_thunk_needed.
23727 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
23729 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
23730 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
23731 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
23733 2018-03-15 David Malcolm <dmalcolm@redhat.com>
23734 Paul Hua <paul.hua.gm@gmail.com>
23737 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
23739 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
23741 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
23742 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
23743 resp. SFmode cases.
23745 2018-03-15 Tamar Christina <tamar.christina@arm.com>
23748 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
23749 instead of GET_MODE_SIZE when comparing Units.
23751 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
23754 * varasm.c (hash_section): Return an unchangeble hash value
23755 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
23756 Return !aarch64_can_use_per_function_literal_pools_p ().
23758 2018-03-15 Jakub Jelinek <jakub@redhat.com>
23761 * optabs.c (emit_conditional_move): Pass address of cmode's copy
23762 rather than address of cmode as last argument to prepare_cmp_insn.
23764 2018-03-15 Julia Koval <julia.koval@intel.com>
23766 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
23767 F_AVX512VNNI, F_AVX512BITALG): New.
23769 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
23772 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
23773 insn for floating-point loads and stores.
23775 2018-03-14 Carl Love <cel@us.ibm.com>
23777 * config/rs6000/rs6000-c.c: Add macro definitions for
23778 ALTIVEC_BUILTIN_VEC_PERMXOR.
23779 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
23780 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
23781 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
23782 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
23784 * config/doc/extend.texi: Add prototypes for vec_permxor.
23786 2018-03-14 David Malcolm <dmalcolm@redhat.com>
23789 * diagnostic-show-locus.c (class layout_point): Convert m_line
23790 from int to linenum_type.
23791 (line_span::comparator): Use linenum "compare" function when
23792 comparing line numbers.
23793 (test_line_span): New function.
23794 (layout_range::contains_point): Convert param "row" from int to
23796 (layout_range::intersects_line_p): Likewise.
23797 (layout::will_show_line_p): Likewise.
23798 (layout::print_source_line): Likewise.
23799 (layout::should_print_annotation_line_p): Likewise.
23800 (layout::print_annotation_line): Likewise.
23801 (layout::print_leading_fixits): Likewise.
23802 (layout::annotation_line_showed_range_p): Likewise.
23803 (struct line_corrections): Likewise for field m_row.
23804 (line_corrections::line_corrections): Likewise for param "row".
23805 (layout::print_trailing_fixits): Likewise.
23806 (layout::get_state_at_point): Likewise.
23807 (layout::get_x_bound_for_row): Likewise.
23808 (layout::print_line): Likewise.
23809 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
23810 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
23811 * input.c (selftest::test_linenum_comparisons): New function.
23812 (selftest::input_c_tests): Call it.
23813 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
23814 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
23815 * selftest.h (ASSERT_GT): New macro.
23816 (ASSERT_GT_AT): New macro.
23817 (ASSERT_LT): New macro.
23818 (ASSERT_LT_AT): New macro.
23820 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
23822 PR rtl-optimization/84780
23823 * combine.c (distribute_links): Don't make a link based on pc_rtx.
23825 2018-03-14 Martin Liska <mliska@suse.cz>
23827 * tree.c (record_node_allocation_statistics): Use
23828 get_stats_node_kind.
23829 (get_stats_node_kind): New function extracted from
23830 record_node_allocation_statistics.
23831 (free_node): Use get_stats_node_kind.
23833 2018-03-14 Richard Biener <rguenther@suse.de>
23835 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
23836 that the value-set of ANTIC_IN doesn't grow.
23839 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
23841 (BB_VISITED_WITH_VISITED_SUCCS): New define.
23842 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
23844 2018-03-14 Julia Koval <julia.koval@intel.com>
23846 * config.gcc (icelake-client, icelake-server): New.
23848 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
23849 (initial_ix86_arch_features): Ditto.
23850 (PTA_SKYLAKE): Add SGX.
23851 (PTA_ICELAKE): Remove.
23852 (PTA_ICELAKE_CLIENT): New.
23853 (PTA_ICELAKE_SERVER): New.
23854 (ix86_option_override_internal): Split up icelake on icelake client and
23856 (get_builtin_code_for_version): Ditto.
23857 (fold_builtin_cpu): Ditto.
23858 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
23859 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
23860 * config/i386/i386.h (processor_type): Ditto.
23861 * doc/invoke.texi: Ditto.
23863 2018-03-14 Jakub Jelinek <jakub@redhat.com>
23866 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
23867 INTEGER_CST offset, add it together with bitpos / 8 and
23868 sign extend based on POINTER_SIZE.
23872 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
23875 * config/i386/constraints.md (Yc): New register constraint.
23876 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
23877 Use Yc constraint for alternative 2 of operand 0. Remove
23878 preferred_for_speed attribute.
23880 2018-03-14 Richard Biener <rguenther@suse.de>
23882 PR tree-optimization/84830
23883 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
23884 with the old one to avoid oscillations.
23886 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
23889 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
23891 (assign_by_spills): Return a flag of reload assignment failure.
23892 Do not process the reload assignment failures. Do not spill other
23893 reload pseudos if they has the same reg class. Update n if
23895 (lra_assign): Add a return arg. Set up from the result of
23896 assign_by_spills call.
23897 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23898 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
23899 usage_insns if it is not NULL.
23900 (spill_hard_reg_in_range): New function.
23901 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23902 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23903 function prototypes.
23904 (lra_assign): Change prototype.
23905 * lra.c (lra): Add code to deal with fails by splitting hard reg
23908 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
23910 * config/riscv/riscv.opt (mrelax): New option.
23911 * config/riscv/riscv.c (riscv_file_start): Emit ".option
23912 "norelax" when riscv_mrelax is disabled.
23913 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
23915 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
23918 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
23919 reassociation for int modes.
23921 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
23923 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
23924 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
23926 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
23927 * config/aarch64/aarch64-sve.md
23928 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
23929 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
23930 (*extend<mode><Vwide>2): Rename to...
23931 (aarch64_sve_extend<mode><Vwide>2): ...this.
23932 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
23933 renaming the old pattern to...
23934 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
23936 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
23937 define_expand, renaming the old pattern to...
23938 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
23939 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
23940 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
23941 account when deciding which SVE instruction the optab should use.
23942 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
23944 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
23946 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
23947 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
23948 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
23949 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
23950 (tlsdesc_small_<mode>): Turn a define_expand and use
23951 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
23952 (tlsdesc_small_advsimd_<mode>): ...this.
23953 (tlsdesc_small_sve_<mode>): New pattern.
23955 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
23957 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
23958 (UNSPEC_UMUL_HIGHPART): New constants.
23959 (MUL_HIGHPART): New int iteraor.
23960 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
23961 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
23963 (*<su>mul<mode>3_highpart): New define_insn.
23965 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
23968 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
23971 2018-03-13 Martin Liska <mliska@suse.cz>
23974 * (sem_item_optimizer::sem_item_optimizer): Initialize new
23976 (sem_item_optimizer::~sem_item_optimizer): Release it.
23977 (sem_item_optimizer::merge_classes): Register variable aliases.
23978 (sem_item_optimizer::fixup_pt_set): New function.
23979 (sem_item_optimizer::fixup_points_to_sets): Likewise.
23980 * ipa-icf.h: Declare new variables and functions.
23982 2018-03-13 Jakub Jelinek <jakub@redhat.com>
23984 PR middle-end/84834
23985 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
23986 integer_pow2p@2 and test integer_pow2p in condition.
23987 (A < 0 ? C : 0): Similarly for @1.
23989 PR middle-end/84831
23990 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
23991 characters starting at p contain '\0' character, don't look beyond
23995 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
23996 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
23999 * reg-stack.c (change_stack): Change update_end var from int to
24000 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
24001 also call set_block_for_insn on the newly added insns and rescan.
24004 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
24005 on the last operand.
24008 * tree.c (stabilize_reference_1): Return save_expr (e) for
24009 STATEMENT_LIST even if it doesn't have side-effects.
24011 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
24013 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
24015 2018-03-12 Renlin Li <renlin.li@arm.com>
24017 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
24018 aarch64_output_scalar_simd_mov_immediate.
24020 2018-03-12 Martin Sebor <msebor@redhat.com>
24022 PR tree-optimization/83456
24023 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
24024 for perfectly overlapping calls to memcpy.
24025 (gimple_fold_builtin_memory_chk): Same.
24026 (gimple_fold_builtin_strcpy): Handle no-warning.
24027 (gimple_fold_builtin_stxcpy_chk): Same.
24028 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
24030 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
24032 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
24033 parameter. Use it for SFmode.
24034 (rs6000_function_arg_advance_1): Adjust.
24035 (rs6000_function_arg): Adjust.
24036 (rs6000_gimplify_va_arg): Pass false for that new parameter.
24038 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
24040 PR rtl-optimization/84169
24041 PR rtl-optimization/84780
24042 * combine.c (can_combine_p): Check for a 2-insn combination whether
24043 the destination register is used between the two insns, too.
24045 2018-03-12 Richard Biener <rguenther@suse.de>
24047 PR tree-optimization/84803
24048 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
24049 for refs DR analysis didn't process.
24051 2018-03-12 Richard Biener <rguenther@suse.de>
24053 PR tree-optimization/84777
24054 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
24055 force-vectorize loops ignore whether we are optimizing for size.
24057 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
24059 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
24060 (TARGET_MD_ASM_ADJUST): Define.
24062 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
24063 Kito Cheng <kito.cheng@gmail.com>
24064 Chung-Ju Wu <jasonwucj@gmail.com>
24066 * config/nds32/nds32.c (nds32_compute_stack_frame,
24067 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
24068 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
24069 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
24070 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
24071 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
24072 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
24073 * config/nds32/nds32.md (prologue, epilogue): Use macro
24074 NDS32_V3PUSH_AVAILABLE_P to do checking.
24076 2018-03-11 Jakub Jelinek <jakub@redhat.com>
24079 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
24080 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
24081 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
24082 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
24083 addition of most attributes on !orig_type_die or the attribute not
24084 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
24086 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
24087 Chung-Ju Wu <jasonwucj@gmail.com>
24089 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
24090 __NDS32_VH__ macro.
24091 * config/nds32/nds32.opt (mvh): New option.
24093 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
24094 Chung-Ju Wu <jasonwucj@gmail.com>
24096 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
24098 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
24099 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
24102 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
24103 Chung-Ju Wu <jasonwucj@gmail.com>
24105 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
24107 * config/nds32/nds32-multiple.md (strlensi): New pattern.
24108 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
24110 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
24111 Kito Cheng <kito.cheng@gmail.com>
24112 Chung-Ju Wu <jasonwucj@gmail.com>
24114 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
24115 UNSPEC_FFMISM and UNSPEC_FLMISM.
24116 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
24117 for ffb, ffmism and flmism.
24118 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
24119 (unspec_ffmism): Ditto.
24120 (unspec_flmism): Ditto.
24121 (nds32_expand_builtin_impl): Check if string extension is available.
24122 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
24123 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
24125 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
24128 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
24131 * lra-assigns.c (assign_by_spills): Return a flag of reload
24132 assignment failure. Do not process the reload assignment
24133 failures. Do not spill other reload pseudos if they has the same
24135 (lra_assign): Add a return arg. Set up from the result of
24136 assign_by_spills call.
24137 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
24138 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
24139 usage_insns if it is not NULL.
24140 (spill_hard_reg_in_range): New function.
24141 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
24142 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
24143 function prototypes.
24144 (lra_assign): Change prototype.
24145 * lra.c (lra): Add code to deal with fails by splitting hard reg
24148 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
24151 * config/i386/i386.opt: Replace Enforcment with Enforcement.
24153 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
24156 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
24157 (dw_val_node): Add val_symbolic_view.
24158 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
24159 (symview_upper_bound): New.
24160 (new_line_info_table): Initialize symviews_since_reset.
24161 (dwarf2out_source_line): Count symviews_since_reset and set
24162 symview_upper_bound.
24163 (dw_val_equal_p): Handle symview.
24164 (add_AT_symview): New.
24165 (print_dw_val): Handle symview.
24166 (attr_checksum, attr_checksum_ordered): Likewise.
24167 (same_dw_val_p, size_of_die): Likewise.
24168 (value_format, output_die): Likewise.
24169 (add_high_low_attributes): Use add_AT_symview for entry_view.
24170 (dwarf2out_finish): Reset symview_upper_bound, clear
24173 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
24176 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
24177 Add strict argument and use it.
24178 (rs6000_split_multireg_move): Update for new strict argument.
24179 (mem_operand_gpr): Disallow all non-offsettable addresses.
24180 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
24182 2018-03-09 Jakub Jelinek <jakub@redhat.com>
24185 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
24186 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
24187 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
24190 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
24191 decl, use remap_type if we want to use the type.
24193 2018-03-09 Martin Sebor <msebor@redhat.com>
24195 PR tree-optimization/84526
24196 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
24198 (builtin_access::generic_overlap): Be prepared to handle non-array
24201 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
24203 PR rtl-optimization/84682
24204 * lra-constraints.c (process_address_1): Check is_address flag
24205 for address constraints.
24206 (process_alt_operands): Likewise.
24207 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
24208 preprocess_constraints.
24209 * recog.h (preprocess_constraints): Add oploc parameter.
24211 * recog.c (preprocess_constraints): Test address_operand for
24212 CT_ADDRESS constraints.
24214 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
24217 * lra-assigns.c (assign_by_spills): Return a flag of reload
24218 assignment failure. Do not process the reload assignment
24219 failures. Do not spill other reload pseudos if they has the same
24221 (lra_assign): Add a return arg. Set up from the result of
24222 assign_by_spills call.
24223 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
24224 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
24225 usage_insns if it is not NULL.
24226 (spill_hard_reg_in_range): New function.
24227 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
24228 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
24229 function prototypes.
24230 (lra_assign): Change prototype.
24231 * lra.c (lra): Add code to deal with fails by splitting hard reg
24234 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24237 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
24238 Accept complain bool parameter. Only emit errors if it is true.
24239 (arm_parse_cpu_option_name): Likewise.
24240 (arm_target_thumb_only): Adjust callers of the above.
24241 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
24242 prototype to take a default true bool parameter.
24243 (arm_parse_arch_option_name): Likewise.
24245 2018-03-09 David Malcolm <dmalcolm@redhat.com>
24246 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
24250 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
24251 * configure: Regenerate.
24252 * configure.ac ("linker --version-script option"): New.
24253 ("linker soname option"): New.
24255 2018-03-09 Richard Biener <rguenther@suse.de>
24257 PR tree-optimization/84775
24258 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
24259 immediate uses of predicate stmts and mark them modified.
24262 PR tree-optimization/84178
24263 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
24265 (version_loop_for_if_conversion): Delay update_ssa call.
24266 (tree_if_conversion): Delay update_ssa until after predicate
24269 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
24272 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
24273 when the function accesses prior frames.
24275 2018-03-08 Jakub Jelinek <jakub@redhat.com>
24278 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
24279 gen_llsym, otherwise call maybe_gen_llsym.
24281 PR inline-asm/84742
24282 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
24283 has ',' character inside of it.
24285 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
24288 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
24289 as clobbering CC_REGNUM.
24291 2018-03-08 Richard Biener <rguenther@suse.de>
24293 PR middle-end/84552
24294 * tree-scalar-evolution.c: Include tree-into-ssa.h.
24295 (follow_copies_to_constant): Do not follow SSA names registered
24298 2018-03-08 Richard Biener <rguenther@suse.de>
24300 PR tree-optimization/84178
24301 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
24303 (version_loop_for_if_conversion): Delay update_ssa call.
24304 (tree_if_conversion): Delay update_ssa until after predicate
24307 2018-03-08 David Malcolm <dmalcolm@redhat.com>
24309 PR tree-optimization/84178
24310 * tree-if-conv.c (release_bb_predicate): Remove the
24311 the assertion that the stmts have NULL use_ops.
24312 Discard the statements, asserting that they haven't
24313 yet been added to a BB.
24315 2018-03-08 Richard Biener <rguenther@suse.de>
24317 PR tree-optimization/84746
24318 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
24319 (phi_translate): Pass in destination ANTIC_OUT set.
24320 (phi_translate_1): Likewise. For a simplified result lookup
24321 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
24322 (phi_translate_set): Adjust.
24323 (do_pre_regular_insertion): Likewise.
24324 (do_pre_partial_partial_insertion): Likewise.
24326 2018-03-08 Martin Liska <mliska@suse.cz>
24328 PR gcov-profile/84735
24329 * doc/gcov.texi: Document usage of profile files.
24330 * gcov-io.h: Document changes in the format.
24332 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
24336 * dwarf2out.c (struct dw_line_info_table): Update comments for
24338 (FORCE_RESET_NEXT_VIEW): New.
24339 (FORCE_RESETTING_VIEW_P): New.
24340 (RESETTING_VIEW_P): Check for -1 too.
24341 (ZERO_VIEW_P): Likewise.
24342 (new_line_info_table): Force-reset next view.
24343 (dwarf2out_begin_function): Likewise.
24344 (dwarf2out_source_line): Simplify zero_view_p initialization.
24345 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
24346 view directly. Omit view when omitting .loc at line 0.
24348 2018-03-08 Jakub Jelinek <jakub@redhat.com>
24350 PR tree-optimization/84740
24351 * tree-switch-conversion.c (process_switch): Call build_constructors
24352 only if info.phi_count is non-zero.
24354 PR tree-optimization/84739
24355 * tree-tailcall.c (find_tail_calls): Check call arguments against
24356 DECL_ARGUMENTS (current_function_decl) rather than
24357 DECL_ARGUMENTS (func) when checking for tail recursion.
24359 2018-03-07 Jakub Jelinek <jakub@redhat.com>
24361 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
24362 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
24363 Volker Reichelt's entry and add entries for people that perform
24364 GCC fuzzy testing and report numerous bugs.
24366 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
24369 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
24370 readonly data in sdata, if that is disabled.
24371 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
24372 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
24373 -mreadonly-in-sdata option.
24375 2018-03-07 Martin Sebor <msebor@redhat.com>
24377 PR tree-optimization/84468
24378 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
24379 basic block when looking for nul assignment.
24381 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
24384 * except.h (output_function_exception_table): Adjust prototype.
24385 * except.c (output_function_exception_table): Remove FNNAME parameter
24386 and add SECTION parameter. Ouput one part of the table at a time.
24387 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
24388 the first part of the exception table and emit unwind directives.
24389 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
24390 (i386_pe_seh_cold_init): Likewise.
24391 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
24392 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
24393 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
24394 (ix86_output_call_insn): Emit a nop in one more case for SEH.
24395 * config/i386/winnt.c: Include except.h.
24396 (struct seh_frame_state): Add reg_offset, after_prologue and
24397 in_cold_section fields.
24398 (i386_pe_seh_end_prologue): Set seh->after_prologue.
24399 (i386_pe_seh_cold_init): New function.
24400 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
24401 to seh->in_cold_section.
24402 (seh_emit_push): Record the offset of the push.
24403 (seh_emit_save): Record the offet of the save.
24404 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
24405 Test seh->after_prologue to disregard the epilogue.
24406 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
24407 (i386_pe_end_cold_function): New function.
24409 2018-03-07 Jakub Jelinek <jakub@redhat.com>
24412 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
24413 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
24416 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
24418 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
24419 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
24421 PR middle-end/84723
24422 * multiple_target.c: Include tree-inline.h and intl.h.
24423 (expand_target_clones): Diagnose and fail if node->definition and
24424 !tree_versionable_function_p (node->decl).
24426 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
24428 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
24430 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
24431 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
24432 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
24434 2018-03-06 Jakub Jelinek <jakub@redhat.com>
24437 * combine.c (try_combine): Use reg_or_subregno instead of handling
24438 just paradoxical SUBREGs and REGs.
24440 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
24442 * config/arc/arc.c (arc_finalize_pic): Remove function.
24443 (arc_must_save_register): We use single base PIC register, remove
24444 checks to save/restore the PIC register.
24445 (arc_expand_prologue): Likewise.
24446 * config/arc/arc-protos.h (arc_set_default_type_attributes):
24448 (arc_verify_short): Likewise.
24449 (arc_attr_type): Likewise.
24450 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
24451 (walk_stores): Likewise.
24452 (arc_address_cost): Make it static.
24453 (arc_verify_short): Likewise.
24454 (branch_dest): Likewise.
24455 (arc_attr_type): Likewise.
24456 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
24457 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
24458 (arc_final_prescan_insn): Remove inserting the nops due to
24459 hardware hazards. It is done in reorg step.
24460 (insn_length_variant_t): Remove.
24461 (insn_length_parameters_t): Likewise.
24462 (arc_insn_length_parameters): Likewise.
24463 (arc_get_insn_variants): Likewise.
24464 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
24466 2018-03-06 Jakub Jelinek <jakub@redhat.com>
24468 PR inline-asm/84683
24469 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
24472 PR tree-optimization/84687
24473 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
24475 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
24477 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
24479 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
24480 Rename to ppc_speculation_barrier.
24481 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
24482 __builtin_ppc_speculation_barrier.
24484 2018-03-05 Jakub Jelinek <jakub@redhat.com>
24487 * combine.c (combine_simplify_rtx): Don't try to simplify if
24488 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
24491 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
24493 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
24494 to 32 bytes when compiling for POWER9.
24496 2018-03-05 Jakub Jelinek <jakub@redhat.com>
24499 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
24500 regparm >= 3 with no arg reg available also for calls with
24501 flag_force_indirect_call. Pass decl to ix86_function_regparm.
24504 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
24506 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
24508 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
24511 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
24513 2018-03-05 Richard Biener <rguenther@suse.de>
24515 PR tree-optimization/84486
24516 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
24517 When inserting a __builtin_assume_aligned call set the LHS
24518 SSA name alignment info accordingly.
24520 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
24522 PR tree-optimization/84114
24523 * config/aarch64/aarch64.c (aarch64_reassociation_width)
24524 Avoid reassociation of FLOAT_MODE addition.
24526 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
24528 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
24529 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
24530 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
24531 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
24532 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
24533 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
24534 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
24536 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
24537 __builtin_ia32_wbinvd): New builtins.
24538 (SPECIAL_ARGS2): New.
24539 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
24540 (SPECIAL_ARGS2): New.
24541 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
24542 (ix86_valid_target_attribute_inner_p): Ditto.
24543 (ix86_init_mmx_sse_builtins): Add special_args2.
24544 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
24545 TARGET_WBNOINVD_P): New.
24546 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
24547 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
24548 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
24549 * config/i386/immintrin.h (_wbinvd): New intrinsic.
24550 * config/i386/pconfigintrin.h: New file.
24551 * config/i386/wbnoinvdintrin.h: Ditto.
24552 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
24554 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
24556 2018-03-05 Richard Biener <rguenther@suse.de>
24558 PR tree-optimization/84670
24559 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
24561 (BB_VISITED_WITH_VISITED_SUCCS): New define.
24562 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
24563 (compute_antic_aux): Only assert the number of values in ANTIC_IN
24564 doesn't grow if all successors (recursively) were visited at least
24567 2018-03-05 Richard Biener <rguenther@suse.de>
24569 PR tree-optimization/84650
24570 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
24571 if executed in the loop pipeline.
24573 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
24575 * doc/configfiles.texi (Configuration Files): Move info about
24576 conditionalizing $target-protos.h to...
24577 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
24578 differs from $target-protos.h.
24580 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
24581 Chung-Ju Wu <jasonwucj@gmail.com>
24583 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
24584 * config/nds32/nds32-multiple.md (setmemsi): Define.
24585 * config/nds32/nds32-memory-manipulation.c
24586 (nds32_gen_dup_4_byte_to_word_value): New.
24587 (emit_setmem_word_loop): New.
24588 (emit_setmem_byte_loop): New.
24589 (nds32_expand_setmem_loop): New.
24590 (nds32_expand_setmem_loop_v3m): New.
24591 (nds32_expand_setmem_unroll): New.
24592 (nds32_expand_setmem): New.
24594 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
24595 Chung-Ju Wu <jasonwucj@gmail.com>
24597 * config/nds32/nds32-memory-manipulation.c
24598 (nds32_emit_load_store): New.
24599 (nds32_emit_post_inc_load_store): New.
24600 (nds32_emit_mem_move): New.
24601 (nds32_emit_mem_move_block): New.
24602 (nds32_expand_movmemsi_loop_unknown_size): New.
24603 (nds32_expand_movmemsi_loop_known_size): New.
24604 (nds32_expand_movmemsi_loop): New.
24605 (nds32_expand_movmemsi_unroll): New.
24606 (nds32_expand_movmemqi): Rename ...
24607 (nds32_expand_movmemsi): ... to this.
24608 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
24609 (movmemsi): ... to this.
24610 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
24611 (nds32_expand_movmemsi): ... to this.
24613 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
24614 Monk Chiang <sh.chiang04@gmail.com>
24615 Chung-Ju Wu <jasonwucj@gmail.com>
24617 * config/nds32/nds32-protos.h
24618 (nds32_expand_load_multiple): New arguments.
24619 (nds32_expand_store_multiple): Ditto.
24620 (nds32_valid_multiple_load_store): Rename ...
24621 (nds32_valid_multiple_load_store_p): ... to this.
24622 * config/nds32/nds32-memory-manipulation.c
24623 (nds32_expand_load_multiple): Refine implementation.
24624 (nds32_expand_store_multiple): Ditto.
24625 * config/nds32/nds32-multiple.md
24626 (load_multiple): Update nds32_expand_load_multiple interface.
24627 (store_multiple): Update nds32_expand_store_multiple interface.
24628 * config/nds32/nds32-predicates.c
24629 (nds32_valid_multiple_load_store): Rename ...
24630 (nds32_valid_multiple_load_store_p): ... to this and refine
24632 * config/nds32/predicates.md
24633 (nds32_load_multiple_and_update_address_operation): New predicate.
24634 (nds32_store_multiple_and_update_address_operation): New predicate.
24636 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
24637 Chung-Ju Wu <jasonwucj@gmail.com>
24639 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
24640 (combo): New attribute.
24641 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
24643 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
24645 * config/nds32/nds32.opt: Change -mcmodel= default value.
24647 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
24648 Monk Chiang <sh.chiang04@gmail.com>
24649 Chung-Ju Wu <jasonwucj@gmail.com>
24651 * config/nds32/constants.md (unspec_element): New enum.
24652 * config/nds32/constraints.md (Umw): New constraint.
24653 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
24654 * config/nds32/nds32-intrinsic.md: Likewise.
24655 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
24656 (nds32_valid_smw_lwm_base_p): New.
24657 (nds32_output_smw_single_word): New.
24658 (nds32_output_lmw_single_word): New.
24659 (nds32_expand_unaligned_load): New.
24660 (nds32_expand_unaligned_store): New.
24661 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
24662 (nds32_output_smw_single_word): Declare.
24663 (nds32_output_lmw_single_word): Declare.
24664 (nds32_expand_unaligned_load): Declare.
24665 (nds32_expand_unaligned_store): Declare.
24666 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
24667 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
24668 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
24669 NDS32_BUILTIN_UASTORE_DW.
24670 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
24673 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
24674 Kito Cheng <kito.cheng@gmail.com>
24675 Chung-Ju Wu <jasonwucj@gmail.com>
24677 * config/nds32/nds32-intrinsic.c
24678 (nds32_expand_builtin_null_ftype_reg): Delete.
24679 (nds32_expand_builtin_reg_ftype_imm): Ditto.
24680 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
24681 (nds32_read_argument): New.
24682 (nds32_legitimize_target): Ditto.
24683 (nds32_legitimize_argument): Ditto.
24684 (nds32_check_constant_argument): Ditto.
24685 (nds32_expand_unop_builtin): Ditto.
24686 (nds32_expand_unopimm_builtin): Ditto.
24687 (nds32_expand_binop_builtin): Ditto.
24688 (nds32_builtin_decl_impl): Ditto.
24689 (builtin_description): Ditto.
24690 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
24691 (nds32_init_builtins_impl): Ditto.
24692 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
24693 (nds32_builtin_decl): New.
24694 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
24695 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
24697 2018-03-02 Jeff Law <law@redhat.com>
24699 * reorg.c (stop_search_p): Handle DEBUG_INSN.
24700 (redundant_insn, fill_simple_delay_slots): Likewise.
24701 (fill_slots_from_thread): Likewise.
24702 * resource.c (mark_referenced_resources): Likewise.
24703 (mark_set_resources, find_dead_or_set_registers): Likewise.
24705 2018-03-02 Jakub Jelinek <jakub@redhat.com>
24707 * substring-locations.h (format_warning_va): Formatting fix for
24708 ATTRIBUTE_GCC_DIAG.
24709 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
24711 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
24712 * substring-locations.c: Include intl.h.
24713 (format_warning_va): Turned into small wrapper around
24714 format_warning_n_va, renamed to ...
24715 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
24716 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
24718 (format_warning_at_substring_n): New function.
24719 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
24720 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
24721 format_warning_at_substring with just a shorter name instead of
24722 const function pointer.
24723 (fmtwarn_n): New function.
24724 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
24725 appropriate, get rid of all the fmtstr temporaries, move conditionals
24726 with G_() wrapped string literals directly into fmtwarn arguments,
24727 cast dir.len to (int), formatting fixes.
24729 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
24731 * doc/invoke.texi: Remove "Cilk Plus" references.
24733 2018-03-02 Jakub Jelinek <jakub@redhat.com>
24734 Richard Biener <rguenther@suse.de>
24737 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
24738 for error or warning attributes if CALL_FROM_THUNK_P is set.
24741 2018-03-02 Jakub Jelinek <jakub@redhat.com>
24744 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
24745 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
24748 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
24749 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
24751 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
24752 instead of -1U in last predictors element's probability member.
24754 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
24757 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
24758 arguments if they are comparable.
24760 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
24762 PR tree-optimization/84634
24763 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
24764 masks and masked_loop_p with a single loop_masks, making sure it's
24765 null for bb vectorization.
24767 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
24769 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
24770 (vect_analyze_data_ref_access): Use loop->safe_len rather than
24771 loop->force_vectorize to check whether there is no alias.
24773 2018-03-02 Jakub Jelinek <jakub@redhat.com>
24776 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
24778 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
24780 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
24781 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
24782 instead of a loop around prev_real_insn.
24783 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
24786 PR inline-asm/84625
24787 * config/i386/i386.c (ix86_print_operand): Use conditional
24788 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
24791 2018-03-02 Richard Biener <rguenther@suse.de>
24793 PR tree-optimization/84427
24794 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
24795 (bitmap_set_subtract_values): Rewrite to handle multiple
24798 (prune_clobbered_mems): Likewise.
24799 (phi_translate): Take edge instead of pred/phiblock.
24800 (phi_translate_1): Likewise.
24801 (phi_translate_set): Likewise. Insert all translated
24802 exprs for a value into the set, keeping possibly multiple
24803 expressions per value.
24804 (compute_antic_aux): Adjust for phi_translate changes.
24805 When intersecting union the expressions and prune those
24806 not in the final value set, keeping possibly multiple
24807 expressions per value. Do not use value-insertion
24808 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
24809 all expressions. Add verification that the value-sets
24810 only shrink during iteration.
24811 (compute_partial_antic_aux): Adjust for the phi_translate changes.
24812 (do_pre_regular_insertion): Likewise.
24813 (do_pre_partial_partial_insertion): Likewise.
24815 2018-03-02 Richard Biener <rguenther@suse.de>
24818 * config/darwin.c (saved_debug_info_level): New static global.
24819 (darwin_asm_lto_start): Disable debug info generation for LTO out.
24820 (darwin_asm_lto_end): Restore debug info generation settings.
24822 2018-03-01 Martin Liska <mliska@suse.cz>
24825 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
24826 volatile arguments.
24828 2018-03-01 Richard Biener <rguenther@suse.de>
24831 * dwarf2out.c (gen_variable_die): Properly handle late VLA
24832 type annotation with LTO when debug was disabled at compile-time.
24834 2018-03-01 Matthew Fortune <mfortune@gmail.com>
24836 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
24838 (mips_final_postscan_insn): Likewise.
24840 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
24842 PR rtl-optimization/84528
24843 * alias.c (init_alias_target): Add commentary.
24844 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
24845 a unique base value if the frame pointer is not eliminated
24846 to the stack pointer.
24848 2018-03-01 Tom de Vries <tom@codesourcery.com>
24850 PR rtl-optimization/83327
24851 * lra-int.h (hard_regs_spilled_into): Declare.
24852 * lra.c (hard_regs_spilled_into): Define.
24853 (init_reg_info): Init hard_regs_spilled_into.
24854 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
24855 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
24856 (process_bb_lives): Handle hard_regs_spilled_into.
24857 (lra_create_live_ranges_1): Before doing liveness propagation, clear
24858 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
24860 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
24862 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
24863 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
24864 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
24865 * config/rs6000/aix72.h: New file.
24867 2018-02-28 Jakub Jelinek <jakub@redhat.com>
24869 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
24870 instead of warning_at with conditional singular and plural messages
24874 * stor-layout.c (update_alignment_for_field): For
24875 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
24876 && !DECL_PACKED (field), do the alignment update, just use
24877 only desired_align instead of MAX (type_align, desired_align)
24879 (place_field): Don't do known_align < desired_align handling
24880 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
24881 is non-NULL, instead do it after rli->prev_field handling and
24882 only if not within a bitfield word. For DECL_PACKED (field)
24883 use type_align of BITS_PER_UNIT.
24885 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
24887 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
24888 superfluous parentheses and trailing spaces.
24890 2018-02-28 Richard Biener <rguenther@suse.de>
24892 PR tree-optimization/84584
24893 * graphite-scop-detection.c (scop_detection::add_scop): Discard
24894 SCoPs with fake exit edge.
24896 2018-02-28 Martin Liska <mliska@suse.cz>
24899 * timevar.c (timer::print): Fix format to properly print 100%
24902 2018-02-28 Richard Biener <rguenther@suse.de>
24904 PR middle-end/84607
24905 * genmatch.c (capture_info::walk_match): Do not mark
24906 captured expressions without operands as expr_p given
24907 they act more like predicates and should be subject to
24908 "lost tail" side-effect preserving.
24910 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
24912 PR rtl-optimization/81611
24913 * auto-inc-dec.c (attempt_change): Move dead note from
24914 mem_insn if it's the next use of regno
24915 (find_address): Take address use of reg holding
24916 non-incremented value. Add parm to limit search to the named
24918 (merge_in_block): Attempt to use a mem insn that is the next
24919 use of the original regno.
24921 2018-02-27 Martin Sebor <msebor@redhat.com>
24924 * doc/invoke.texi (-Wmissing-attributes): New option.
24925 * print-tree.c (print_node): Handle DECL_UNINLINABLE.
24927 2018-02-27 Martin Sebor <msebor@redhat.com>
24929 PR translation/84207
24930 * diagnostic-core.h (warning_n, error_n, inform_n): Change
24931 n argument to unsigned HOST_WIDE_INT.
24932 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
24933 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
24934 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
24935 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
24937 2018-02-27 Richard Biener <rguenther@suse.de>
24939 PR tree-optimization/84512
24940 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
24941 Do not use the estimate returned from record_stmt_cost for
24942 the scalar iteration cost but sum properly using add_stmt_cost.
24944 2018-02-27 Richard Biener <rguenther@suse.de>
24946 PR tree-optimization/84466
24947 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
24948 Adjust last change to less strictly validate use operands.
24950 2018-02-27 Martin Liska <mliska@suse.cz>
24952 PR gcov-profile/84548
24953 * gcov.c (process_file): Allow partial overlap and consider it
24954 also as group functions.
24955 (output_lines): Properly calculate range of lines for a group.
24957 2018-02-27 Martin Liska <mliska@suse.cz>
24959 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
24960 'ggc' suffixes. Change first column width.
24961 (timer::print): Fix formatting of the column.
24963 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
24965 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
24966 preserve inline entry blocks for the sake of debug inline
24967 entry point markers alone.
24968 (remove_unused_locals): Suggest in comments a better place to
24969 force the preservation of inline entry blocks that are
24970 otherwise unused, but do not preserve them.
24972 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
24974 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
24976 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
24979 * config/i386/constraints.md (Bs): Replace
24980 ix86_indirect_branch_register with
24981 TARGET_INDIRECT_BRANCH_REGISTER.
24983 * config/i386/i386.md (indirect_jump): Likewise.
24984 (tablejump): Likewise.
24985 (*sibcall_memory): Likewise.
24986 (*sibcall_value_memory): Likewise.
24987 Peepholes of indirect call and jump via memory: Likewise.
24988 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
24989 (*sibcall_value_GOT_32): Likewise.
24990 * config/i386/predicates.md (indirect_branch_operand): Likewise.
24991 (GOT_memory_operand): Likewise.
24992 (call_insn_operand): Likewise.
24993 (sibcall_insn_operand): Likewise.
24994 (GOT32_symbol_operand): Likewise.
24995 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
24997 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
24999 PR rtl-optimization/83496
25000 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
25001 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
25002 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
25003 redundant insn, if any.
25004 (relax_delay_slots): Likewise.
25005 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
25007 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
25009 PR tree-optimization/83965
25010 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
25011 that grouped statements are part of a reduction chain. Return
25012 true if the statement is not marked as a reduction itself but
25013 is part of a group.
25014 (vect_recog_dot_prod_pattern): Don't check whether the statement
25015 is part of a group here.
25016 (vect_recog_sad_pattern): Likewise.
25017 (vect_recog_widen_sum_pattern): Likewise.
25019 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
25022 * final.c (rest_of_clean_state): Also look for calls inside sequences.
25024 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
25027 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
25029 (ix86_output_indirect_function_return): New prototype.
25030 (ix86_split_simple_return_pop_internal): Likewise.
25031 * config/i386/i386.c (indirect_return_via_cx): New.
25032 (indirect_return_via_cx_bnd): Likewise.
25033 (indirect_thunk_name): Handle return va CX_REG.
25034 (output_indirect_thunk_function): Create alias for
25035 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
25036 (ix86_output_indirect_jmp): Remove the bool argument.
25037 (ix86_output_indirect_function_return): New function.
25038 (ix86_split_simple_return_pop_internal): Likewise.
25039 * config/i386/i386.md (*indirect_jump): Don't pass false
25040 to ix86_output_indirect_jmp.
25041 (*tablejump_1): Likewise.
25042 (simple_return_pop_internal): Change it to define_insn_and_split.
25043 Call ix86_split_simple_return_pop_internal to split it for
25044 -mfunction-return=.
25045 (simple_return_indirect_internal): Call
25046 ix86_output_indirect_function_return instead of
25047 ix86_output_indirect_jmp.
25049 2018-02-26 Jakub Jelinek <jakub@redhat.com>
25052 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
25053 memset and value initialization afterwards.
25055 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
25057 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
25059 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
25062 * common/config/aarch64/aarch64-common.c
25063 (aarch_option_optimization_table[]): Switch
25064 off fomit-frame-pointer
25066 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
25067 Chung-Ju Wu <jasonwucj@gmail.com>
25069 * config/nds32/nds32-multiple.md (load_multiple): Disallow
25071 (store_multiple): Ditto.
25073 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
25075 * config.gcc: Add --with-cpu support for nds32 target.
25076 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
25077 * config/nds32/nds32.opt: Add -mcpu= option.
25079 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
25081 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
25082 isel=yes): Warn for these deprecated options.
25084 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
25086 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
25087 ISA_2_5_MASKS_EMBEDDED.
25089 2018-02-23 Jakub Jelinek <jakub@redhat.com>
25091 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
25092 p->max as pointers rather than using iterative_hash_expr.
25094 2018-02-23 Carl Love <cel@us.ibm.com>
25096 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
25097 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
25099 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
25100 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
25101 P8V_BUILTIN_VEC_VUNSIGNED2.
25103 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
25106 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
25107 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
25109 * lra-constraints.c (curr_insn_transform): Set up
25110 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
25112 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
25113 * lra-eliminations.c (spill_pseudos): Ditto.
25114 (process_insn_for_elimination): Ditto.
25115 * lra-lives.c (reg_early_clobber_p): Use the new macros.
25116 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
25117 LRA_NON_CLOBBERED_ALT.
25119 2018-02-22 Martin Sebor <msebor@redhat.com>
25121 PR tree-optimization/84480
25122 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
25123 to maybe_diag_stxncpy_trunc. Call it.
25124 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
25125 from gimple_fold_builtin_strcpy. Print inlining stack.
25126 (handle_builtin_stxncpy): Print inlining stack.
25127 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
25129 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
25132 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
25133 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
25134 and -fcheck-pointer-bounds are used together.
25135 (indirect_thunk_prefix): New enum.
25136 (indirect_thunk_need_prefix): New function.
25137 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
25138 "_nt" instead of "_bnd" for NOTRACK prefix.
25139 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
25140 (output_indirect_thunk_function): Likewise.
25142 (ix86_code_end): Update output_indirect_thunk_function calls.
25143 (ix86_output_indirect_branch_via_reg): Replace
25144 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
25145 (ix86_output_indirect_branch_via_push): Likewise.
25146 (ix86_output_function_return): Likewise.
25147 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
25148 incompatible with -fcf-protection=branch and
25149 -fcheck-pointer-bounds.
25151 2018-02-22 Steve Ellcey <sellcey@cavium.com>
25154 * config/aarch64/aarch64.c (aarch64_print_address_internal):
25155 Change gcc_assert call to output_operand_lossage.
25157 2018-02-22 Steve Ellcey <sellcey@cavium.com>
25159 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
25161 2018-02-22 DJ Delorie <dj@redhat.com>
25162 Sebastian Perta <sebastian.perta@renesas.com>
25163 Oleg Endo <olegendo@gcc.gnu.org>
25165 * config/rx/rx.c (rx_rtx_costs): New function.
25166 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
25168 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
25170 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
25172 2018-02-22 Martin Liska <mliska@suse.cz>
25175 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
25176 Add "native" as a possible value.
25178 2018-02-22 Martin Liska <mliska@suse.cz>
25181 * config/i386/i386.c (ix86_option_override_internal):
25182 Add "native" as a possible value for -march and -mtune.
25184 2018-02-22 Jakub Jelinek <jakub@redhat.com>
25187 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
25188 to all type variants.
25190 PR tree-optimization/84503
25191 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
25192 width as info->bitpos + info->bitsize - start.
25193 (merged_store_group::merge_overlapping): Simplify width computation.
25194 (check_no_overlap): New function.
25195 (imm_store_chain_info::try_coalesce_bswap): Compute expected
25196 start + width and last_order of the group, fail if check_no_overlap
25198 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
25199 to group if check_no_overlap fails.
25201 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
25203 * config/rs6000/altivec.md: Delete contraint arguments to
25204 define_expand, define_split, and define_peephole2, and in
25205 define_insn_and_split if always unused.
25206 * config/rs6000/darwin.md: Ditto.
25207 * config/rs6000/dfp.md: Ditto.
25208 * config/rs6000/rs6000.md: Ditto.
25209 * config/rs6000/sync.md: Ditto.
25210 * config/rs6000/vector.md: Ditto.
25211 * config/rs6000/vsx.md: Ditto.
25213 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
25215 * config/rs6000/altivec.md: Write output control strings as braced
25216 blocks instead of double-quoted strings.
25217 * config/rs6000/darwin.md: Ditto.
25218 * config/rs6000/rs6000.md: Ditto.
25219 * config/rs6000/vector.md: Ditto.
25220 * config/rs6000/vsx.md: Ditto.
25222 2018-02-21 Jason Merrill <jason@redhat.com>
25224 PR c++/84314 - ICE with templates and fastcall attribute.
25225 * attribs.c (build_type_attribute_qual_variant): Remove assert.
25227 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
25229 * ipa-cp.c (determine_versionability): Fix comment typos.
25231 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
25234 * ipa-cp.c (determine_versionability): Do not version functions caling
25237 2018-02-21 Martin Liska <mliska@suse.cz>
25240 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
25241 Add "native" as a possible value.
25242 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
25243 the macro when native cpu detection is available.
25245 2018-02-21 Martin Liska <mliska@suse.cz>
25248 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
25249 Add "native" as a possible value.
25250 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
25251 when native cpu detection is available.
25253 2018-02-21 Jakub Jelinek <jakub@redhat.com>
25254 Martin Sebor <msebor@redhat.com>
25256 PR tree-optimization/84478
25257 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
25259 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
25260 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
25261 support which is conservatively correct, for 2 only stay conservative
25262 for maxlen. Formatting and comment capitalization fixes. Add STRICT
25263 argument to the 2 argument get_range_strlen, adjust 6 arg
25264 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
25266 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
25267 (gimple_fold_builtin_strlen): Pass true as last argument to
25270 2018-02-20 Martin Sebor <msebor@redhat.com>
25272 PR middle-end/84095
25273 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
25274 (builtin_memref::set_base_and_offset): Same. Handle inner references.
25275 (builtin_memref::builtin_memref): Factor out parts into
25276 set_base_and_offset and call it.
25278 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
25280 PR middle-end/84406
25281 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
25282 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
25283 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
25284 search at the associated MODE_INT.
25286 2018-02-20 Jeff Law <law@redhat.com>
25288 PR middle-end/82123
25289 PR tree-optimization/81592
25290 PR middle-end/79257
25291 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
25292 for range data rather than using global data.
25293 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
25294 range data rather than using global data.
25295 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
25296 pass it to children as needed.
25297 (struct directive::fmtresult): Similarly.
25298 (struct directive::set_width): Similarly.
25299 (struct directive::set_precision): Similarly.
25300 (format_integer, format_directive, parse_directive): Similarly.
25301 (format_none): Accept unnamed vr_values parameter.
25302 (format_percent, format_floating, format_character): Similarly.
25303 (format_string, format_plain): Similarly.
25304 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
25305 the EVRP range analyzer for range data rather than using global data.
25306 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
25307 gimple-ssa-evrp-analyze.h
25308 (class sprintf_dom_walker): Add after_dom_children member function.
25309 Add evrp_range_analyzer member.
25310 (sprintf_dom_walker::before_dom_children): Call into the EVRP
25311 range analyzer as needed.
25312 (sprintf_dom_walker::after_dom_children): New member function.
25313 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
25315 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
25316 (evrp_range_analyzer::pop_to_marker): Likewise.
25318 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
25320 PR tree-optimization/84419
25321 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
25322 with the required type if its current type is compatible but
25325 2018-02-20 Jakub Jelinek <jakub@redhat.com>
25327 PR middle-end/82004
25328 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
25329 after vectorization.
25331 2018-02-20 Martin Liska <mliska@suse.cz>
25334 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
25335 possible values if we don't have a hint.
25337 2018-02-20 Martin Liska <mliska@suse.cz>
25341 * final.c (shorten_branches): Build align_tab array with one
25343 * opts.c (finish_options): Add alignment option limit check.
25344 (MAX_CODE_ALIGN): Likewise.
25345 (MAX_CODE_ALIGN_VALUE): Likewise.
25346 * doc/invoke.texi: Document maximum allowed option value for
25347 all -falign-* options.
25349 2018-02-19 Jakub Jelinek <jakub@redhat.com>
25352 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
25353 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
25354 * var-tracking.c (emit_note_insn_var_location): Remove all references
25355 to NOTE_INSN_CALL_ARG_LOCATION.
25356 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
25357 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
25358 Use copy_rtx_if_shared.
25359 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
25360 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
25361 (dwarf2out_var_location): Remove handling of
25362 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
25364 * final.c (final_scan_insn): Remove all references to
25365 NOTE_INSN_CALL_ARG_LOCATION.
25366 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
25367 before dumping final insns.
25368 * except.c (emit_note_eh_region_end): Remove all references to
25369 NOTE_INSN_CALL_ARG_LOCATION.
25370 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
25371 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
25372 * config/arc/arc.c (hwloop_optimize): Likewise.
25373 * config/arm/arm.c (create_fix_barrier): Likewise.
25374 * config/s390/s390.c (s390_chunkify_start): Likewise.
25375 * config/sh/sh.c (find_barrier): Likewise.
25376 * config/i386/i386.c (rest_of_insert_endbranch,
25377 ix86_seh_fixup_eh_fallthru): Likewise.
25378 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
25379 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
25380 * config/frv/frv.c (frv_function_prologue): Likewise.
25381 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
25383 (note_outside_basic_block_p): Remove all references to
25384 NOTE_INSN_CALL_ARG_LOCATION.
25385 * gengtype.c (adjust_field_rtx_def): Likewise.
25386 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
25388 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
25389 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
25392 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
25395 PR tree-optimization/84452
25396 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
25397 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
25400 2018-02-19 Martin Liska <mliska@suse.cz>
25403 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
25405 2018-02-19 Martin Liska <mliska@suse.cz>
25406 Richard Sandiford <richard.sandiford@linaro.org>
25408 PR tree-optimization/82491
25409 * gimple-fold.c (get_base_constructor): Make earlier bail out
25412 2018-02-19 Carl Love <cel@us.ibm.com>
25414 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
25415 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
25417 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
25418 P8V_BUILTIN_VEC_NEG.
25420 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
25422 * config/rl78/rl78.md (movdf): New define expand.
25424 2018-02-19 Martin Liska <mliska@suse.cz>
25427 * doc/invoke.texi: Fix typo.
25428 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
25430 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
25432 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
25433 handle rs6000_single_float and rs6000_double_float specially for
25436 2018-02-16 Jeff Law <law@redhat.com>
25438 * config/rx/rx.c (add_pop_cfi_notes): New function.;
25439 (pop_regs): Use it.
25441 2018-02-16 Jakub Jelinek <jakub@redhat.com>
25444 * ipa-inline.c (inline_small_functions): Fix a typo.
25446 2018-02-16 Nathan Sidwell <nathan@acm.org>
25448 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
25450 2018-02-16 Carl Love <cel@us.ibm.com>
25452 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
25453 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
25454 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
25455 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
25456 expansion to P8V_BUILTIN_VEC_FLOAT2.
25458 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
25460 PR rtl-optimization/70023
25461 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
25462 src_regno into account.
25464 2018-02-16 Carl Love <cel@us.ibm.com>
25466 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
25467 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
25468 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
25469 * config/rs6000/rs6000.c: Remove case statements for
25470 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
25471 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
25472 and P9V_BUILTIN_VEC_VINSERT4B.
25473 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
25474 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
25475 * config/rs6000/vsx.md:
25476 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
25479 2018-02-16 Carl Love <cel@us.ibm.com>
25481 * config/rs6000/altivec.h: Add builtin names vec_extract4b
25483 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
25485 * config/rs6000/rs6000-c.c: Add the definitions for
25486 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
25487 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
25488 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
25489 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
25490 definition for insert4b and define insn *insert3b_internal.
25491 * doc/extend.texi: Add documentation for vec_extract4b.
25493 2018-02-16 Nathan Sidwell <nathan@acm.org>
25495 * doc/extend.texi (Backwards Compatibility): Mention friend
25496 injection. Note for-scope is deprecated.
25497 * doc/invoke.texi (-ffriend-injection): Deprecate.
25499 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
25501 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
25502 that moved to I2, also allow destinations that are a paradoxical
25503 subreg (instead of a normal reg).
25505 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
25508 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
25511 2018-02-16 Richard Biener <rguenther@suse.de>
25513 PR tree-optimization/84037
25514 PR tree-optimization/84016
25516 * config/i386/i386.c (ix86_builtin_vectorization_cost):
25517 Adjust vec_construct for the fact we need additional higher latency
25518 128bit inserts for AVX256 and AVX512 vector builds.
25519 (ix86_add_stmt_cost): Scale vector construction cost for
25522 2018-02-16 Richard Biener <rguenther@suse.de>
25524 PR tree-optimization/84417
25525 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
25526 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
25527 (non_rewritable_lvalue_p): Likewise, use poly-ints.
25529 2018-02-16 Martin Liska <mliska@suse.cz>
25532 * internal-fn.def (ASAN_CHECK): Set proper flags.
25533 (ASAN_MARK): Likewise.
25535 2018-02-16 Julia Koval <julia.koval@intel.com>
25537 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
25538 from PTA_CANNONLAKE.
25540 2018-02-16 Jakub Jelinek <jakub@redhat.com>
25543 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
25544 Use ++iter rather than iter++ for std::list iterators.
25545 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
25546 defer deleting them until all nodes in the forest are processed. Do
25547 free even leaf nodes. Change to_process into auto_vec.
25550 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
25551 * vec.h (vec_default_construct): Use memset instead of placement new
25552 if BROKEN_VALUE_INITIALIZATION is defined.
25553 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
25554 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
25557 PR rtl-optimization/83723
25558 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
25559 * lra.c (lra_substitute_pseudo): Likewise. If true, use
25560 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
25562 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
25564 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
25566 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
25568 PR rtl-optimization/81443
25569 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
25570 from inner REGs to paradoxical SUBREGs.
25572 2018-02-16 Richard Biener <rguenther@suse.de>
25574 PR tree-optimization/84399
25575 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
25576 For operands we can analyze at their definition make sure we can
25577 analyze them at each use as well.
25579 2018-02-16 Richard Biener <rguenther@suse.de>
25581 PR tree-optimization/84190
25582 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
25583 volatile accesses if the decl isn't volatile.
25585 2018-02-15 Jason Merrill <jason@redhat.com>
25587 PR c++/84314 - ICE with templates and fastcall attribute.
25588 * attribs.c (build_type_attribute_qual_variant): Don't clobber
25589 TYPE_CANONICAL on an existing type.
25591 2018-02-15 Jakub Jelinek <jakub@redhat.com>
25593 PR tree-optimization/84383
25594 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
25595 dstoff nor call operand_equal_p if dstbase is NULL.
25597 PR tree-optimization/84334
25598 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
25599 also a CONSTANT_CLASS_P, punt.
25601 2018-02-14 Jim Wilson <jimw@sifive.com>
25603 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
25604 first SMALL_OPERAND check. New local min_second_step. Move assert
25605 to where locals are set. Add TARGET_RVC support.
25606 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
25608 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
25610 * doc/invoke.texi: Correct -Wformat-overflow code sample.
25612 2018-02-14 Martin Sebor <msebor@redhat.com>
25614 PR tree-optimization/83698
25615 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
25616 arrays constrain the offset range to their bounds.
25617 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
25618 (builtin_access::overlap): Avoid setting the size of overlap if it's
25620 (maybe_diag_overlap): Also consider arrays when deciding what values
25621 of offsets to include in diagnostics.
25623 2018-02-14 Martin Sebor <msebor@redhat.com>
25626 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
25627 that correspond to the kind of a declaration.
25629 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
25632 * config/pa/pa.md: Load address of PIC label using the linkage table
25633 if the label is nonlocal.
25635 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
25637 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
25638 warning message if user requests -maltivec=be.
25639 * doc/invoke.texi: Document deprecation of -maltivec=be.
25641 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
25644 * config/rs6000/rs6000-c.c: Update definitions for
25645 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
25646 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
25648 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
25651 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
25652 add _get_ssp intrinsics. Remove argument from
25653 __builtin_ia32_rdssp[d|q].
25654 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
25655 * config/i386/i386-builtin.def: Remove argument from
25656 __builtin_ia32_rdssp[d|q].
25657 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
25658 ix86_expand_special_args_builtin for _rdssp[d|q].
25659 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
25660 Clear register before usage.
25661 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
25662 Add documentation for new _get_ssp and _inc_ssp intrinsics.
25664 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
25666 PR tree-optimization/84357
25667 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
25668 operand 1 of an ARRAY_REF too.
25670 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
25673 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
25674 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25676 (set_of_reg): New struct.
25677 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
25678 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
25679 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25681 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
25682 Split into bitclr, bitset, bitinvert patterns if appropriate.
25683 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
25684 use rx_fuse_in_memory_bitop.
25685 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
25686 to named insn, correct maximum insn length.
25688 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
25691 * machmode.def: Define a complex mode for PARTIAL_INT.
25692 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
25694 * doc/rtl.texi: Document CSPImode.
25695 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
25697 (msp430_hard_regno_nregs_with_padding): Likewise.
25699 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
25702 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
25704 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
25706 PR rtl-optimization/84169
25707 * combine.c (try_combine): New variable split_i2i3. Set it to true if
25708 we generated a parallel as new i3 and we split that to new i2 and i3
25709 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
25710 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
25711 those to i2, not i1. Partially rewrite this scan code.
25713 2018-02-13 Jakub Jelinek <jakub@redhat.com>
25716 * stor-layout.c (place_field): For variable length fields, adjust
25717 offset_align afterwards not just based on the field's alignment,
25718 but also on the size.
25720 PR middle-end/84309
25721 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
25722 of exps and logs in the use_exp2 case.
25724 2018-02-13 Jeff Law <law@redhat.com>
25726 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
25727 entry for "vector".
25729 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
25732 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
25736 * common.opt (gas-loc-support, gas-locview-support): New.
25737 (ginline-points, ginternal-reset-location-views): New.
25738 * doc/invoke.texi: Document them. Use @itemx where intended.
25739 (gvariable-location-views): Adjust.
25740 * target.def (reset_location_view): New.
25741 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
25742 (TARGET_RESET_LOCATION_VIEW): New.
25743 * doc/tm.texi: Rebuilt.
25744 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
25745 (dwarf2out_default_as_locview_support): New.
25746 (output_asm_line_debug_info): Use option variables.
25747 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
25748 (output_loc_list): Likewise.
25749 (add_high_low_attributes): Check option variables.
25750 Don't output entry view attribute in strict mode.
25751 (gen_inlined_subroutine_die): Check option variables.
25752 (dwarf2out_inline_entry): Likewise.
25753 (init_sections_and_labels): Likewise.
25754 (dwarf2out_early_finish): Likewise.
25755 (maybe_reset_location_view): New, from...
25756 (dwarf2out_var_location): ... here. Call it.
25757 * debug.h (dwarf2out_default_as_loc_support): Declare.
25758 (dwarf2out_default_as_locview_support): Declare.
25759 * hooks.c (hook_int_rtx_insn_0): New.
25760 * hooks.h (hook_int_rtx_insn_0): Declare.
25761 * toplev.c (process_options): Take -gas-loc-support and
25762 -gas-locview-support from dwarf2out. Enable
25763 -gvariable-location-views by default only with locview
25764 assembler support. Enable -ginternal-reset-location-views by
25765 default only if the target defines the corresponding hook.
25766 Enable -ginline-points by default if location views are
25767 enabled; force it disabled if statement frontiers are
25769 * tree-inline.c (expand_call_inline): Check option variables.
25770 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
25772 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
25774 PR tree-optimization/84321
25775 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
25776 handling. Also check whether the anti-range contains any values
25777 that satisfy the mask; switch to a VR_RANGE if not.
25779 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
25782 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
25784 2018-02-13 Martin Jambor <mjambor@suse.cz>
25787 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
25788 of call statements, also set location of a load to a temporary.
25790 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
25792 * config/rl78/rl78.c (add_vector_labels): New function.
25793 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
25794 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
25795 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
25796 which checks that no arguments are passed.
25797 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
25798 * doc/extend.texi: Documentation for the new attribute.
25800 2018-02-13 Andreas Schwab <schwab@suse.de>
25802 * config/riscv/linux.h (CPP_SPEC): Define.
25804 2018-02-13 Jakub Jelinek <jakub@redhat.com>
25807 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
25808 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
25809 OPTION_MASK_ISA_AES as first argument to def_builtin_const
25810 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
25811 instead of OPTION_MASK_ISA_PCLMUL as first argument to
25812 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
25813 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
25814 temporarily for AES and PCLMUL builtins.
25816 PR tree-optimization/84339
25817 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
25818 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
25821 PR middle-end/84309
25822 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
25823 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
25824 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
25826 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
25828 * omp-simd-clone.h: New file.
25829 * omp-simd-clone.c: Include omp-simd-clone.h.
25830 (expand_simd_clones): No longer static.
25831 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
25832 cgraph.h and omp-simd-clone.h.
25833 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
25834 (vect_recog_widen_shift_pattern): Formatting fix.
25835 (vect_pattern_recog_1): Don't check optab for calls.
25838 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
25839 operands[2] into a REG before using gen_lowpart on it.
25841 2018-02-12 Jeff Law <law@redhat.com>
25844 * config/sh/sh.c (find_barrier): Consider a sibling call
25847 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
25848 successfully back substituting a reg.
25850 2018-02-12 Richard Biener <rguenther@suse.de>
25852 PR tree-optimization/84037
25853 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
25854 parameter, move visited init to caller.
25855 (vect_slp_analyze_operations): Separate cost from validity
25856 check, initialize visited once for all instances.
25857 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
25859 * tree-vect-stmts.c (vect_model_simple_cost): Make early
25861 (vect_model_promotion_demotion_cost): Likewise.
25862 (vectorizable_bswap): Guard cost modeling with !slp_node
25863 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
25865 (vectorizable_call): Likewise.
25866 (vectorizable_conversion): Likewise.
25867 (vectorizable_assignment): Likewise.
25868 (vectorizable_shift): Likewise.
25869 (vectorizable_operation): Likewise.
25870 (vectorizable_store): Likewise.
25871 (vectorizable_load): Likewise.
25872 (vectorizable_condition): Likewise.
25873 (vectorizable_comparison): Likewise.
25875 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
25878 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
25879 (ASAN_MARK): Fix fnspec to account for return value, change pointer
25880 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
25882 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
25884 PR middle-end/83665
25885 * params.def (inline-min-speedup): Increase from 8 to 15.
25886 (max-inline-insns-auto): Decrease from 40 to 30.
25887 * ipa-split.c (consider_split): Add some buffer for function to
25888 be considered inlining candidate.
25889 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
25892 2018-02-12 Richard Biener <rguenther@suse.de>
25894 PR tree-optimization/84037
25895 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
25896 matched stmts if we cannot swap the non-matched ones.
25898 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
25900 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
25901 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
25902 _mm_maskz_scalef_round_ss): New intrinsics.
25903 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
25904 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
25905 __builtin_ia32_scalefss_round): Remove.
25906 (__builtin_ia32_scalefsd_mask_round,
25907 __builtin_ia32_scalefss_mask_round): New intrinsics.
25908 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
25909 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
25910 ((match_operand:VF_128 2 "<round_nimm_predicate>"
25911 "<round_constraint>")): Changed to ...
25912 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
25913 "<round_scalar_constraint>")): ... this.
25914 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
25915 %0, %1, %2<round_op3>}"): Changed to ...
25916 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
25917 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
25918 %2<round_scalar_mask_op3>}"): ... this.
25919 * config/i386/subst.md (round_scalar_nimm_predicate): New.
25921 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
25923 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
25924 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
25925 (_mm_maskz_sqrt_round_ss): New intrinsics.
25926 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
25927 (__builtin_ia32_sqrtsd_mask_round)
25928 (__builtin_ia32_sqrtss_mask_round): New builtins.
25929 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
25930 (__builtin_ia32_sqrtss_round): Remove.
25931 (__builtin_ia32_sqrtsd_mask_round)
25932 (__builtin_ia32_sqrtss_mask_round): New builtins.
25933 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
25934 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
25935 ((match_operand:VF_128 1 "vector_operand"
25936 "xBm,<round_constraint>")): Changed to ...
25937 ((match_operand:VF_128 1 "vector_operand"
25938 "xBm,<round_scalar_constraint>")): ... this.
25939 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
25940 %0, %2, %<iptr>1<round_op3>}): Changed to ...
25941 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
25942 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
25943 %<iptr>1<round_scalar_mask_op3>}): ... this.
25944 ((set_attr "prefix" "<round_prefix>")): Changed to ...
25945 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
25947 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
25950 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
25951 Cast vec_cmpeq result to correct type.
25952 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
25953 Cast vec_cmpgt result to correct type.
25955 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
25957 * final.c (final_scan_insn_1): Renamed from...
25958 (final_scan_insn): ... this. New wrapper, to recover
25959 seen from the outermost call in recursive ones.
25960 * config/sparc/sparc.c (output_return): Drop seen from call.
25961 (output_sibcall): Likewise.
25962 * config/visium/visium.c (output_branch): Likewise.
25964 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
25966 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
25969 2018-02-10 Alan Modra <amodra@gmail.com>
25972 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
25973 Specify LR as an input.
25975 2018-02-10 Jakub Jelinek <jakub@redhat.com>
25978 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
25979 remove_member_access_dummy_vars): New functions.
25980 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
25981 lower_omp_1, execute_lower_omp): Use them.
25983 PR rtl-optimization/84308
25984 * shrink-wrap.c (spread_components): Release todo vector.
25986 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
25988 PR rtl-optimization/57193
25989 * ira-color.c (struct allocno_color_data): Add member
25990 conflict_allocno_hard_prefs.
25991 (update_conflict_allocno_hard_prefs): New.
25992 (bucket_allocno_compare_func): Add a preference based on
25993 conflict_allocno_hard_prefs.
25994 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
25995 (color_allocnos): Remove a dead code. Initiate
25996 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
25998 2018-02-09 Jakub Jelinek <jakub@redhat.com>
26001 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
26002 constraint from =wa to wa. Avoid a subreg on the output operand,
26003 instead use a pseudo and subreg it in a move.
26004 (p9_xxbrd_<mode>): Changed to ...
26005 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
26006 (p9_xxbrd_v2df): New expander.
26007 (p9_xxbrw_<mode>): Changed to ...
26008 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
26009 (p9_xxbrw_v4sf): New expander.
26011 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
26013 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
26015 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
26018 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
26019 multiply in 32-bit mode.
26020 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
26021 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
26024 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
26026 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
26027 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
26028 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
26029 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
26031 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
26034 * dwarf2out.c (is_trivial_indirect_ref): New function.
26035 (dwarf2out_late_global_decl): Do not generate a location
26036 attribute for variables that have a non-trivial DECL_VALUE_EXPR
26037 and that are not defined in the current unit.
26039 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
26041 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
26042 instead of a libcall for UNORDERED.
26044 2018-02-09 Tamar Christina <tamar.christina@arm.com>
26047 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
26048 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
26050 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
26053 * config/s390/s390.c (s390_set_current_function): Invoke
26054 s390_indirect_branch_settings also if fndecl didn't change.
26056 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
26058 * config/rs6000/rs6000.md (blockage): Set length to zero.
26060 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
26062 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
26064 2018-02-09 Jakub Jelinek <jakub@redhat.com>
26067 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
26068 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
26072 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
26073 PARALLEL incoming that failed vt_get_decl_and_offset check.
26075 PR middle-end/84237
26076 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
26077 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
26079 (get_variable_section): For decls in named .bss* sections pass true as
26080 second argument to bss_initializer_p.
26082 2018-02-09 Marek Polacek <polacek@redhat.com>
26083 Jakub Jelinek <jakub@redhat.com>
26086 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
26087 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
26088 Sync some changes from cxx_fold_indirect_ref.
26090 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
26092 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
26094 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
26095 (BLOCK_INLINE_ENTRY_LABEL): New.
26096 (dwarf2out_var_location): Disregard inline entry markers.
26097 (inline_entry_data): New struct.
26098 (inline_entry_data_hasher): New hashtable type.
26099 (inline_entry_data_hasher::hash): New.
26100 (inline_entry_data_hasher::equal): New.
26101 (inline_entry_data_table): New variable.
26102 (add_high_low_attributes): Add DW_AT_entry_pc and
26103 DW_AT_GNU_entry_view attributes if a pending entry is found
26104 in inline_entry_data_table. Add old entry_pc attribute only
26105 if debug nonbinding markers are disabled.
26106 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
26107 markers are enabled.
26108 (block_within_block_p, dwarf2out_inline_entry): New.
26109 (dwarf2out_finish): Check that no entries remained in
26110 inline_entry_data_table.
26111 * final.c (reemit_insn_block_notes): Handle inline entry notes.
26112 (final_scan_insn, notice_source_line): Likewise.
26113 (rest_of_clean_state): Skip inline entry markers.
26114 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
26116 * gimple.c (gimple_build_debug_inline_entry): New.
26117 * gimple.h (enum gimple_debug_subcode): Add
26118 GIMPLE_DEBUG_INLINE_ENTRY.
26119 (gimple_build_debug_inline_entry): Declare.
26120 (gimple_debug_inline_entry_p): New.
26121 (gimple_debug_nonbind_marker_p): Adjust.
26122 * insn-notes.def (INLINE_ENTRY): New.
26123 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
26124 inline entry marker notes.
26125 (print_insn): Likewise.
26126 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
26127 (INSN_DEBUG_MARKER_KIND): Likewise.
26128 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
26129 * tree-inline.c (expand_call_inline): Build and insert
26130 debug_inline_entry stmt.
26131 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
26132 inline entry blocks early, if nonbind markers are enabled.
26133 (dump_scope_block): Dump fragment info.
26134 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
26135 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
26136 (gimple_build_debug_inline_entry): New.
26137 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
26138 Enable/disable inline entry points too.
26139 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
26140 (DEBUG_INSN): Describe inline entry markers.
26142 * common.opt (gvariable-location-views): New.
26143 (gvariable-location-views=incompat5): New.
26144 * config.in: Rebuilt.
26145 * configure: Rebuilt.
26146 * configure.ac: Test assembler for view support.
26147 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
26148 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
26149 * dwarf2out.c (var_loc_view): New typedef.
26150 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
26151 (dwarf2out_locviews_in_attribute): New.
26152 (dwarf2out_locviews_in_loclist): New.
26153 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
26154 (enum dw_line_info_opcode): Add LI_adv_address.
26155 (struct dw_line_info_table): Add view.
26156 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
26157 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
26158 (zero_view_p): New variable.
26159 (ZERO_VIEW_P): New macro.
26160 (output_asm_line_debug_info): New.
26161 (struct var_loc_node): Add view.
26162 (add_AT_view_list, AT_loc_list): New.
26163 (add_var_loc_to_decl): Add view param. Test it against last.
26164 (new_loc_list): Add view params. Record them.
26165 (AT_loc_list_ptr): Handle loc and view lists.
26166 (view_list_to_loc_list_val_node): New.
26167 (print_dw_val): Handle dw_val_class_view_list.
26168 (size_of_die): Likewise.
26169 (value_format): Likewise.
26170 (loc_list_has_views): New.
26171 (gen_llsym): Set vl_symbol too.
26172 (maybe_gen_llsym, skip_loc_list_entry): New.
26173 (dwarf2out_maybe_output_loclist_view_pair): New.
26174 (output_loc_list): Output view list or entries too.
26175 (output_view_list_offset): New.
26176 (output_die): Handle dw_val_class_view_list.
26177 (output_dwarf_version): New.
26178 (output_compilation_unit_header): Use it.
26179 (output_skeleton_debug_sections): Likewise.
26180 (output_rnglists, output_line_info): Likewise.
26181 (output_pubnames, output_aranges): Update version comments.
26182 (output_one_line_info_table): Output view numbers in asm comments.
26183 (dw_loc_list): Determine current endview, pass it to new_loc_list.
26184 Call maybe_gen_llsym.
26185 (loc_list_from_tree_1): Adjust.
26186 (add_AT_location_description): Create view list attribute if
26187 needed, check it's absent otherwise.
26188 (convert_cfa_to_fb_loc_list): Adjust.
26189 (maybe_emit_file): Call output_asm_line_debug_info for test.
26190 (dwarf2out_var_location): Reset views as needed. Precompute
26191 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
26192 attribute. Set view.
26193 (new_line_info_table): Reset next view.
26194 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
26195 (dwarf2out_source_line): Likewise. Output view resets and labels to
26196 the assembler, or select appropriate line info opcodes.
26197 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
26198 (optimize_string_length): Catch it. Adjust.
26199 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
26200 dw_val_class_view_list, and remove it if no longer needed.
26201 (hash_loc_list): Hash view numbers.
26202 (loc_list_hasher::equal): Compare them.
26203 (optimize_location_lists): Check whether a view list symbol is
26204 needed, and whether the locview attribute is present, and
26205 whether they match. Remove the locview attribute if no longer
26207 (index_location_lists): Call skip_loc_list_entry for test.
26208 (dwarf2out_finish): Call output_asm_line_debug_info for test.
26209 Use output_dwarf_version.
26210 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
26211 (struct dw_val_node): Add val_view_list.
26212 * final.c (SEEN_NEXT_VIEW): New.
26213 (set_next_view_needed): New.
26214 (clear_next_view_needed): New.
26215 (maybe_output_next_view): New.
26216 (final_start_function): Rename to...
26217 (final_start_function_1): ... this. Take pointer to FIRST,
26218 add SEEN parameter. Emit param bindings in the initial view.
26219 (final_start_function): Reintroduce SEEN-less interface.
26220 (final): Rename to...
26221 (final_1): ... this. Take SEEN parameter. Output final pending
26222 next view at the end.
26223 (final): Reintroduce seen-less interface.
26224 (final_scan_insn): Output pending next view before switching
26225 sections or ending a block. Mark the next view as needed when
26226 outputting variable locations. Notify debug backend of section
26227 changes, and of location view changes.
26228 (rest_of_handle_final): Adjust.
26229 * toplev.c (process_options): Autodetect value for debug variable
26230 location views option. Warn on incompat5 without -gdwarf-5.
26231 * doc/invoke.texi (gvariable-location-views): New.
26232 (gvariable-location-views=incompat5): New.
26233 (gno-variable-location-views): New.
26235 2018-02-08 David Malcolm <dmalcolm@redhat.com>
26237 PR tree-optimization/84136
26238 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
26239 that the result of find_edge is non-NULL.
26241 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
26244 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
26245 storing integer register in SImode. Fix cost of 256 and 512
26246 byte aligned SSE register store.
26248 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
26250 * config/i386/i386.c (ix86_multiplication_cost): Fix
26251 multiplication cost for TARGET_AVX512DQ.
26253 2018-02-08 Marek Polacek <polacek@redhat.com>
26255 PR tree-optimization/84238
26256 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
26259 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
26261 PR tree-optimization/84265
26262 * tree-vect-stmts.c (vectorizable_store): Don't treat
26263 VMAT_CONTIGUOUS accesses as grouped.
26264 (vectorizable_load): Likewise.
26266 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
26268 PR tree-optimization/81635
26269 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
26270 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
26271 (test_round_for_mask): New functions.
26272 (wide_int_cc_tests): Call test_round_for_mask.
26273 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
26274 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
26275 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
26276 range returned by get_range_info.
26278 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
26281 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
26282 * symtab.c: Include builtins.h
26283 (symtab_node::output_to_lto_symbol_table_p): Move here
26284 from lto-streamer-out.c:output_symbol_p.
26285 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
26286 (output_symbol_p): Move all logic to symtab.c
26287 (produce_symtab): Update.
26289 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
26291 * config/s390/s390-opts.h (enum indirect_branch): Define.
26292 * config/s390/s390-protos.h (s390_return_addr_from_memory)
26293 (s390_indirect_branch_via_thunk)
26294 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
26295 (enum s390_indirect_branch_type): Define.
26296 * config/s390/s390.c (struct s390_frame_layout, struct
26297 machine_function): Remove.
26298 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
26299 (indirect_branch_table_label_no, indirect_branch_table_name):
26301 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
26302 (enum s390_indirect_branch_option): Define.
26303 (s390_return_addr_from_memory): New function.
26304 (s390_handle_string_attribute): New function.
26305 (s390_attribute_table): Add new attribute handler.
26306 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
26307 (s390_indirect_branch_via_thunk): New function.
26308 (s390_indirect_branch_via_inline_thunk): New function.
26309 (s390_function_ok_for_sibcall): When jumping via thunk disallow
26310 sibling call optimization for non z10 compiles.
26311 (s390_emit_call): Force indirect branch target to be a single
26312 register. Add r1 clobber for non-z10 compiles.
26313 (s390_emit_epilogue): Emit return jump via return_use expander.
26314 (s390_reorg): Handle JUMP_INSNs as execute targets.
26315 (s390_option_override_internal): Perform validity checks for the
26316 new command line options.
26317 (s390_indirect_branch_attrvalue): New function.
26318 (s390_indirect_branch_settings): New function.
26319 (s390_set_current_function): Invoke s390_indirect_branch_settings.
26320 (s390_output_indirect_thunk_function): New function.
26321 (s390_code_end): Implement target hook.
26322 (s390_case_values_threshold): Implement target hook.
26323 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
26325 * config/s390/s390.h (struct s390_frame_layout)
26326 (struct machine_function): Move here from s390.c.
26327 (TARGET_INDIRECT_BRANCH_NOBP_RET)
26328 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
26329 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
26330 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
26331 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
26332 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
26333 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
26334 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
26335 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
26336 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
26337 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
26338 (mnemonic attribute): Add values which aren't recognized
26340 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
26341 pattern for branch conversion. Fix mnemonic attribute.
26342 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
26343 indirect branch via thunk if requested.
26344 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
26345 ("*indirect_jump"): Disable for branch conversion using out of
26347 ("indirect_jump_via_thunk<mode>_z10")
26348 ("indirect_jump_via_thunk<mode>")
26349 ("indirect_jump_via_inlinethunk<mode>_z10")
26350 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
26351 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
26352 ("casesi_jump_via_inlinethunk<mode>_z10")
26353 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
26354 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
26355 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
26356 ("*indirect2_jump"): Disable for branch conversion.
26357 ("casesi_jump"): Turn into expander and expand patterns for branch
26359 ("return_use"): New expander.
26360 ("*return"): Emit return via thunk and rename it to ...
26361 ("*return<mode>"): ... this one.
26362 * config/s390/s390.opt: Add new options and and enum for the
26365 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
26367 * lra-constraints.c (match_reload): Unconditionally use
26368 gen_lowpart_SUBREG, rather than selecting between that
26369 and equivalent gen_rtx_SUBREG code.
26371 2018-02-08 Richard Biener <rguenther@suse.de>
26373 PR tree-optimization/84233
26374 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
26375 changed flag instead of boguously re-using phi_inserted.
26377 2018-02-08 Martin Jambor <mjambor@suse.cz>
26379 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
26380 static local variables.
26382 2018-02-08 Richard Biener <rguenther@suse.de>
26384 PR tree-optimization/84278
26385 * tree-vect-stmts.c (vectorizable_store): When looking for
26386 smaller vector types to perform grouped strided loads/stores
26387 make sure the mode is supported by the target.
26388 (vectorizable_load): Likewise.
26390 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
26392 * config/aarch64/aarch64.c (aarch64_components_for_bb):
26393 Increase LDP/STP opportunities by adding adjacent callee-saves.
26395 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
26397 PR rtl-optimization/84068
26398 PR rtl-optimization/83459
26399 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
26401 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
26403 PR tree-optimization/84224
26404 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
26405 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
26406 non-zero arguments.
26408 2018-02-07 Iain Sandoe <iain@codesourcery.com>
26411 * config/rs6000/altivec.md (*restore_world): Remove LR use.
26412 * config/rs6000/predicates.md (restore_world_operation): Adjust op
26413 count, remove one USE.
26415 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
26417 * doc/install.texi (Configuration): Document the
26418 --with-long-double-format={ibm,ieee} PowerPC configuration
26422 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
26423 Convert from define_expand to be define_insn_and_split. Rework
26424 float/double/_Float128 conversions to QI/HI/SImode to work with
26425 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
26426 conversions to QI/HImode types did a store and then a load to
26427 truncate the value. For conversions to VSX registers, don't split
26428 the insn, instead emit the code directly. Use the code iterator
26429 any_fix to combine signed and unsigned conversions.
26430 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
26431 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
26432 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
26433 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
26434 (fix_<mode>di2_hw): Likewise.
26435 (fixuns_<mode>di2_hw): Likewise.
26436 (fix_<mode>si2_hw): Likewise.
26437 (fixuns_<mode>si2_hw): Likewise.
26438 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
26439 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
26440 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
26441 fix<uns>_trunc<SFDF:mode>si2_p8.
26442 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
26443 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
26444 (fix<uns>_<mode>_mem): Likewise.
26445 (fctiw<u>z_<mode>_mem): Likewise.
26446 (fix<uns>_<mode>_mem): Likewise.
26447 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
26448 the register allocator from doing a direct move to the GPRs to do
26449 a store, and instead use the ISA 3.0 store byte/half-word from
26450 vector register instruction. For IEEE 128-bit floating point,
26451 also optimize stores of 32-bit ints.
26452 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
26454 2018-02-07 Alan Hayward <alan.hayward@arm.com>
26456 * genextract.c (push_pathstr_operand): New function to support
26458 (walk_rtx): Call push_pathstr_operand.
26459 (print_path): Support [a-zA-Z].
26461 2018-02-07 Richard Biener <rguenther@suse.de>
26463 PR tree-optimization/84037
26464 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
26465 (cse_and_gimplify_to_preheader): Declare.
26466 (vect_get_place_in_interleaving_chain): Likewise.
26467 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26469 (_loop_vec_info::~_loop_vec_info): Delete it.
26470 (cse_and_gimplify_to_preheader): New function.
26471 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
26472 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
26473 (vectorizable_load): Likewise. For grouped stores always base
26474 the IV on the first element.
26475 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
26476 condition before gimplifying.
26478 2018-02-07 Jakub Jelinek <jakub@redhat.com>
26480 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
26481 *DIV_EXPR and *MOD_EXPR.
26483 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
26486 * config/i386/i386.c (ix86_option_override_internal): Mask out
26487 the CF_SET bit when checking -fcf-protection.
26489 2018-02-07 Tom de Vries <tom@codesourcery.com>
26492 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
26495 2018-02-07 Richard Biener <rguenther@suse.de>
26497 PR tree-optimization/84204
26498 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
26501 PR tree-optimization/84205
26502 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
26503 special-case isl_ast_op_zdiv_r.
26505 PR tree-optimization/84223
26506 * graphite-scop-detection.c (gather_bbs::before_dom_children):
26507 Only add conditions from within the region.
26508 (gather_bbs::after_dom_children): Adjust.
26510 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
26513 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
26514 * config/avr/avr.md: Only post-reload split REG-REG moves if
26515 either register is GENERAL_REG_P.
26517 2018-02-07 Jakub Jelinek <jakub@redhat.com>
26519 PR tree-optimization/84235
26520 * tree-ssa-scopedtables.c
26521 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
26522 if the subtraction is performed in floating point type where NaNs are
26523 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
26524 build 1. Formatting fix.
26526 2018-02-06 Jakub Jelinek <jakub@redhat.com>
26529 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
26530 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
26531 and skip it regardless of bb boundaries. Use CALL_P macro,
26532 don't test INSN_P (insn) together with CALL_P or JUMP_P check
26533 unnecessarily, formatting fix.
26535 2018-02-06 Michael Collison <michael.collison@arm.com>
26537 * config/arm/thumb2.md:
26538 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
26539 (*thumb_mov_notscc): Ditto.
26541 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
26544 * config/rs6000/rs6000.md (su code attribute): Use "u" for
26545 unsigned_fix, not "s".
26547 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
26549 * configure.ac (gcc_fn_eh_frame_ro): New function.
26550 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
26551 correct .eh_frame permissions.
26552 * configure: Regenerate.
26554 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
26556 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
26557 irrelevant options.
26559 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
26561 * config/rs6000/rs6000.c (rs6000_option_override_internal):
26562 Display warning message for -mno-speculate-indirect-jumps.
26564 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
26566 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
26568 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
26570 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
26572 PR tree-optimization/84225
26573 * tree-eh.c (find_trapping_overflow): Only call
26574 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
26576 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
26579 * config/i386/i386.c: Reimplement the check of possible options
26580 -mibt/-mshstk conbination. Change error messages.
26581 * doc/invoke.texi: Fix a typo: remove extra '='.
26583 2018-02-06 Marek Polacek <polacek@redhat.com>
26585 PR tree-optimization/84228
26586 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
26588 2018-02-06 Tamar Christina <tamar.christina@arm.com>
26591 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
26592 emitted arch directives.
26593 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
26594 __ARM_FEATURE_COPROC before changing architectures.
26596 2018-02-06 Richard Biener <rguenther@suse.de>
26598 * config/i386/i386.c (print_reg): Fix typo.
26599 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
26601 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
26603 * configure: Regenerate.
26605 2018-02-05 Martin Sebor <msebor@redhat.com>
26607 PR tree-optimization/83369
26608 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
26611 2018-02-05 Martin Liska <mliska@suse.cz>
26613 * doc/invoke.texi: Cherry-pick upstream r323995.
26615 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
26617 * ira.c (ira_init_register_move_cost): Adjust comment.
26619 2018-02-05 Martin Liska <mliska@suse.cz>
26621 PR gcov-profile/84137
26622 * doc/gcov.texi: Fix typo in documentation.
26624 2018-02-05 Martin Liska <mliska@suse.cz>
26626 PR gcov-profile/83879
26627 * doc/gcov.texi: Document necessity of --dynamic-list-data when
26628 using dlopen functionality.
26630 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
26632 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
26633 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
26634 _mm_maskz_range_ss, _mm_mask_range_round_ss,
26635 _mm_maskz_range_round_ss): New intrinsics.
26636 (__builtin_ia32_rangesd128_round)
26637 (__builtin_ia32_rangess128_round): Remove.
26638 (__builtin_ia32_rangesd128_mask_round,
26639 __builtin_ia32_rangess128_mask_round): New builtins.
26640 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
26641 __builtin_ia32_rangess128_round): Remove.
26642 (__builtin_ia32_rangesd128_mask_round,
26643 __builtin_ia32_rangess128_mask_round): New builtins.
26644 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
26645 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
26646 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
26647 "<round_saeonly_constraint>")): Changed to ...
26648 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
26649 "<round_saeonly_scalar_constraint>")): ... this.
26650 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
26651 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
26652 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
26653 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
26654 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
26656 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
26658 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
26660 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
26661 Remove all values except native, 8540 and 8548.
26663 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
26665 * config/i386/i386.c (ix86_output_function_return): Pass
26666 INVALID_REGNUM, instead of -1, as invalid register number to
26667 indirect_thunk_name and output_indirect_thunk.
26669 2018-02-02 Julia Koval <julia.koval@intel.com>
26671 * config.gcc: Add -march=icelake.
26672 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
26673 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
26674 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
26675 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
26676 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
26677 (processor_target_table): Add icelake.
26678 (ix86_option_override_internal): Handle new PTAs.
26679 (get_builtin_code_for_version): Handle icelake.
26680 (M_INTEL_COREI7_ICELAKE): New.
26681 (fold_builtin_cpu): Handle icelake.
26682 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
26683 * doc/invoke.texi: Add -march=icelake.
26685 2018-02-02 Julia Koval <julia.koval@intel.com>
26687 * config/i386/i386.c (ix86_option_override_internal): Change flags type
26688 to wide_int_bitmask.
26689 * wide-int-bitmask.h: New.
26691 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
26694 * config/i386/i386.md: Replace Pmode with word_mode in
26695 builtin_setjmp_setup and builtin_longjmp to support x32.
26697 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
26701 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
26703 (rs6000_supported_cpu_names): New static variable.
26704 (linux_cpu_translation_table): Likewise.
26705 (elf_platform) <cpu>: Define new static variable and use it.
26706 Translate kernel AT_PLATFORM name to canonical name if needed.
26707 Error if platform name is unknown.
26709 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
26712 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
26714 2018-02-01 Jeff Law <law@redhat.com>
26717 * config/i386/i386.c (release_scratch_register_on_entry): Add new
26718 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
26719 the scratch if RELEASE_VIA_POP is false.
26720 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
26721 If we have to save a temporary register, decrement SIZE appropriately.
26722 Pass new arguments to release_scratch_register_on_entry.
26723 (ix86_adjust_stack_and_probe): Likewise.
26724 (ix86_emit_probe_stack_range): Pass new arguments to
26725 release_scratch_register_on_entry.
26727 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
26729 PR rtl-optimization/84157
26730 * combine.c (change_zero_ext): Use REG_P predicate in
26731 front of HARD_REGISTER_P predicate.
26733 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
26735 * config/avr/avr.c (avr_option_override): Move disabling of
26736 -fdelete-null-pointer-checks to...
26737 * common/config/avr/avr-common.c (avr_option_optimization_table):
26740 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
26742 PR tree-optimization/81635
26743 * tree-data-ref.c (split_constant_offset_1): For types that
26744 wrap on overflow, try to use range info to prove that wrapping
26747 2018-02-01 Renlin Li <renlin.li@arm.com>
26750 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
26751 TAILCALL_ADDR_REGS.
26752 (aarch64_register_move_cost): Likewise.
26753 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
26754 TAILCALL_ADDR_REGS.
26755 (REG_CLASS_NAMES): Likewise.
26756 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
26757 TAILCALL_ADDR_REGS. Remove IP registers.
26758 * config/aarch64/aarch64.md (Ucs): Update register constraint.
26760 2018-02-01 Richard Biener <rguenther@suse.de>
26762 * domwalk.h (dom_walker::dom_walker): Add additional constructor
26763 for specifying RPO order and allow NULL for that.
26764 * domwalk.c (dom_walker::dom_walker): Likewise.
26765 (dom_walker::walk): Handle NULL RPO order.
26766 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
26768 (rewrite_update_dom_walker): Likewise.
26769 (mark_def_dom_walker): Likewise.
26771 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
26773 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
26774 (aarch64_maybe_expand_sve_subreg_move): Declare.
26775 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
26776 * config/aarch64/predicates.md (aarch64_any_register_operand): New
26778 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
26779 that are semantically a reverse operation.
26780 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
26781 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
26782 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
26784 (aarch64_can_change_mode_class): For big-endian, forbid changes
26785 between two SVE modes if they have different element sizes.
26787 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
26789 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
26790 the TImode handling for big-endian targets.
26792 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
26794 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
26795 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
26797 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
26798 Remove BSWAP handing for big-endian targets and use the form of
26799 LD1RQ appropariate for the mode.
26801 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
26803 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
26804 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
26805 duplicated element.
26807 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
26810 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
26811 check for operands that need to go through aarch64_sve_reload_be.
26813 2018-02-01 Jakub Jelinek <jakub@redhat.com>
26815 PR tree-optimization/81661
26816 PR tree-optimization/84117
26817 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
26818 * tree-eh.c: Include gimplify.h.
26819 (find_trapping_overflow, replace_trapping_overflow,
26820 rewrite_to_non_trapping_overflow): New functions.
26821 * tree-vect-loop.c: Include tree-eh.h.
26822 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
26823 * tree-data-ref.c: Include tree-eh.h.
26824 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
26826 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
26828 PR rtl-optimization/84123
26829 * combine.c (change_zero_ext): Check if hard register satisfies
26830 can_change_dest_mode before calling gen_lowpart_SUBREG.
26832 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
26835 * ira.c (ira_init_register_move_cost): Remove assert.
26837 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
26839 PR rtl-optimization/84071
26840 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
26841 * doc/tm.texi: Regenerate.
26843 2018-01-31 Richard Biener <rguenther@suse.de>
26845 PR tree-optimization/84132
26846 * tree-data-ref.c (analyze_miv_subscript): Properly
26847 check whether evolution_function_is_affine_multivariate_p
26848 before calling gcd_of_steps_may_divide_p.
26850 2018-01-31 Julia Koval <julia.koval@intel.com>
26853 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
26854 * config/i386/i386.md (rdpid_rex64) New.
26855 (rdpid): Make 32bit only.
26857 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
26860 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
26861 an IDENTIFIER_NODE for FUNCTION_TYPE's.
26863 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
26866 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
26868 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
26870 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
26872 PR rtl-optimization/84071
26873 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
26874 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
26876 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
26878 * config/arc/arc.c (arc_handle_aux_attribute): New function.
26879 (arc_attribute_table): Add 'aux' attribute.
26880 (arc_in_small_data_p): Consider aux like variables.
26881 (arc_is_aux_reg_p): New function.
26882 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
26883 (arc_get_aux_arg): New function.
26884 (prepare_move_operands): Handle aux-register access.
26885 (arc_handle_aux_attribute): New function.
26886 * doc/extend.texi (ARC Variable attributes): Add subsection.
26888 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
26890 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
26891 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
26892 (arc_attribute_table): Add 'uncached' attribute.
26893 (arc_print_operand): Print '.di' flag for uncached memory
26895 (arc_in_small_data_p): Do not consider for small data the uncached
26897 (arc_is_uncached_mem_p): New function.
26898 * config/arc/predicates.md (compact_store_memory_operand): Check
26899 for uncached memory accesses.
26900 (nonvol_nonimm_operand): Likewise.
26901 * doc/extend.texi (ARC Type Attribute): New subsection.
26903 2018-01-31 Jakub Jelinek <jakub@redhat.com>
26906 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
26907 falign-loops=): Add Optimization flag.
26909 2018-01-30 Jeff Law <law@redhat.com>
26912 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
26913 INT_REGISTERS_SAVED. Check it prior to calling
26914 get_scratch_register_on_entry.
26915 (ix86_adjust_stack_and_probe): Similarly.
26916 (ix86_emit_probe_stack_range): Similarly.
26917 (ix86_expand_prologue): Corresponding changes.
26919 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
26922 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
26923 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
26925 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
26928 * lra-constraints.c (curr_insn_transform): Process AND in the
26931 2018-01-30 Jakub Jelinek <jakub@redhat.com>
26933 PR rtl-optimization/83986
26934 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
26935 dependence against last_pending_memory_flush in addition to
26936 pending_jump_insns.
26938 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
26940 PR tree-optimization/81611
26941 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
26944 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
26947 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
26950 2018-01-30 Richard Biener <rguenther@suse.de>
26951 Jakub Jelinek <jakub@redhat.com>
26953 PR tree-optimization/84111
26954 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
26955 inner loops added during recursion, as they don't have up-to-date
26958 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
26961 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
26962 (can_inline_edge_by_limits_p): ... here.
26963 (can_early_inline_edge_p, check_callers,
26964 update_caller_keys, update_callee_keys, recursive_inlining,
26965 add_new_edges_to_heap, speculation_useful_p,
26966 inline_small_functions,
26967 inline_small_functions, flatten_function,
26968 inline_to_all_callers_1): Update.
26970 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
26972 * profile-count.c (profile_count::combine_with_ipa_count): Handle
26975 2018-01-30 Richard Biener <rguenther@suse.de>
26977 PR tree-optimization/83008
26978 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
26979 invariant and constant vector uses in stmts when they need
26980 more than one stmt.
26982 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
26985 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
26986 * configure: Regenerate.
26988 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
26990 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
26992 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
26993 Use gen_rtx_REG rather than gen_lowpart.
26995 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
26997 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
26998 rather than 0 when creating partial subregs.
27000 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
27002 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
27005 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
27008 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
27009 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
27010 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
27011 flags. This restores the settings used before the 2017-07-24.
27012 Turning off pre increment/decrement/modify allows IVOPTS to
27013 optimize DF/SF loops where the index is an int.
27015 2018-01-29 Richard Biener <rguenther@suse.de>
27016 Kelvin Nilsen <kelvin@gcc.gnu.org>
27019 * tree-vect-stmts.c (vectorizable_call): Don't call
27020 targetm.vectorize_builtin_md_vectorized_function if callee is
27023 2018-01-22 Carl Love <cel@us.ibm.com>
27025 * doc/extend.tex: Fix typo in second arg in
27026 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
27028 2018-01-29 Richard Biener <rguenther@suse.de>
27030 PR tree-optimization/84086
27031 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
27032 (flush_ssaname_freelist): When SSA names were released reset
27033 the SCEV hash table.
27035 2018-01-29 Richard Biener <rguenther@suse.de>
27037 PR tree-optimization/84057
27038 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
27039 removed paths when removing edges.
27041 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
27043 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
27044 -mfunction-return=@var{choice}.
27046 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
27048 PR diagnostic/84034
27049 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
27050 Handle CR like TAB.
27051 (layout::print_source_line): Likewise.
27052 (test_get_line_width_without_trailing_whitespace): Add test cases.
27054 2018-01-27 Jakub Jelinek <jakub@redhat.com>
27056 PR middle-end/84040
27057 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
27060 2018-01-26 Jim Wilson <jimw@sifive.com>
27062 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
27064 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
27067 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27069 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
27070 and CMP + SUB-immediate -> SUBS.
27072 2018-01-26 Martin Sebor <msebor@redhat.com>
27074 PR tree-optimization/83896
27075 * tree-ssa-strlen.c (get_string_len): Rename...
27076 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
27077 Avoid assuming length is constant.
27078 (handle_char_store): Use HOST_WIDE_INT for string length.
27080 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
27083 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
27084 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
27086 2018-01-26 Richard Biener <rguenther@suse.de>
27088 PR rtl-optimization/84003
27089 * dse.c (record_store): Only record redundant stores when
27090 the earlier store aliases at least all accesses the later one does.
27092 2018-01-26 Jakub Jelinek <jakub@redhat.com>
27094 PR rtl-optimization/83985
27095 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
27096 REG_CFA_RESTORE insns.
27097 (delete_unmarked_insns): Don't ignore separate shrink wrapping
27098 REG_CFA_RESTORE insns here.
27101 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
27102 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
27104 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
27106 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
27107 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
27108 (arc_init): Likewise.
27109 (arc_override_options): Likewise.
27110 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
27112 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
27114 * config/arc/arc.h (TARGET_DBNZ): Define.
27115 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
27116 properly set the tune attribute.
27117 (dbnz): Use TARGET_DBNZ guard.
27118 * config/arc/arc.opt (mtune): Add core3 option.
27120 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
27122 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
27123 recognize new pic like addresses.
27124 (arc_delegitimize_address): Clean up.
27126 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
27128 * config/arc/arc-arches.def: Option mrf16 valid for all
27130 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
27131 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
27132 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
27133 * config/arc/arc-tables.opt: Regenerate.
27134 * config/arc/arc.c (arc_conditional_register_usage): Handle
27135 reduced register file case.
27136 (arc_file_start): Set must have build attributes.
27137 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
27138 mrf16 option value.
27139 * config/arc/arc.opt (mrf16): Add new option.
27140 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
27141 * config/arc/genmultilib.awk: Handle new mrf16 option.
27142 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
27143 * config/arc/t-multilib: Regenerate.
27144 * doc/invoke.texi (ARC Options): Document mrf16 option.
27146 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
27148 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
27149 * config/arc/arc.c (arc_handle_secure_attribute): New function.
27150 (arc_attribute_table): Add 'secure_call' attribute.
27151 (arc_print_operand): Print secure call operand.
27152 (arc_function_ok_for_sibcall): Don't optimize tail calls when
27154 (arc_is_secure_call_p): New function. * config/arc/arc.md
27155 (call_i): Add support for sjli instruction.
27156 (call_value_i): Likewise.
27157 * config/arc/constraints.md (Csc): New constraint.
27159 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
27160 John Eric Martin <John.Martin@emmicro-us.com>
27162 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
27163 * config/arc/arc.c (_arc_jli_section): New struct.
27164 (arc_jli_section): New type.
27165 (rc_jli_sections): New static variable.
27166 (arc_handle_jli_attribute): New function.
27167 (arc_attribute_table): Add jli_always and jli_fixed attribute.
27168 (arc_file_end): New function.
27169 (TARGET_ASM_FILE_END): Define.
27170 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
27171 (arc_add_jli_section): New function.
27172 (jli_call_scan): Likewise.
27173 (arc_reorg): Call jli_call_scan.
27174 (arc_output_addsi): Remove 'S' from printing asm operand.
27175 (arc_is_jli_call_p): New function.
27176 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
27178 (movhi_insn): Likewise.
27179 (movsi_insn): Likewise.
27180 (movsi_set_cc_insn): Likewise.
27181 (loadqi_update): Likewise.
27182 (load_zeroextendqisi_update): Likewise.
27183 (load_signextendqisi_update): Likewise.
27184 (loadhi_update): Likewise.
27185 (load_zeroextendhisi_update): Likewise.
27186 (load_signextendhisi_update): Likewise.
27187 (loadsi_update): Likewise.
27188 (loadsf_update): Likewise.
27189 (movsicc_insn): Likewise.
27190 (bset_insn): Likewise.
27191 (bxor_insn): Likewise.
27192 (bclr_insn): Likewise.
27193 (bmsk_insn): Likewise.
27194 (bicsi3_insn): Likewise.
27195 (cmpsi_cc_c_insn): Likewise.
27196 (movsi_ne): Likewise.
27197 (movsi_cond_exec): Likewise.
27198 (clrsbsi2): Likewise.
27199 (norm_f): Likewise.
27209 (call_i): Remove 'S' asm letter, add jli instruction.
27210 (call_value_i): Likewise.
27211 * config/arc/arc.op (mjli-always): New option.
27212 * config/arc/constraints.md (Cji): New constraint.
27213 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
27215 (subsf3_fpx): Likewise.
27216 (mulsf3_fpx): Likewise.
27217 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
27219 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
27220 function attrbutes.
27221 * doc/invoke.texi (ARC): Document mjli-always option.
27223 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
27225 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
27226 avoid addition with 0 and use incw and decw where possible.
27228 2018-01-26 Richard Biener <rguenther@suse.de>
27230 PR tree-optimization/81082
27231 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
27232 association if it requires casting to unsigned.
27233 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
27234 from fold_plusminus_mult_expr to catch important cases late when
27235 range info is available.
27237 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
27239 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
27240 * configure.ac (hidden_linkonce): New test.
27241 * configure: Regenerate.
27242 * config.in: Regenerate.
27244 2018-01-26 Julia Koval <julia.koval@intel.com>
27246 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
27247 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
27248 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
27249 _mm_mask_bitshuffle_epi64_mask): Fix type.
27250 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
27251 USI_FTYPE_V4DI_V4DI_USI): Remove.
27252 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
27253 __builtin_ia32_vpshufbitqmb256_mask,
27254 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
27255 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
27256 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
27258 2018-01-26 Alan Modra <amodra@gmail.com>
27261 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
27262 UNSPEC_VBPERMQ. Sort other unspecs.
27264 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
27266 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
27268 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
27270 PR middle-end/83055
27271 * predict.c (drop_profile): Do not push/pop cfun; update also
27273 (handle_missing_profiles): Fix logic looking for zero profiles.
27275 2018-01-25 Jakub Jelinek <jakub@redhat.com>
27277 PR middle-end/83977
27278 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
27279 on functions with #pragma omp declare simd or functions with simd
27281 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
27282 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
27283 Remove trailing \n from warning_at calls.
27285 2018-01-25 Tom de Vries <tom@codesourcery.com>
27288 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
27289 for neutered workers.
27291 2018-01-24 Joseph Myers <joseph@codesourcery.com>
27294 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
27295 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
27297 2018-01-24 Jeff Law <law@redhat.com>
27300 * i386.c (get_probe_interval): Move to earlier point.
27301 (ix86_compute_frame_layout): If -fstack-clash-protection and
27302 the frame is larger than the probe interval, then use pushes
27303 to save registers rather than reg->mem moves.
27304 (ix86_expand_prologue): Remove conditional for int_registers_saved
27307 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
27310 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
27311 min/max for never referenced object.
27313 2018-01-24 Jakub Jelinek <jakub@redhat.com>
27315 PR middle-end/83977
27316 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
27318 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
27319 attributes from DECL_ATTRIBUTES (decl) without affecting
27320 DECL_ATTRIBUTES (current_function_decl).
27321 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
27322 functions with non-NULL DECL_ABSTRACT_ORIGIN.
27324 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
27326 PR tree-optimization/83979
27327 * fold-const.c (fold_comparison): Use constant_boolean_node
27328 instead of boolean_{true,false}_node.
27330 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
27332 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
27335 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27337 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27338 Simplify the clause that sets the length attribute.
27339 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27340 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
27341 clause that sets the length attribute.
27342 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27344 2018-01-24 Tom de Vries <tom@codesourcery.com>
27347 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
27348 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
27349 Add strict parameter.
27350 (prevent_branch_around_nothing): Insert dummy insn between branch to
27351 label and label with no ptx insn inbetween.
27352 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
27354 2018-01-24 Tom de Vries <tom@codesourcery.com>
27357 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
27358 for neutered threads in warp.
27359 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
27361 2018-01-24 Richard Biener <rguenther@suse.de>
27363 PR tree-optimization/83176
27364 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
27367 2018-01-24 Richard Biener <rguenther@suse.de>
27369 PR tree-optimization/82819
27370 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
27371 code generating pluses that are no-ops in the target precision.
27373 2018-01-24 Richard Biener <rguenther@suse.de>
27375 PR middle-end/84000
27376 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
27378 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
27380 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
27381 to merge probabilities.
27382 * predict.c (probably_never_executed): Also mark as cold functions
27383 with global 0 profile and guessed local profile.
27384 * profile-count.c (profile_probability::combine_with_count): New
27386 * profile-count.h (profile_probability::operator*,
27387 profile_probability::operator*=, profile_probability::operator/,
27388 profile_probability::operator/=): Reduce precision to adjusted
27389 and set value to guessed on contradictory divisions.
27390 (profile_probability::combine_with_freq): Remove.
27391 (profile_probability::combine_wiht_count): Declare.
27392 (profile_count::force_nonzero):: Set to adjusted.
27393 (profile_count::probability_in):: Set quality to adjusted.
27394 * tree-ssa-tail-merge.c (replace_block_by): Use
27395 combine_with_count.
27397 2018-01-23 Andrew Waterman <andrew@sifive.com>
27398 Jim Wilson <jimw@sifive.com>
27400 * config/riscv/riscv.c (riscv_stack_boundary): New.
27401 (riscv_option_override): Set riscv_stack_boundary. Handle
27402 riscv_preferred_stack_boundary_arg.
27403 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
27404 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
27405 (STACK_BOUNDARY): Set to riscv_stack_boundary.
27406 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
27407 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
27408 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
27410 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
27413 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
27414 of struct ix86_frame.
27415 (ix86_expand_epilogue): Likewise. Add a local variable for
27416 the reg_save_offset field in struct ix86_frame.
27418 2018-01-23 Bin Cheng <bin.cheng@arm.com>
27420 PR tree-optimization/82604
27421 * tree-loop-distribution.c (enum partition_kind): New enum item
27422 PKIND_PARTIAL_MEMSET.
27423 (partition_builtin_p): Support above new enum item.
27424 (generate_code_for_partition): Ditto.
27425 (compute_access_range): Differentiate cases that equality can be
27426 proven at all loops, the innermost loops or no loops.
27427 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
27428 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
27429 (finalize_partitions, distribute_loop): Don't fuse partition of
27430 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
27431 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
27432 parloop is enabled.
27434 2018-01-23 Martin Liska <mliska@suse.cz>
27436 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
27437 order to ignore the predictor.
27438 (PRED_POLYMORPHIC_CALL): Likewise.
27439 (PRED_RECURSIVE_CALL): Likewise.
27441 2018-01-23 Martin Liska <mliska@suse.cz>
27443 * tree-profile.c (tree_profiling): Print function header to
27444 aware reader which function we are working on.
27445 * value-prof.c (gimple_find_values_to_profile): Do not print
27446 not interesting value histograms.
27448 2018-01-23 Martin Liska <mliska@suse.cz>
27450 * profile-count.h (enum profile_quality): Add
27451 profile_uninitialized as the first value. Do not number values
27452 as they are zero based.
27453 (profile_count::verify): Update sanity check.
27454 (profile_probability::verify): Likewise.
27456 2018-01-23 Nathan Sidwell <nathan@acm.org>
27458 * doc/invoke.texi (ffor-scope): Deprecate.
27460 2018-01-23 David Malcolm <dmalcolm@redhat.com>
27462 PR tree-optimization/83510
27463 * domwalk.c (set_all_edges_as_executable): New function.
27464 (dom_walker::dom_walker): Convert bool param
27465 "skip_unreachable_blocks" to enum reachability. Move setup of
27466 edge flags to set_all_edges_as_executable and only do it when
27467 reachability is REACHABLE_BLOCKS.
27468 * domwalk.h (enum dom_walker::reachability): New enum.
27469 (dom_walker::dom_walker): Convert bool param
27470 "skip_unreachable_blocks" to enum reachability.
27471 (set_all_edges_as_executable): New decl.
27472 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
27473 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
27475 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
27476 but converting true to REACHABLE_BLOCKS.
27477 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
27479 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
27480 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
27481 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
27483 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
27484 if check_all_array_refs will be called.
27486 2018-01-23 David Malcolm <dmalcolm@redhat.com>
27488 * tree.c (selftest::test_location_wrappers): Add more test
27491 2018-01-23 David Malcolm <dmalcolm@redhat.com>
27493 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
27494 (selftest::test_bit_in_range): Likewise.
27496 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
27499 * doc/sourcebuild.texi (vect_float): Say that the selector
27500 only describes the situation when -funsafe-math-optimizations is on.
27501 (vect_float_strict): Document.
27503 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
27505 PR tree-optimization/83965
27506 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
27507 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
27508 instead of checking only for a reduction.
27509 (vect_recog_widen_sum_pattern): Likewise.
27511 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
27513 * predict.c (probably_never_executed): Only use precise profile info.
27514 (compute_function_frequency): Skip after inlining hack since we now
27515 have quality checking.
27517 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
27519 * profile-count.h (profile_probability::very_unlikely,
27520 profile_probability::unlikely, profile_probability::even): Set
27521 precision to guessed.
27523 2018-01-23 Richard Biener <rguenther@suse.de>
27525 PR tree-optimization/83963
27526 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
27527 Properly terminate dominator walk when crossing the exit edge not
27528 when visiting its source block.
27530 2018-01-23 Jakub Jelinek <jakub@redhat.com>
27533 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
27534 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
27536 2018-01-22 Jakub Jelinek <jakub@redhat.com>
27538 PR tree-optimization/83957
27539 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
27540 semicolon after for body surrounded by braces.
27542 PR tree-optimization/83081
27543 * profile-count.h (profile_probability::split): New method.
27544 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
27545 Use profile_probability::split.
27546 (do_compare_rtx_and_jump): Fix adjustment of probabilities
27547 when splitting a single conditional jump into 2.
27549 2018-01-22 David Malcolm <dmalcolm@redhat.com>
27551 PR tree-optimization/69452
27552 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
27555 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
27557 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
27558 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
27559 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
27561 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
27563 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
27565 * config/rl78/rl78.md (movdi): New define_expand.
27566 * config/rl78/rl78.c (rl78_split_movdi): New function.
27568 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
27571 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
27573 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
27574 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
27575 128-bit to produce an UNSPEC move to get the double word with the
27576 signbit and then a shift directly to do signbit.
27577 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
27578 implementation with a new version that just does either a direct
27579 move or a regular move. Move memory interface to separate insns.
27580 Move insns so they are next to the expander.
27581 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
27582 with signbit move. Split big and little endian case.
27583 (signbit<mode>2_dm_mem_le): Likewise.
27584 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
27585 (signbit<mode>2_dm2): Likewise.
27587 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
27589 * config/rl78/rl78.md (anddi3): New define_expand.
27591 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
27593 * config/rl78/rl78.md (umindi3): New define_expand.
27595 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
27597 * config/rl78/rl78.md (smindi3): New define_expand.
27599 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
27601 * config/rl78/rl78.md (smaxdi3): New define_expand.
27603 2018-01-22 Carl Love <cel@us.ibm.com>
27605 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
27606 LVX_V1TI): Add macro expansion.
27607 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
27608 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
27609 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
27610 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
27611 Change check to determine if the instruction is a byte reversing
27612 entry. Fix typo in comment.
27613 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
27614 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
27615 Add def_builtin calls for new builtins.
27616 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
27617 Add define_insn expansion.
27619 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
27621 * config/rl78/rl78.md (umaxdi3): New define_expand.
27623 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
27625 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
27626 for non-QImode registers.
27628 2018-01-22 Richard Biener <rguenther@suse.de>
27630 PR tree-optimization/83963
27631 * graphite-scop-detection.c (scop_detection::get_sese): Delay
27632 including the loop exit block.
27633 (scop_detection::merge_sese): Likewise.
27634 (scop_detection::add_scop): Do it here instead.
27636 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27638 * doc/sourcebuild.texi (arm_softfloat): Document.
27640 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
27643 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
27644 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
27645 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
27647 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27648 David Edelsohn <dje.gcc@gmail.com>
27651 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27652 Change "crset eq" to "crset 2".
27653 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27654 (*call_indirect_aix<mode>_nospec): Likewise.
27655 (*call_value_indirect_aix<mode>_nospec): Likewise.
27656 (*call_indirect_elfv2<mode>_nospec): Likewise.
27657 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
27658 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
27659 change assembly output from . to $.
27660 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27661 (indirect_jump<mode>_nospec): Change assembly output from . to $.
27662 (*tablejump<mode>_internal1_nospec): Likewise.
27664 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
27667 * config/sh/sh_optimize_sett_clrt.cc:
27668 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
27670 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
27672 PR tree-optimization/83940
27673 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
27674 offset_dt to vect_constant_def rather than vect_unknown_def_type.
27675 (vect_check_load_store_mask): Add a mask_dt_out parameter and
27676 use it to pass back the definition type.
27677 (vect_check_store_rhs): Likewise rhs_dt_out.
27678 (vect_build_gather_load_calls): Add a mask_dt argument and use
27679 it instead of a call to vect_is_simple_use.
27680 (vectorizable_store): Update calls to vect_check_load_store_mask
27681 and vect_check_store_rhs. Use the dt returned by the latter instead
27682 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
27683 instead of calls to vect_is_simple_use. Pass the scalar rather
27684 than the vector operand to vect_is_simple_use when handling
27685 second and subsequent copies of an rhs value.
27686 (vectorizable_load): Update calls to vect_check_load_store_mask
27687 and vect_build_gather_load_calls. Use the cached mask_dt and
27688 gs_info.offset_dt instead of calls to vect_is_simple_use.
27690 2018-01-20 Jakub Jelinek <jakub@redhat.com>
27692 PR middle-end/83945
27693 * tree-emutls.c: Include gimplify.h.
27694 (lower_emutls_2): New function.
27695 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
27696 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
27697 it before further processing.
27700 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
27701 UINTVAL (trueop1) instead of INTVAL (op1).
27703 2018-01-19 Jakub Jelinek <jakub@redhat.com>
27707 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
27708 INCOMING_FRAME_SP_OFFSET if not defined.
27709 (scan_trace): Add ENTRY argument. If true and
27710 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
27711 emit a note to adjust the CFA offset.
27712 (create_cfi_notes): Adjust scan_trace callers.
27713 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
27714 INCOMING_FRAME_SP_OFFSET in the CIE.
27715 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
27716 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
27718 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
27719 * doc/tm.texi: Regenerated.
27721 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
27723 PR rtl-optimization/83147
27724 * lra-constraints.c (remove_inheritance_pseudos): Use
27725 lra_substitute_pseudo_within_insn.
27727 2018-01-19 Tom de Vries <tom@codesourcery.com>
27728 Cesar Philippidis <cesar@codesourcery.com>
27731 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
27733 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
27736 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
27737 spaces for function labels.
27739 2018-01-19 Martin Liska <mliska@suse.cz>
27741 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
27742 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
27743 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
27744 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
27745 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
27746 (PRED_CONST_RETURN): Change from 69 to 65.
27747 (PRED_NULL_RETURN): Change from 91 to 71.
27748 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
27749 (PRED_LOOP_GUARD): Change from 66 to 73.
27751 2018-01-19 Martin Liska <mliska@suse.cz>
27753 * predict.c (predict_insn_def): Add new assert.
27754 (struct branch_predictor): Change type to signed integer.
27755 (test_prediction_value_range): Amend test to cover
27756 PROB_UNINITIALIZED.
27757 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
27758 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
27759 (PRED_LOOP_ITERATIONS_MAX): Likewise.
27760 (PRED_LOOP_IV_COMPARE): Likewise.
27761 * predict.h (PROB_UNINITIALIZED): Define new constant.
27763 2018-01-19 Martin Liska <mliska@suse.cz>
27765 * predict.c (dump_prediction): Add new format for
27766 analyze_brprob.py script which is enabled with -details
27768 * profile-count.h (precise_p): New function.
27770 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
27772 PR tree-optimization/83922
27773 * tree-vect-loop.c (vect_verify_full_masking): Return false if
27774 there are no statements that need masking.
27775 (vect_active_double_reduction_p): New function.
27776 (vect_analyze_loop_operations): Use it when handling phis that
27777 are not in the loop header.
27779 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
27781 PR tree-optimization/83914
27782 * tree-vect-loop.c (vectorizable_induction): Don't convert
27783 init_expr or apply the peeling adjustment for inductions
27784 that are nested within the vectorized loop.
27786 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
27788 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
27791 2018-01-18 Jakub Jelinek <jakub@redhat.com>
27795 * function.h (gimplify_parameters): Add gimple_seq * argument.
27796 * function.c: Include gimple.h and options.h.
27797 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
27798 for the added local temporaries if needed.
27799 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
27800 if there are any parameter cleanups, wrap whole body into a
27801 try/finally with the cleanups.
27803 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
27806 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27807 Use GET_MODE_CLASS for scalar floating point.
27809 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
27813 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
27814 Fix call of call_cgraph_insertion_hooks.
27816 2018-01-18 Martin Sebor <msebor@redhat.com>
27818 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
27820 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
27823 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
27826 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
27829 * common.opt: (-ffile-prefix-map): New option.
27830 * opts.c (common_handle_option): Defer it.
27831 * opts-global.c (handle_common_deferred_options): Handle it.
27832 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
27833 * file-prefix-map.h: New file.
27834 (remap_debug_filename, add_debug_prefix_map): ...here.
27835 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
27836 * final.c (debug_prefix_map, add_debug_prefix_map
27837 remap_debug_filename): Move to...
27838 * file-prefix-map.c: New file.
27839 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
27840 generalize, get rid of alloca(), use strrchr() instead of strchr().
27841 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
27842 Implement in terms of add_prefix_map().
27843 (remap_macro_filename, remap_debug_filename): Implement in term of
27845 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
27846 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
27847 * dbxout.c: Include file-prefix-map.h.
27848 * varasm.c: Likewise.
27849 * vmsdbgout.c: Likewise.
27850 * xcoffout.c: Likewise.
27851 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
27852 * doc/cppopts.texi (-fmacro-prefix-map): Document.
27853 * doc/invoke.texi (-ffile-prefix-map): Document.
27854 (-fdebug-prefix-map): Update description.
27856 2018-01-18 Martin Liska <mliska@suse.cz>
27858 * config/i386/i386.c (indirect_thunk_name): Document that also
27860 (output_indirect_thunk): Document why both instructions
27861 (pause and lfence) are generated.
27863 2018-01-18 Richard Biener <rguenther@suse.de>
27865 PR tree-optimization/83887
27866 * graphite-scop-detection.c
27867 (scop_detection::get_nearest_dom_with_single_entry): Remove.
27868 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
27869 (scop_detection::merge_sese): Re-implement with a flood-fill
27870 algorithm that properly finds a SESE region if it exists.
27872 2018-01-18 Jakub Jelinek <jakub@redhat.com>
27875 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
27876 pointer_diff optimizations use view_convert instead of convert.
27878 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
27880 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27881 Generate different code for -mno-speculate-indirect-jumps.
27882 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27883 (*call_indirect_aix<mode>): Disable for
27884 -mno-speculate-indirect-jumps.
27885 (*call_indirect_aix<mode>_nospec): New define_insn.
27886 (*call_value_indirect_aix<mode>): Disable for
27887 -mno-speculate-indirect-jumps.
27888 (*call_value_indirect_aix<mode>_nospec): New define_insn.
27889 (*sibcall_nonlocal_sysv<mode>): Generate different code for
27890 -mno-speculate-indirect-jumps.
27891 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27893 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
27895 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
27896 long double type, set the flags for noting the default long double
27897 type, even if we don't pass or return a long double type.
27899 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
27902 * ipa-inline.c (flatten_function): Do not overwrite final inlining
27905 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
27907 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
27908 support for merge[hl].
27909 (fold_mergehl_helper): New helper function.
27910 (tree-vector-builder.h): New #include for tree_vector_builder usage.
27911 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
27912 (altivec_vmrglw_direct): Add xxmrglw insn.
27914 2018-01-17 Andrew Waterman <andrew@sifive.com>
27916 * config/riscv/riscv.c (riscv_conditional_register_usage): If
27917 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
27919 2018-01-17 David Malcolm <dmalcolm@redhat.com>
27922 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
27923 call the lto_location_cache before reading the
27924 DECL_SOURCE_LOCATION of the types.
27926 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
27927 Richard Sandiford <richard.sandiford@linaro.org>
27929 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
27930 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
27931 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
27932 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
27933 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
27935 * config/aarch64/constraints.md (aarch64_movti_operand):
27937 * config/aarch64/predicates.md (Uti): Add new constraint.
27939 2018-01-17 Carl Love <cel@us.ibm.com>
27941 * config/rs6000/vsx.md (define_expand xl_len_r,
27942 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
27943 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
27945 (define_expand, define_insn): Move the shift left from the
27946 define_insn to the define_expand for lxvl and stxvl instructions.
27947 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
27948 and XL_LEN_R definitions to PURE.
27950 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
27952 * config/i386/i386.c (indirect_thunk_name): Declare regno
27953 as unsigned int. Compare regno with INVALID_REGNUM.
27954 (output_indirect_thunk): Ditto.
27955 (output_indirect_thunk_function): Ditto.
27956 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
27957 in the call to output_indirect_thunk_function.
27959 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
27961 PR middle-end/83884
27962 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
27963 rather than the size of inner_type to determine the stack slot size
27964 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
27966 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
27969 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
27972 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
27974 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
27975 endian Linux systems to optionally enable multilibs for selecting
27976 the long double type if the user configured an explicit type.
27977 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
27978 have no long double multilibs if not defined.
27979 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
27980 warn if the user used -mabi={ieee,ibm}longdouble and we built
27981 multilibs for long double.
27982 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
27983 appropriate multilib option.
27984 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
27986 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
27987 for building long double multilibs.
27988 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
27990 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
27992 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
27995 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
27997 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
28000 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
28003 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
28006 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
28008 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
28009 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
28011 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
28013 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
28014 different rtl trees depending on TARGET_64BIT.
28015 (rs6000_gen_lvx): Likewise.
28017 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
28019 * config/visium/visium.md (nop): Tweak comment.
28020 (hazard_nop): Likewise.
28022 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
28024 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
28025 -mspeculate-indirect-jumps.
28026 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
28027 for -mno-speculate-indirect-jumps.
28028 (*call_indirect_elfv2<mode>_nospec): New define_insn.
28029 (*call_value_indirect_elfv2<mode>): Disable for
28030 -mno-speculate-indirect-jumps.
28031 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
28032 (indirect_jump): Emit different RTL for
28033 -mno-speculate-indirect-jumps.
28034 (*indirect_jump<mode>): Disable for
28035 -mno-speculate-indirect-jumps.
28036 (*indirect_jump<mode>_nospec): New define_insn.
28037 (tablejump): Emit different RTL for
28038 -mno-speculate-indirect-jumps.
28039 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
28040 (tablejumpsi_nospec): New define_expand.
28041 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
28042 (tablejumpdi_nospec): New define_expand.
28043 (*tablejump<mode>_internal1): Disable for
28044 -mno-speculate-indirect-jumps.
28045 (*tablejump<mode>_internal1_nospec): New define_insn.
28046 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
28049 2018-01-16 Artyom Skrobov tyomitch@gmail.com
28051 * caller-save.c (insert_save): Drop unnecessary parameter. All
28054 2018-01-16 Jakub Jelinek <jakub@redhat.com>
28055 Richard Biener <rguenth@suse.de>
28058 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
28059 return early, inline manually is_gimple_sizepos. Make sure if we
28060 call gimplify_expr we don't end up with a gimple constant.
28061 * tree.c (variably_modified_type_p): Don't return true for
28062 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
28063 * gimplify.h (is_gimple_sizepos): Remove.
28065 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
28067 PR tree-optimization/83857
28068 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
28069 vectorizable_live_operation for pure SLP statements.
28070 (vectorizable_live_operation): Handle PHIs.
28072 2018-01-16 Richard Biener <rguenther@suse.de>
28074 PR tree-optimization/83867
28075 * tree-vect-stmts.c (vect_transform_stmt): Precompute
28076 nested_in_vect_loop_p since the scalar stmt may get invalidated.
28078 2018-01-16 Jakub Jelinek <jakub@redhat.com>
28081 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
28082 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
28083 If off is not INTEGER_CST, issue a may not be aligned warning
28084 rather than isn't aligned. Use isn%'t rather than isn't.
28085 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
28087 <case MULT_EXPR>: Improve the case when bottom and one of the
28088 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
28089 operand, in that case check if the other operand is multiple of
28090 bottom divided by the INTEGER_CST operand.
28092 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
28095 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
28096 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
28097 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
28098 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
28099 * config/pa/pa.c (pa_function_arg_advance): Likewise.
28100 (pa_function_arg, pa_arg_partial_bytes): Likewise.
28101 (pa_function_arg_size): New function.
28103 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
28105 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
28106 in a separate statement.
28108 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
28110 PR tree-optimization/83847
28111 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
28112 group gathers and scatters.
28114 2018-01-16 Jakub Jelinek <jakub@redhat.com>
28116 PR rtl-optimization/86620
28117 * params.def (max-sched-ready-insns): Bump minimum value to 1.
28119 PR rtl-optimization/83213
28120 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
28121 to last if both are JUMP_INSNs.
28123 PR tree-optimization/83843
28124 * gimple-ssa-store-merging.c
28125 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
28126 store_immediate_info for bswap/nop orig_stores.
28128 2018-01-15 Andrew Waterman <andrew@sifive.com>
28130 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
28132 <UDIV>: Increase cost if !TARGET_DIV.
28134 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
28136 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
28137 (define_attr "cr_logical_3op"): New.
28138 (cceq_ior_compare): Adjust.
28139 (cceq_ior_compare_complement): Adjust.
28140 (*cceq_rev_compare): Adjust.
28141 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
28142 (is_cracked_insn): Adjust.
28143 (insn_must_be_first_in_group): Adjust.
28144 * config/rs6000/40x.md: Adjust.
28145 * config/rs6000/440.md: Adjust.
28146 * config/rs6000/476.md: Adjust.
28147 * config/rs6000/601.md: Adjust.
28148 * config/rs6000/603.md: Adjust.
28149 * config/rs6000/6xx.md: Adjust.
28150 * config/rs6000/7450.md: Adjust.
28151 * config/rs6000/7xx.md: Adjust.
28152 * config/rs6000/8540.md: Adjust.
28153 * config/rs6000/cell.md: Adjust.
28154 * config/rs6000/e300c2c3.md: Adjust.
28155 * config/rs6000/e500mc.md: Adjust.
28156 * config/rs6000/e500mc64.md: Adjust.
28157 * config/rs6000/e5500.md: Adjust.
28158 * config/rs6000/e6500.md: Adjust.
28159 * config/rs6000/mpc.md: Adjust.
28160 * config/rs6000/power4.md: Adjust.
28161 * config/rs6000/power5.md: Adjust.
28162 * config/rs6000/power6.md: Adjust.
28163 * config/rs6000/power7.md: Adjust.
28164 * config/rs6000/power8.md: Adjust.
28165 * config/rs6000/power9.md: Adjust.
28166 * config/rs6000/rs64.md: Adjust.
28167 * config/rs6000/titan.md: Adjust.
28169 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
28171 * config/i386/predicates.md (indirect_branch_operand): Rewrite
28172 ix86_indirect_branch_register logic.
28174 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
28176 * config/i386/constraints.md (Bs): Update
28177 ix86_indirect_branch_register check. Don't check
28178 ix86_indirect_branch_register with GOT_memory_operand.
28180 * config/i386/predicates.md (GOT_memory_operand): Don't check
28181 ix86_indirect_branch_register here.
28182 (GOT32_symbol_operand): Likewise.
28184 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
28186 * config/i386/predicates.md (constant_call_address_operand):
28187 Rewrite ix86_indirect_branch_register logic.
28188 (sibcall_insn_operand): Likewise.
28190 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
28192 * config/i386/constraints.md (Bs): Replace
28193 ix86_indirect_branch_thunk_register with
28194 ix86_indirect_branch_register.
28196 * config/i386/i386.md (indirect_jump): Likewise.
28197 (tablejump): Likewise.
28198 (*sibcall_memory): Likewise.
28199 (*sibcall_value_memory): Likewise.
28200 Peepholes of indirect call and jump via memory: Likewise.
28201 * config/i386/i386.opt: Likewise.
28202 * config/i386/predicates.md (indirect_branch_operand): Likewise.
28203 (GOT_memory_operand): Likewise.
28204 (call_insn_operand): Likewise.
28205 (sibcall_insn_operand): Likewise.
28206 (GOT32_symbol_operand): Likewise.
28208 2018-01-15 Jakub Jelinek <jakub@redhat.com>
28210 PR middle-end/83837
28211 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
28212 type rather than type addr's type points to.
28213 (expand_omp_atomic_mutex): Likewise.
28214 (expand_omp_atomic): Likewise.
28216 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
28219 * config/i386/i386.c (output_indirect_thunk_function): Use
28220 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
28221 for __x86_return_thunk.
28223 2018-01-15 Richard Biener <rguenther@suse.de>
28225 PR middle-end/83850
28226 * expmed.c (extract_bit_field_1): Fix typo.
28228 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
28231 * config/arm/iterators.md (VF): New mode iterator.
28232 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
28233 Remove integer-related logic from pattern.
28234 (neon_vabd<mode>_3): Likewise.
28236 2018-01-15 Jakub Jelinek <jakub@redhat.com>
28238 PR middle-end/82694
28239 * common.opt (fstrict-overflow): No longer an alias.
28240 (fwrapv-pointer): New option.
28241 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
28242 also for pointer types based on flag_wrapv_pointer.
28243 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
28244 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
28245 opts->x_flag_wrapv got set.
28246 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
28247 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
28248 POINTER_TYPE_OVERFLOW_UNDEFINED.
28249 * match.pd: Likewise in address comparison pattern.
28250 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
28252 2018-01-15 Richard Biener <rguenther@suse.de>
28255 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
28256 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
28257 Reset type names to their identifier if their TYPE_DECL doesn't
28258 have linkage (and thus is used for ODR and devirt).
28259 (save_debug_info_for_decl): Remove.
28260 (save_debug_info_for_type): Likewise.
28261 (add_tree_to_fld_list): Adjust.
28262 * tree-pretty-print.c (dump_generic_node): Make dumping of
28263 type names more robust.
28265 2018-01-15 Richard Biener <rguenther@suse.de>
28267 * BASE-VER: Bump to 8.0.1.
28269 2018-01-14 Martin Sebor <msebor@redhat.com>
28272 * builtins.c (check_access): Avoid warning when the no-warning bit
28275 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
28277 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
28278 * ira-color (allocno_hard_regs_compare): Likewise.
28280 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
28283 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
28284 Use .pushsection/.popsection.
28286 2018-01-14 Martin Sebor <msebor@redhat.com>
28289 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
28291 2018-01-14 Jakub Jelinek <jakub@redhat.com>
28293 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
28294 entry from extra_headers.
28295 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
28296 extra_headers, make the list bitwise identical to the i?86-*-* one.
28298 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
28300 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
28301 -mcmodel=large with -mindirect-branch=thunk,
28302 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
28303 -mfunction-return=thunk-extern.
28304 * doc/invoke.texi: Document -mcmodel=large is incompatible with
28305 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
28306 -mfunction-return=thunk and -mfunction-return=thunk-extern.
28308 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
28310 * config/i386/i386.c (print_reg): Print the name of the full
28311 integer register without '%'.
28312 (ix86_print_operand): Handle 'V'.
28313 * doc/extend.texi: Document 'V' modifier.
28315 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
28317 * config/i386/constraints.md (Bs): Disallow memory operand for
28318 -mindirect-branch-register.
28320 * config/i386/predicates.md (indirect_branch_operand): Likewise.
28321 (GOT_memory_operand): Likewise.
28322 (call_insn_operand): Likewise.
28323 (sibcall_insn_operand): Likewise.
28324 (GOT32_symbol_operand): Likewise.
28325 * config/i386/i386.md (indirect_jump): Call convert_memory_address
28326 for -mindirect-branch-register.
28327 (tablejump): Likewise.
28328 (*sibcall_memory): Likewise.
28329 (*sibcall_value_memory): Likewise.
28330 Disallow peepholes of indirect call and jump via memory for
28331 -mindirect-branch-register.
28332 (*call_pop): Replace m with Bw.
28333 (*call_value_pop): Likewise.
28334 (*sibcall_pop_memory): Replace m with Bs.
28335 * config/i386/i386.opt (mindirect-branch-register): New option.
28336 * doc/invoke.texi: Document -mindirect-branch-register option.
28338 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
28340 * config/i386/i386-protos.h (ix86_output_function_return): New.
28341 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
28342 set function_return_type.
28343 (indirect_thunk_name): Add ret_p to indicate thunk for function
28345 (output_indirect_thunk_function): Pass false to
28346 indirect_thunk_name.
28347 (ix86_output_indirect_branch_via_reg): Likewise.
28348 (ix86_output_indirect_branch_via_push): Likewise.
28349 (output_indirect_thunk_function): Create alias for function
28350 return thunk if regno < 0.
28351 (ix86_output_function_return): New function.
28352 (ix86_handle_fndecl_attribute): Handle function_return.
28353 (ix86_attribute_table): Add function_return.
28354 * config/i386/i386.h (machine_function): Add
28355 function_return_type.
28356 * config/i386/i386.md (simple_return_internal): Use
28357 ix86_output_function_return.
28358 (simple_return_internal_long): Likewise.
28359 * config/i386/i386.opt (mfunction-return=): New option.
28360 (indirect_branch): Mention -mfunction-return=.
28361 * doc/extend.texi: Document function_return function attribute.
28362 * doc/invoke.texi: Document -mfunction-return= option.
28364 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
28366 * config/i386/i386-opts.h (indirect_branch): New.
28367 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
28368 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
28369 with local indirect jump when converting indirect call and jump.
28370 (ix86_set_indirect_branch_type): New.
28371 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
28372 (indirectlabelno): New.
28373 (indirect_thunk_needed): Likewise.
28374 (indirect_thunk_bnd_needed): Likewise.
28375 (indirect_thunks_used): Likewise.
28376 (indirect_thunks_bnd_used): Likewise.
28377 (INDIRECT_LABEL): Likewise.
28378 (indirect_thunk_name): Likewise.
28379 (output_indirect_thunk): Likewise.
28380 (output_indirect_thunk_function): Likewise.
28381 (ix86_output_indirect_branch_via_reg): Likewise.
28382 (ix86_output_indirect_branch_via_push): Likewise.
28383 (ix86_output_indirect_branch): Likewise.
28384 (ix86_output_indirect_jmp): Likewise.
28385 (ix86_code_end): Call output_indirect_thunk_function if needed.
28386 (ix86_output_call_insn): Call ix86_output_indirect_branch if
28388 (ix86_handle_fndecl_attribute): Handle indirect_branch.
28389 (ix86_attribute_table): Add indirect_branch.
28390 * config/i386/i386.h (machine_function): Add indirect_branch_type
28391 and has_local_indirect_jump.
28392 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
28394 (tablejump): Likewise.
28395 (*indirect_jump): Use ix86_output_indirect_jmp.
28396 (*tablejump_1): Likewise.
28397 (simple_return_indirect_internal): Likewise.
28398 * config/i386/i386.opt (mindirect-branch=): New option.
28399 (indirect_branch): New.
28402 (thunk-inline): Likewise.
28403 (thunk-extern): Likewise.
28404 * doc/extend.texi: Document indirect_branch function attribute.
28405 * doc/invoke.texi: Document -mindirect-branch= option.
28407 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
28410 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
28412 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
28414 * ipa-inline.c (want_inline_small_function_p): Return false if
28415 inlining has already failed with CIF_FINAL_ERROR.
28416 (update_caller_keys): Call want_inline_small_function_p before
28418 (update_callee_keys): Likewise.
28420 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
28422 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
28424 (rs6000_quadword_masked_address_p): Likewise.
28425 (quad_aligned_load_p): Likewise.
28426 (quad_aligned_store_p): Likewise.
28427 (const_load_sequence_p): Add comment to describe the outer-most loop.
28428 (mimic_memory_attributes_and_flags): New function.
28429 (rs6000_gen_stvx): Likewise.
28430 (replace_swapped_aligned_store): Likewise.
28431 (rs6000_gen_lvx): Likewise.
28432 (replace_swapped_aligned_load): Likewise.
28433 (replace_swapped_load_constant): Capitalize argument name in
28434 comment describing this function.
28435 (rs6000_analyze_swaps): Add a third pass to search for vector loads
28436 and stores that access quad-word aligned addresses and replace
28437 with stvx or lvx instructions when appropriate.
28438 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
28439 New function prototype.
28440 (rs6000_quadword_masked_address_p): Likewise.
28441 (rs6000_gen_lvx): Likewise.
28442 (rs6000_gen_stvx): Likewise.
28443 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
28444 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
28445 when memory address is aligned.
28446 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
28447 this split to select lvx instruction when memory address is aligned.
28448 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
28449 instruction when memory address is aligned.
28450 (*vsx_le_perm_load_v16qi): Likewise.
28451 (four unnamed splitters): Modify to select the stvx instruction
28452 when memory is aligned.
28454 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
28456 * predict.c (determine_unlikely_bbs): Handle correctly BBs
28457 which appears in the queue multiple times.
28459 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28460 Alan Hayward <alan.hayward@arm.com>
28461 David Sherwood <david.sherwood@arm.com>
28463 * tree-vectorizer.h (vec_lower_bound): New structure.
28464 (_loop_vec_info): Add check_nonzero and lower_bounds.
28465 (LOOP_VINFO_CHECK_NONZERO): New macro.
28466 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
28467 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
28468 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
28469 fields. Make seg_len the distance travelled, not including the
28471 (dr_direction_indicator): Declare.
28472 (dr_zero_step_indicator): Likewise.
28473 (dr_known_forward_stride_p): Likewise.
28474 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
28476 (runtime_alias_check_p): Allow runtime alias checks with
28478 (operator ==): Compare access_size and align.
28479 (prune_runtime_alias_test_list): Rework for new distinction between
28480 the access_size and seg_len.
28481 (create_intersect_range_checks_index): Likewise. Cope with polynomial
28483 (get_segment_min_max): New function.
28484 (create_intersect_range_checks): Use it.
28485 (dr_step_indicator): New function.
28486 (dr_direction_indicator): Likewise.
28487 (dr_zero_step_indicator): Likewise.
28488 (dr_known_forward_stride_p): Likewise.
28489 * tree-loop-distribution.c (data_ref_segment_size): Return
28490 DR_STEP * (niters - 1).
28491 (compute_alias_check_pairs): Update call to the dr_with_seg_len
28493 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
28494 (vect_preserves_scalar_order_p): New function, split out from...
28495 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
28496 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
28497 (vect_vfa_access_size): New function.
28498 (vect_vfa_align): Likewise.
28499 (vect_compile_time_alias): Take access_size_a and access_b arguments.
28500 (dump_lower_bound): New function.
28501 (vect_check_lower_bound): Likewise.
28502 (vect_small_gap_p): Likewise.
28503 (vectorizable_with_step_bound_p): Likewise.
28504 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
28505 depencies if the vectorization factor is 1. Convert the checks
28506 for nonzero steps into checks on the bounds of DR_STEP. Try using
28507 a bunds check for variable steps if the minimum required step is
28508 relatively small. Update calls to the dr_with_seg_len
28509 constructor and to vect_compile_time_alias.
28510 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
28512 (vect_loop_versioning): Call it.
28513 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
28515 (vect_estimate_min_profitable_iters): Account for any bounds checks.
28517 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28518 Alan Hayward <alan.hayward@arm.com>
28519 David Sherwood <david.sherwood@arm.com>
28521 * doc/sourcebuild.texi (vect_scatter_store): Document.
28522 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
28524 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
28526 * genopinit.c (main): Add supports_vec_scatter_store and
28527 supports_vec_scatter_store_cached to target_optabs.
28528 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
28529 IFN_MASK_SCATTER_STORE.
28530 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
28532 * internal-fn.h (internal_store_fn_p): Declare.
28533 (internal_fn_stored_value_index): Likewise.
28534 * internal-fn.c (scatter_store_direct): New macro.
28535 (expand_scatter_store_optab_fn): New function.
28536 (direct_scatter_store_optab_supported_p): New macro.
28537 (internal_store_fn_p): New function.
28538 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
28539 IFN_MASK_SCATTER_STORE.
28540 (internal_fn_mask_index): Likewise.
28541 (internal_fn_stored_value_index): New function.
28542 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
28543 for scatter stores.
28544 * optabs-query.h (supports_vec_scatter_store_p): Declare.
28545 * optabs-query.c (supports_vec_scatter_store_p): New function.
28546 * tree-vectorizer.h (vect_get_store_rhs): Declare.
28547 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
28548 true for scatter stores.
28549 (vect_gather_scatter_fn_p): Handle scatter stores too.
28550 (vect_check_gather_scatter): Consider using scatter stores if
28551 supports_vec_scatter_store_p.
28552 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
28553 scatter stores too.
28554 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28555 internal_fn_stored_value_index.
28556 (check_load_store_masking): Handle scatter stores too.
28557 (vect_get_store_rhs): Make public.
28558 (vectorizable_call): Use internal_store_fn_p.
28559 (vectorizable_store): Handle scatter store internal functions.
28560 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
28561 when deciding whether the end of the group has been reached.
28562 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
28563 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
28564 (mask_scatter_store<mode>): New insns.
28566 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28567 Alan Hayward <alan.hayward@arm.com>
28568 David Sherwood <david.sherwood@arm.com>
28570 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
28571 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
28572 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
28574 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
28575 Use vect_truncate_gather_scatter_offset if we can't treat the
28576 operation as a normal gather load or scatter store.
28577 (get_group_load_store_type): Take the gather_scatter_info
28578 as argument. Try using a gather load or scatter store for
28579 single-element groups.
28580 (get_load_store_type): Update calls to get_group_load_store_type
28581 and vect_use_strided_gather_scatters_p.
28583 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28584 Alan Hayward <alan.hayward@arm.com>
28585 David Sherwood <david.sherwood@arm.com>
28587 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
28588 optional tree argument.
28589 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
28591 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
28592 but continue to use the current value as a fallback.
28593 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
28594 to compare the updates.
28595 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
28596 (get_load_store_type): Use it when handling a strided access.
28597 (vect_get_strided_load_store_ops): New function.
28598 (vect_get_data_ptr_increment): Likewise.
28599 (vectorizable_load): Handle strided gather loads. Always pass
28600 a step to vect_create_data_ref_ptr and bump_vector_ptr.
28602 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28603 Alan Hayward <alan.hayward@arm.com>
28604 David Sherwood <david.sherwood@arm.com>
28606 * doc/md.texi (gather_load@var{m}): Document.
28607 (mask_gather_load@var{m}): Likewise.
28608 * genopinit.c (main): Add supports_vec_gather_load and
28609 supports_vec_gather_load_cached to target_optabs.
28610 * optabs-tree.c (init_tree_optimization_optabs): Use
28611 ggc_cleared_alloc to allocate target_optabs.
28612 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
28613 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
28615 * internal-fn.h (internal_load_fn_p): Declare.
28616 (internal_gather_scatter_fn_p): Likewise.
28617 (internal_fn_mask_index): Likewise.
28618 (internal_gather_scatter_fn_supported_p): Likewise.
28619 * internal-fn.c (gather_load_direct): New macro.
28620 (expand_gather_load_optab_fn): New function.
28621 (direct_gather_load_optab_supported_p): New macro.
28622 (direct_internal_fn_optab): New function.
28623 (internal_load_fn_p): Likewise.
28624 (internal_gather_scatter_fn_p): Likewise.
28625 (internal_fn_mask_index): Likewise.
28626 (internal_gather_scatter_fn_supported_p): Likewise.
28627 * optabs-query.c (supports_at_least_one_mode_p): New function.
28628 (supports_vec_gather_load_p): Likewise.
28629 * optabs-query.h (supports_vec_gather_load_p): Declare.
28630 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
28631 and memory_type field.
28632 (NUM_PATTERNS): Bump to 15.
28633 * tree-vect-data-refs.c: Include internal-fn.h.
28634 (vect_gather_scatter_fn_p): New function.
28635 (vect_describe_gather_scatter_call): Likewise.
28636 (vect_check_gather_scatter): Try using internal functions for
28637 gather loads. Recognize existing calls to a gather load function.
28638 (vect_analyze_data_refs): Consider using gather loads if
28639 supports_vec_gather_load_p.
28640 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
28641 (vect_get_gather_scatter_offset_type): Likewise.
28642 (vect_convert_mask_for_vectype): Likewise.
28643 (vect_add_conversion_to_patterm): Likewise.
28644 (vect_try_gather_scatter_pattern): Likewise.
28645 (vect_recog_gather_scatter_pattern): New pattern recognizer.
28646 (vect_vect_recog_func_ptrs): Add it.
28647 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28648 internal_fn_mask_index and internal_gather_scatter_fn_p.
28649 (check_load_store_masking): Take the gather_scatter_info as an
28650 argument and handle gather loads.
28651 (vect_get_gather_scatter_ops): New function.
28652 (vectorizable_call): Check internal_load_fn_p.
28653 (vectorizable_load): Likewise. Handle gather load internal
28655 (vectorizable_store): Update call to check_load_store_masking.
28656 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
28657 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
28658 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
28659 (aarch64_gather_scale_operand_d): New predicates.
28660 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
28661 (mask_gather_load<mode>): New insns.
28663 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28664 Alan Hayward <alan.hayward@arm.com>
28665 David Sherwood <david.sherwood@arm.com>
28667 * optabs.def (fold_left_plus_optab): New optab.
28668 * doc/md.texi (fold_left_plus_@var{m}): Document.
28669 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
28670 * internal-fn.c (fold_left_direct): Define.
28671 (expand_fold_left_optab_fn): Likewise.
28672 (direct_fold_left_optab_supported_p): Likewise.
28673 * fold-const-call.c (fold_const_fold_left): New function.
28674 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
28675 * tree-parloops.c (valid_reduction_p): New function.
28676 (gather_scalar_reductions): Use it.
28677 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
28678 (vect_finish_replace_stmt): Declare.
28679 * tree-vect-loop.c (fold_left_reduction_fn): New function.
28680 (needs_fold_left_reduction_p): New function, split out from...
28681 (vect_is_simple_reduction): ...here. Accept reductions that
28682 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
28683 (vect_force_simple_reduction): Also store the reduction type in
28684 the assignment's STMT_VINFO_REDUC_TYPE.
28685 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
28686 (merge_with_identity): New function.
28687 (vect_expand_fold_left): Likewise.
28688 (vectorize_fold_left_reduction): Likewise.
28689 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
28690 scalar phi in place for it. Check for target support and reject
28691 cases that would reassociate the operation. Defer the transform
28692 phase to vectorize_fold_left_reduction.
28693 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
28694 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
28695 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
28697 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28699 * tree-if-conv.c (predicate_mem_writes): Remove redundant
28700 call to ifc_temp_var.
28702 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28703 Alan Hayward <alan.hayward@arm.com>
28704 David Sherwood <david.sherwood@arm.com>
28706 * target.def (legitimize_address_displacement): Take the original
28707 offset as a poly_int.
28708 * targhooks.h (default_legitimize_address_displacement): Update
28710 * targhooks.c (default_legitimize_address_displacement): Likewise.
28711 * doc/tm.texi: Regenerate.
28712 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
28713 as an argument, moving assert of ad->disp == ad->disp_term to...
28714 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
28715 Try calling targetm.legitimize_address_displacement before expanding
28716 the address rather than afterwards, and adjust for the new interface.
28717 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
28718 Match the new hook interface. Handle SVE addresses.
28719 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
28720 new hook interface.
28722 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28724 * Makefile.in (OBJS): Add early-remat.o.
28725 * target.def (select_early_remat_modes): New hook.
28726 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
28727 * doc/tm.texi: Regenerate.
28728 * targhooks.h (default_select_early_remat_modes): Declare.
28729 * targhooks.c (default_select_early_remat_modes): New function.
28730 * timevar.def (TV_EARLY_REMAT): New timevar.
28731 * passes.def (pass_early_remat): New pass.
28732 * tree-pass.h (make_pass_early_remat): Declare.
28733 * early-remat.c: New file.
28734 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
28736 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
28738 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28739 Alan Hayward <alan.hayward@arm.com>
28740 David Sherwood <david.sherwood@arm.com>
28742 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
28743 vfm1 with a bound_epilog parameter.
28744 (vect_do_peeling): Update calls accordingly, and move the prologue
28745 call earlier in the function. Treat the base bound_epilog as 0 for
28746 fully-masked loops and retain vf - 1 for other loops. Add 1 to
28747 this base when peeling for gaps.
28748 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
28749 with fully-masked loops.
28750 (vect_estimate_min_profitable_iters): Handle the single peeled
28751 iteration in that case.
28753 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28754 Alan Hayward <alan.hayward@arm.com>
28755 David Sherwood <david.sherwood@arm.com>
28757 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
28758 single-element interleaving even if the size is not a power of 2.
28759 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
28760 accesses for single-element interleaving if the group size is
28763 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28764 Alan Hayward <alan.hayward@arm.com>
28765 David Sherwood <david.sherwood@arm.com>
28767 * doc/md.texi (fold_extract_last_@var{m}): Document.
28768 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
28769 * optabs.def (fold_extract_last_optab): New optab.
28770 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
28771 * internal-fn.c (fold_extract_direct): New macro.
28772 (expand_fold_extract_optab_fn): Likewise.
28773 (direct_fold_extract_optab_supported_p): Likewise.
28774 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
28775 * tree-vect-loop.c (vect_model_reduction_cost): Handle
28776 EXTRACT_LAST_REDUCTION.
28777 (get_initial_def_for_reduction): Do not create an initial vector
28778 for EXTRACT_LAST_REDUCTION reductions.
28779 (vectorizable_reduction): Leave the scalar phi in place for
28780 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
28781 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
28782 epilogue code for EXTRACT_LAST_REDUCTION and defer the
28783 transform phase to vectorizable_condition.
28784 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
28786 (vect_finish_stmt_generation): ...here.
28787 (vect_finish_replace_stmt): New function.
28788 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
28789 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
28791 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
28793 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28794 Alan Hayward <alan.hayward@arm.com>
28795 David Sherwood <david.sherwood@arm.com>
28797 * doc/md.texi (extract_last_@var{m}): Document.
28798 * optabs.def (extract_last_optab): New optab.
28799 * internal-fn.def (EXTRACT_LAST): New internal function.
28800 * internal-fn.c (cond_unary_direct): New macro.
28801 (expand_cond_unary_optab_fn): Likewise.
28802 (direct_cond_unary_optab_supported_p): Likewise.
28803 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
28804 loops using EXTRACT_LAST.
28805 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
28806 (extract_last_<mode>): ...this optab.
28807 (vec_extract<mode><Vel>): Update accordingly.
28809 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28810 Alan Hayward <alan.hayward@arm.com>
28811 David Sherwood <david.sherwood@arm.com>
28813 * target.def (empty_mask_is_expensive): New hook.
28814 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
28815 * doc/tm.texi: Regenerate.
28816 * targhooks.h (default_empty_mask_is_expensive): Declare.
28817 * targhooks.c (default_empty_mask_is_expensive): New function.
28818 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
28819 if the target says that empty masks are expensive.
28820 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
28822 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
28824 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28825 Alan Hayward <alan.hayward@arm.com>
28826 David Sherwood <david.sherwood@arm.com>
28828 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
28829 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
28830 (vect_use_loop_mask_for_alignment_p): New function.
28831 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
28832 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
28833 niters_skip argument. Make sure that the first niters_skip elements
28834 of the first iteration are inactive.
28835 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
28836 Update call to vect_set_loop_masks_directly.
28837 (get_misalign_in_elems): New function, split out from...
28838 (vect_gen_prolog_loop_niters): ...here.
28839 (vect_update_init_of_dr): Take a code argument that specifies whether
28840 the adjustment should be added or subtracted.
28841 (vect_update_init_of_drs): Likewise.
28842 (vect_prepare_for_masked_peels): New function.
28843 (vect_do_peeling): Skip prologue peeling if we're using a mask
28844 instead. Update call to vect_update_inits_of_drs.
28845 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28847 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
28848 alignment. Do not include the number of peeled iterations in
28849 the minimum threshold in that case.
28850 (vectorizable_induction): Adjust the start value down by
28851 LOOP_VINFO_MASK_SKIP_NITERS iterations.
28852 (vect_transform_loop): Call vect_prepare_for_masked_peels.
28853 Take the number of skipped iterations into account when calculating
28855 * tree-vect-stmts.c (vect_gen_while_not): New function.
28857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28858 Alan Hayward <alan.hayward@arm.com>
28859 David Sherwood <david.sherwood@arm.com>
28861 * doc/sourcebuild.texi (vect_fully_masked): Document.
28862 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
28863 default value to 0.
28864 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
28866 (vect_analyze_loop_2): ...here. Don't check the vectorization
28867 factor against the number of loop iterations if the loop is
28870 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28871 Alan Hayward <alan.hayward@arm.com>
28872 David Sherwood <david.sherwood@arm.com>
28874 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
28875 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
28876 (dump_groups): Update accordingly.
28877 (iv_use::mem_type): New member variable.
28878 (address_p): New function.
28879 (record_use): Add a mem_type argument and initialize the new
28881 (record_group_use): Add a mem_type argument. Use address_p.
28882 Remove obsolete null checks of base_object. Update call to record_use.
28883 (find_interesting_uses_op): Update call to record_group_use.
28884 (find_interesting_uses_cond): Likewise.
28885 (find_interesting_uses_address): Likewise.
28886 (get_mem_type_for_internal_fn): New function.
28887 (find_address_like_use): Likewise.
28888 (find_interesting_uses_stmt): Try find_address_like_use before
28889 calling find_interesting_uses_op.
28890 (addr_offset_valid_p): Use the iv mem_type field as the type
28891 of the addressed memory.
28892 (add_autoinc_candidates): Likewise.
28893 (get_address_cost): Likewise.
28894 (split_small_address_groups_p): Use address_p.
28895 (split_address_groups): Likewise.
28896 (add_iv_candidate_for_use): Likewise.
28897 (autoinc_possible_for_pair): Likewise.
28898 (rewrite_groups): Likewise.
28899 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
28900 (determine_group_iv_cost): Update after split of USE_ADDRESS.
28901 (get_alias_ptr_type_for_ptr_address): New function.
28902 (rewrite_use_address): Rewrite address uses in calls that were
28903 identified by find_address_like_use.
28905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28906 Alan Hayward <alan.hayward@arm.com>
28907 David Sherwood <david.sherwood@arm.com>
28909 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
28911 * gimple-expr.h (is_gimple_addressable: Likewise.
28912 * gimple-expr.c (is_gimple_address): Likewise.
28913 * internal-fn.c (expand_call_mem_ref): New function.
28914 (expand_mask_load_optab_fn): Use it.
28915 (expand_mask_store_optab_fn): Likewise.
28917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28918 Alan Hayward <alan.hayward@arm.com>
28919 David Sherwood <david.sherwood@arm.com>
28921 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
28922 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
28923 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
28924 (cond_umax@var{mode}): Document.
28925 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
28926 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
28927 (cond_umin_optab, cond_umax_optab): New optabs.
28928 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
28929 (COND_IOR, COND_XOR): New internal functions.
28930 * internal-fn.h (get_conditional_internal_fn): Declare.
28931 * internal-fn.c (cond_binary_direct): New macro.
28932 (expand_cond_binary_optab_fn): Likewise.
28933 (direct_cond_binary_optab_supported_p): Likewise.
28934 (get_conditional_internal_fn): New function.
28935 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
28936 Cope with reduction statements that are vectorized as calls rather
28938 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
28939 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
28940 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
28941 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
28942 (UNSPEC_COND_EOR): New unspecs.
28943 (optab): Add mappings for them.
28944 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
28945 (sve_int_op, sve_fp_op): New int attributes.
28947 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
28948 Alan Hayward <alan.hayward@arm.com>
28949 David Sherwood <david.sherwood@arm.com>
28951 * optabs.def (while_ult_optab): New optab.
28952 * doc/md.texi (while_ult@var{m}@var{n}): Document.
28953 * internal-fn.def (WHILE_ULT): New internal function.
28954 * internal-fn.h (direct_internal_fn_supported_p): New override
28955 that takes two types as argument.
28956 * internal-fn.c (while_direct): New macro.
28957 (expand_while_optab_fn): New function.
28958 (convert_optab_supported_p): Likewise.
28959 (direct_while_optab_supported_p): New macro.
28960 * wide-int.h (wi::udiv_ceil): New function.
28961 * tree-vectorizer.h (rgroup_masks): New structure.
28962 (vec_loop_masks): New typedef.
28963 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
28964 and fully_masked_p.
28965 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
28966 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
28967 (vect_max_vf): New function.
28968 (slpeel_make_loop_iterate_ntimes): Delete.
28969 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
28970 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
28971 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
28972 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
28973 internal-fn.h, stor-layout.h and optabs-query.h.
28974 (vect_set_loop_mask): New function.
28975 (add_preheader_seq): Likewise.
28976 (add_header_seq): Likewise.
28977 (interleave_supported_p): Likewise.
28978 (vect_maybe_permute_loop_masks): Likewise.
28979 (vect_set_loop_masks_directly): Likewise.
28980 (vect_set_loop_condition_masked): Likewise.
28981 (vect_set_loop_condition_unmasked): New function, split out from
28982 slpeel_make_loop_iterate_ntimes.
28983 (slpeel_make_loop_iterate_ntimes): Rename to..
28984 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
28985 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
28986 (vect_do_peeling): Update call accordingly.
28987 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
28989 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28990 mask_compare_type, can_fully_mask_p and fully_masked_p.
28991 (release_vec_loop_masks): New function.
28992 (_loop_vec_info): Use it to free the loop masks.
28993 (can_produce_all_loop_masks_p): New function.
28994 (vect_get_max_nscalars_per_iter): Likewise.
28995 (vect_verify_full_masking): Likewise.
28996 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
28997 retries, and free the mask rgroups before retrying. Check loop-wide
28998 reasons for disallowing fully-masked loops. Make the final decision
28999 about whether use a fully-masked loop or not.
29000 (vect_estimate_min_profitable_iters): Do not assume that peeling
29001 for the number of iterations will be needed for fully-masked loops.
29002 (vectorizable_reduction): Disable fully-masked loops.
29003 (vectorizable_live_operation): Likewise.
29004 (vect_halve_mask_nunits): New function.
29005 (vect_double_mask_nunits): Likewise.
29006 (vect_record_loop_mask): Likewise.
29007 (vect_get_loop_mask): Likewise.
29008 (vect_transform_loop): Handle the case in which the final loop
29009 iteration might handle a partial vector. Call vect_set_loop_condition
29010 instead of slpeel_make_loop_iterate_ntimes.
29011 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
29012 (check_load_store_masking): New function.
29013 (prepare_load_store_mask): Likewise.
29014 (vectorizable_store): Handle fully-masked loops.
29015 (vectorizable_load): Likewise.
29016 (supportable_widening_operation): Use vect_halve_mask_nunits for
29018 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
29019 (vect_gen_while): New function.
29020 * config/aarch64/aarch64.md (umax<mode>3): New expander.
29021 (aarch64_uqdec<mode>): New insn.
29023 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29024 Alan Hayward <alan.hayward@arm.com>
29025 David Sherwood <david.sherwood@arm.com>
29027 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
29028 (reduc_xor_scal_optab): New optabs.
29029 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
29030 (reduc_xor_scal_@var{m}): Document.
29031 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
29032 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
29033 internal functions.
29034 * fold-const-call.c (fold_const_call): Handle them.
29035 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
29036 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
29037 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
29038 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
29039 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
29040 (UNSPEC_XORV): New unspecs.
29041 (optab): Add entries for them.
29042 (BITWISEV): New int iterator.
29043 (bit_reduc_op): New int attributes.
29045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29046 Alan Hayward <alan.hayward@arm.com>
29047 David Sherwood <david.sherwood@arm.com>
29049 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
29050 * internal-fn.def (VEC_SHL_INSERT): New internal function.
29051 * optabs.def (vec_shl_insert_optab): New optab.
29052 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
29053 (duplicate_and_interleave): Likewise.
29054 * tree-vect-loop.c: Include internal-fn.h.
29055 (neutral_op_for_slp_reduction): New function, split out from
29056 get_initial_defs_for_reduction.
29057 (get_initial_def_for_reduction): Handle option 2 for variable-length
29058 vectors by loading the neutral value into a vector and then shifting
29059 the initial value into element 0.
29060 (get_initial_defs_for_reduction): Replace the code argument with
29061 the neutral value calculated by neutral_op_for_slp_reduction.
29062 Use gimple_build_vector for constant-length vectors.
29063 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
29064 but the first group_size elements have a neutral value.
29065 Use duplicate_and_interleave otherwise.
29066 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
29067 Update call to get_initial_defs_for_reduction. Handle SLP
29068 reductions for variable-length vectors by creating one vector
29069 result for each scalar result, with the elements associated
29070 with other scalar results stubbed out with the neutral value.
29071 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
29072 Require IFN_VEC_SHL_INSERT for double reductions on
29073 variable-length vectors, or SLP reductions that have
29074 a neutral value. Require can_duplicate_and_interleave_p
29075 support for variable-length unchained SLP reductions if there
29076 is no neutral value, such as for MIN/MAX reductions. Also require
29077 the number of vector elements to be a multiple of the number of
29078 SLP statements when doing variable-length unchained SLP reductions.
29079 Update call to vect_create_epilog_for_reduction.
29080 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
29081 and remove initial values.
29082 (duplicate_and_interleave): Make public.
29083 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
29084 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
29086 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29087 Alan Hayward <alan.hayward@arm.com>
29088 David Sherwood <david.sherwood@arm.com>
29090 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
29091 (can_duplicate_and_interleave_p): New function.
29092 (vect_get_and_check_slp_defs): Take the vector of statements
29093 rather than just the current one. Remove excess parentheses.
29094 Restriction rejectinon of vect_constant_def and vect_external_def
29095 for variable-length vectors to boolean types, or types for which
29096 can_duplicate_and_interleave_p is false.
29097 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
29098 (duplicate_and_interleave): New function.
29099 (vect_get_constant_vectors): Use gimple_build_vector for
29100 constant-length vectors and suitable variable-length constant
29101 vectors. Use duplicate_and_interleave for other variable-length
29102 vectors. Don't defer the update when inserting new statements.
29104 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29105 Alan Hayward <alan.hayward@arm.com>
29106 David Sherwood <david.sherwood@arm.com>
29108 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
29109 min_profitable_iters doesn't go negative.
29111 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29112 Alan Hayward <alan.hayward@arm.com>
29113 David Sherwood <david.sherwood@arm.com>
29115 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
29116 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
29117 * optabs.def (vec_mask_load_lanes_optab): New optab.
29118 (vec_mask_store_lanes_optab): Likewise.
29119 * internal-fn.def (MASK_LOAD_LANES): New internal function.
29120 (MASK_STORE_LANES): Likewise.
29121 * internal-fn.c (mask_load_lanes_direct): New macro.
29122 (mask_store_lanes_direct): Likewise.
29123 (expand_mask_load_optab_fn): Handle masked operations.
29124 (expand_mask_load_lanes_optab_fn): New macro.
29125 (expand_mask_store_optab_fn): Handle masked operations.
29126 (expand_mask_store_lanes_optab_fn): New macro.
29127 (direct_mask_load_lanes_optab_supported_p): Likewise.
29128 (direct_mask_store_lanes_optab_supported_p): Likewise.
29129 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
29131 (vect_load_lanes_supported): Likewise.
29132 * tree-vect-data-refs.c (strip_conversion): New function.
29133 (can_group_stmts_p): Likewise.
29134 (vect_analyze_data_ref_accesses): Use it instead of checking
29135 for a pair of assignments.
29136 (vect_store_lanes_supported): Take a masked_p parameter.
29137 (vect_load_lanes_supported): Likewise.
29138 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
29139 vect_store_lanes_supported and vect_load_lanes_supported.
29140 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
29141 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
29142 parameter. Don't allow gaps for masked accesses.
29143 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
29144 and vect_load_lanes_supported.
29145 (get_load_store_type): Take a masked_p parameter and update
29146 call to get_group_load_store_type.
29147 (vectorizable_store): Update call to get_load_store_type.
29148 Handle IFN_MASK_STORE_LANES.
29149 (vectorizable_load): Update call to get_load_store_type.
29150 Handle IFN_MASK_LOAD_LANES.
29152 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29153 Alan Hayward <alan.hayward@arm.com>
29154 David Sherwood <david.sherwood@arm.com>
29156 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
29158 * config/aarch64/aarch64-protos.h
29159 (aarch64_sve_struct_memory_operand_p): Declare.
29160 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
29161 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
29162 (VPRED, vpred): Handle SVE structure modes.
29163 * config/aarch64/constraints.md (Utx): New constraint.
29164 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
29165 (aarch64_sve_struct_nonimmediate_operand): New predicates.
29166 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
29167 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
29168 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
29169 structure modes. Split into pieces after RA.
29170 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
29171 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
29173 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
29174 SVE structure modes.
29175 (aarch64_classify_address): Likewise.
29176 (sizetochar): Move earlier in file.
29177 (aarch64_print_operand): Handle SVE register lists.
29178 (aarch64_array_mode): New function.
29179 (aarch64_sve_struct_memory_operand_p): Likewise.
29180 (TARGET_ARRAY_MODE): Redefine.
29182 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29183 Alan Hayward <alan.hayward@arm.com>
29184 David Sherwood <david.sherwood@arm.com>
29186 * target.def (array_mode): New target hook.
29187 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
29188 * doc/tm.texi: Regenerate.
29189 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
29190 * hooks.c (hook_optmode_mode_uhwi_none): New function.
29191 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
29192 targetm.array_mode.
29193 * stor-layout.c (mode_for_array): Likewise. Support polynomial
29196 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29197 Alan Hayward <alan.hayward@arm.com>
29198 David Sherwood <david.sherwood@arm.com>
29200 * fold-const.c (fold_binary_loc): Check the argument types
29201 rather than the result type when testing for a vector operation.
29203 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29205 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
29206 * doc/tm.texi: Regenerate.
29208 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29209 Alan Hayward <alan.hayward@arm.com>
29210 David Sherwood <david.sherwood@arm.com>
29212 * doc/invoke.texi (-msve-vector-bits=): Document new option.
29213 (sve): Document new AArch64 extension.
29214 * doc/md.texi (w): Extend the description of the AArch64
29215 constraint to include SVE vectors.
29216 (Upl, Upa): Document new AArch64 predicate constraints.
29217 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
29219 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
29220 (msve-vector-bits=): New option.
29221 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
29222 SVE when these are disabled.
29223 (sve): New extension.
29224 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
29225 modes. Adjust their number of units based on aarch64_sve_vg.
29226 (MAX_BITSIZE_MODE_ANY_MODE): Define.
29227 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
29228 aarch64_addr_query_type.
29229 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
29230 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
29231 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
29232 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
29233 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
29234 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
29235 (aarch64_simd_imm_zero_p): Delete.
29236 (aarch64_check_zero_based_sve_index_immediate): Declare.
29237 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
29238 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
29239 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
29240 (aarch64_sve_float_mul_immediate_p): Likewise.
29241 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
29242 rather than an rtx.
29243 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
29244 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
29245 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
29246 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
29247 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
29248 (aarch64_regmode_natural_size): Likewise.
29249 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
29250 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
29252 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
29253 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
29254 for VG and the SVE predicate registers.
29255 (V_ALIASES): Add a "z"-prefixed alias.
29256 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
29257 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
29258 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
29259 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
29260 (REG_CLASS_NAMES): Add entries for them.
29261 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
29262 and the predicate registers.
29263 (aarch64_sve_vg): Declare.
29264 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
29265 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
29266 (REGMODE_NATURAL_SIZE): Define.
29267 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
29269 * config/aarch64/aarch64.c: Include cfgrtl.h.
29270 (simd_immediate_info): Add a constructor for series vectors,
29271 and an associated step field.
29272 (aarch64_sve_vg): New variable.
29273 (aarch64_dbx_register_number): Handle VG and the predicate registers.
29274 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
29275 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
29276 (VEC_ANY_DATA, VEC_STRUCT): New constants.
29277 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
29278 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
29279 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
29280 (aarch64_get_mask_mode): New functions.
29281 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
29282 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
29283 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
29284 predicate modes and predicate registers. Explicitly restrict
29285 GPRs to modes of 16 bytes or smaller. Only allow FP registers
29286 to store a vector mode if it is recognized by
29287 aarch64_classify_vector_mode.
29288 (aarch64_regmode_natural_size): New function.
29289 (aarch64_hard_regno_caller_save_mode): Return the original mode
29291 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
29292 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
29293 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
29294 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
29296 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
29297 does not overlap dest if the function is frame-related. Handle
29299 (aarch64_split_add_offset): New function.
29300 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
29301 them aarch64_add_offset.
29302 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
29303 and update call to aarch64_sub_sp.
29304 (aarch64_add_cfa_expression): New function.
29305 (aarch64_expand_prologue): Pass extra temporary registers to the
29306 functions above. Handle the case in which we need to emit new
29307 DW_CFA_expressions for registers that were originally saved
29308 relative to the stack pointer, but now have to be expressed
29309 relative to the frame pointer.
29310 (aarch64_output_mi_thunk): Pass extra temporary registers to the
29312 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
29313 IP0 and IP1 values for SVE frames.
29314 (aarch64_expand_vec_series): New function.
29315 (aarch64_expand_sve_widened_duplicate): Likewise.
29316 (aarch64_expand_sve_const_vector): Likewise.
29317 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
29318 Handle SVE constants. Use emit_move_insn to move a force_const_mem
29319 into the register, rather than emitting a SET directly.
29320 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
29321 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
29322 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
29323 (offset_9bit_signed_scaled_p): New functions.
29324 (aarch64_replicate_bitmask_imm): New function.
29325 (aarch64_bitmask_imm): Use it.
29326 (aarch64_cannot_force_const_mem): Reject expressions involving
29327 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
29328 (aarch64_classify_index): Handle SVE indices, by requiring
29329 a plain register index with a scale that matches the element size.
29330 (aarch64_classify_address): Handle SVE addresses. Assert that
29331 the mode of the address is VOIDmode or an integer mode.
29332 Update call to aarch64_classify_symbol.
29333 (aarch64_classify_symbolic_expression): Update call to
29334 aarch64_classify_symbol.
29335 (aarch64_const_vec_all_in_range_p): New function.
29336 (aarch64_print_vector_float_operand): Likewise.
29337 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
29338 "vN" for FP registers with SVE modes. Handle (const ...) vectors
29339 and the FP immediates 1.0 and 0.5.
29340 (aarch64_print_address_internal): Handle SVE addresses.
29341 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
29342 (aarch64_regno_regclass): Handle predicate registers.
29343 (aarch64_secondary_reload): Handle big-endian reloads of SVE
29345 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
29346 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
29347 (aarch64_convert_sve_vector_bits): New function.
29348 (aarch64_override_options): Use it to handle -msve-vector-bits=.
29349 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
29350 rather than an rtx.
29351 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
29352 Handle SVE vector and predicate modes. Accept VL-based constants
29353 that need only one temporary register, and VL offsets that require
29354 no temporary registers.
29355 (aarch64_conditional_register_usage): Mark the predicate registers
29356 as fixed if SVE isn't available.
29357 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
29358 Return true for SVE vector and predicate modes.
29359 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
29360 rather than an unsigned int. Handle SVE modes.
29361 (aarch64_preferred_simd_mode): Update call accordingly. Handle
29363 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
29365 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
29366 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
29367 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
29368 (aarch64_sve_float_mul_immediate_p): New functions.
29369 (aarch64_sve_valid_immediate): New function.
29370 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
29371 Explicitly reject structure modes. Check for INDEX constants.
29372 Handle PTRUE and PFALSE constants.
29373 (aarch64_check_zero_based_sve_index_immediate): New function.
29374 (aarch64_simd_imm_zero_p): Delete.
29375 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
29376 vector modes. Accept constants in the range of CNT[BHWD].
29377 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
29378 ask for an Advanced SIMD mode.
29379 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
29380 (aarch64_simd_vector_alignment): Handle SVE predicates.
29381 (aarch64_vectorize_preferred_vector_alignment): New function.
29382 (aarch64_simd_vector_alignment_reachable): Use it instead of
29384 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
29385 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
29387 (MAX_VECT_LEN): Delete.
29388 (expand_vec_perm_d): Add a vec_flags field.
29389 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
29390 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
29391 (aarch64_evpc_ext): Don't apply a big-endian lane correction
29393 (aarch64_evpc_rev): Rename to...
29394 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
29395 (aarch64_evpc_rev_global): New function.
29396 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
29397 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
29399 (aarch64_evpc_sve_tbl): New function.
29400 (aarch64_expand_vec_perm_const_1): Update after rename of
29401 aarch64_evpc_rev. Handle SVE permutes too, trying
29402 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
29403 than aarch64_evpc_tbl.
29404 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
29405 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
29406 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
29407 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
29408 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
29409 (aarch64_expand_sve_vcond): New functions.
29410 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
29411 of aarch64_vector_mode_p.
29412 (aarch64_dwarf_poly_indeterminate_value): New function.
29413 (aarch64_compute_pressure_classes): Likewise.
29414 (aarch64_can_change_mode_class): Likewise.
29415 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
29416 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
29417 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
29418 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
29419 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
29420 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
29421 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
29422 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
29424 (Dn, Dl, Dr): Accept const as well as const_vector.
29425 (Dz): Likewise. Compare against CONST0_RTX.
29426 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
29427 of "vector" where appropriate.
29428 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
29429 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
29430 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
29431 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
29432 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
29433 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
29434 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
29435 (v_int_equiv): Extend to SVE modes.
29436 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
29438 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
29439 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
29440 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
29441 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
29442 (SVE_COND_FP_CMP): New int iterators.
29443 (perm_hilo): Handle the new unpack unspecs.
29444 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
29446 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
29447 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
29448 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
29449 (aarch64_equality_operator, aarch64_constant_vector_operand)
29450 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
29451 (aarch64_sve_nonimmediate_operand): Likewise.
29452 (aarch64_sve_general_operand): Likewise.
29453 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
29454 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
29455 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
29456 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
29457 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
29458 (aarch64_sve_float_arith_immediate): Likewise.
29459 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
29460 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
29461 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
29462 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
29463 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
29464 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
29465 (aarch64_sve_float_arith_operand): Likewise.
29466 (aarch64_sve_float_arith_with_sub_operand): Likewise.
29467 (aarch64_sve_float_mul_operand): Likewise.
29468 (aarch64_sve_vec_perm_operand): Likewise.
29469 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
29470 (aarch64_mov_operand): Accept const_poly_int and const_vector.
29471 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
29472 as well as const_vector.
29473 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
29474 in file. Use CONST0_RTX and CONSTM1_RTX.
29475 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
29476 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
29477 Use aarch64_simd_imm_zero.
29478 * config/aarch64/aarch64-sve.md: New file.
29479 * config/aarch64/aarch64.md: Include it.
29480 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
29481 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
29482 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
29483 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
29484 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
29485 (sve): New attribute.
29486 (enabled): Disable instructions with the sve attribute unless
29488 (movqi, movhi): Pass CONST_POLY_INT operaneds through
29489 aarch64_expand_mov_immediate.
29490 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
29491 CNT[BHSD] immediates.
29492 (movti): Split CONST_POLY_INT moves into two halves.
29493 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
29494 Split additions that need a temporary here if the destination
29495 is the stack pointer.
29496 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
29497 (*add<mode>3_poly_1): New instruction.
29498 (set_clobber_cc): New expander.
29500 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29502 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
29503 parameter and use it instead of GET_MODE_SIZE (innermode). Use
29504 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
29505 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
29506 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
29507 Change innermode from fixed_mode_size to machine_mode.
29508 (simplify_subreg): Update call accordingly. Handle a constant-sized
29509 subreg of a variable-length CONST_VECTOR.
29511 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
29512 Alan Hayward <alan.hayward@arm.com>
29513 David Sherwood <david.sherwood@arm.com>
29515 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
29516 (add_offset_to_base): New function, split out from...
29517 (create_mem_ref): ...here. When handling a scale other than 1,
29518 check first whether the address is valid without the offset.
29519 Add it into the base if so, leaving the index and scale as-is.
29521 2018-01-12 Jakub Jelinek <jakub@redhat.com>
29524 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
29525 fold_for_warn before checking if arg2 is INTEGER_CST.
29527 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
29529 * config/rs6000/predicates.md (load_multiple_operation): Delete.
29530 (store_multiple_operation): Delete.
29531 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
29532 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
29533 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
29534 guarded by TARGET_STRING.
29535 (rs6000_output_load_multiple): Delete.
29536 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29537 OPTION_MASK_STRING / TARGET_STRING handling.
29538 (print_operand) <'N', 'O'>: Add comment that these are unused now.
29539 (const rs6000_opt_masks) <"string">: Change mask to 0.
29540 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
29541 (MASK_STRING): Delete.
29542 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
29544 (load_multiple): Delete.
29551 (store_multiple): Delete.
29558 (movmemsi_8reg): Delete.
29559 (corresponding unnamed define_insn): Delete.
29560 (movmemsi_6reg): Delete.
29561 (corresponding unnamed define_insn): Delete.
29562 (movmemsi_4reg): Delete.
29563 (corresponding unnamed define_insn): Delete.
29564 (movmemsi_2reg): Delete.
29565 (corresponding unnamed define_insn): Delete.
29566 (movmemsi_1reg): Delete.
29567 (corresponding unnamed define_insn): Delete.
29568 * config/rs6000/rs6000.opt (mno-string): New.
29569 (mstring): Replace by deprecation warning stub.
29570 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
29572 2018-01-12 Jakub Jelinek <jakub@redhat.com>
29574 * regrename.c (regrename_do_replace): If replacing the same
29575 reg multiple times, try to reuse last created gen_raw_REG.
29578 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
29579 main to workaround a bug in GDB.
29581 2018-01-12 Tom de Vries <tom@codesourcery.com>
29584 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
29586 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
29588 PR rtl-optimization/80481
29589 * ira-color.c (get_cap_member): New function.
29590 (allocnos_conflict_by_live_ranges_p): Use it.
29591 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
29592 (setup_slot_coalesced_allocno_live_ranges): Ditto.
29594 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
29597 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
29598 (*saddl_se_1): Ditto.
29599 (*ssubsi_1): Ditto.
29600 (*ssubl_se_1): Ditto.
29602 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
29604 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
29605 rather than wi::to_widest for DR_INITs.
29606 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
29607 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
29608 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
29610 (vect_analyze_group_access_1): Note that here.
29612 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
29614 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
29615 polynomial type sizes.
29617 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
29619 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
29620 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
29621 (gimple_add_tmp_var): Likewise.
29623 2018-01-12 Martin Liska <mliska@suse.cz>
29625 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
29626 (gimple_alloc_sizes): Likewise.
29627 (dump_gimple_statistics): Use PRIu64 in printf format.
29628 * gimple.h: Change uint64_t to int.
29630 2018-01-12 Martin Liska <mliska@suse.cz>
29632 * tree-core.h: Use uint64_t instead of int.
29633 * tree.c (tree_node_counts): Likewise.
29634 (tree_node_sizes): Likewise.
29635 (dump_tree_statistics): Use PRIu64 in printf format.
29637 2018-01-12 Martin Liska <mliska@suse.cz>
29639 * Makefile.in: As qsort_chk is implemented in vec.c, add
29640 vec.o to linkage of gencfn-macros.
29641 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
29642 passing the info to record_node_allocation_statistics.
29643 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
29645 * ggc-common.c (struct ggc_usage): Add operator== and use
29646 it in operator< and compare function.
29647 * mem-stats.h (struct mem_usage): Likewise.
29648 * vec.c (struct vec_usage): Remove operator< and compare
29649 function. Can be simply inherited.
29651 2018-01-12 Martin Jambor <mjambor@suse.cz>
29654 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
29655 * tree-ssa-math-opts.c: Include domwalk.h.
29656 (convert_mult_to_fma_1): New function.
29657 (fma_transformation_info): New type.
29658 (fma_deferring_state): Likewise.
29659 (cancel_fma_deferring): New function.
29660 (result_of_phi): Likewise.
29661 (last_fma_candidate_feeds_initial_phi): Likewise.
29662 (convert_mult_to_fma): Added deferring logic, split actual
29663 transformation to convert_mult_to_fma_1.
29664 (math_opts_dom_walker): New type.
29665 (math_opts_dom_walker::after_dom_children): New method, body moved
29666 here from pass_optimize_widening_mul::execute, added deferring logic
29668 (pass_optimize_widening_mul::execute): Moved most of code to
29669 math_opts_dom_walker::after_dom_children.
29670 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
29671 * config/i386/i386.c (ix86_option_override_internal): Added
29672 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
29674 2018-01-12 Richard Biener <rguenther@suse.de>
29677 * dwarf2out.c (gen_variable_die): Do not reset old_die for
29678 inline instance vars.
29680 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
29683 * config/rx/rx.c (rx_is_restricted_memory_address):
29684 Handle SUBREG case.
29686 2018-01-12 Richard Biener <rguenther@suse.de>
29688 PR tree-optimization/80846
29689 * target.def (split_reduction): New target hook.
29690 * targhooks.c (default_split_reduction): New function.
29691 * targhooks.h (default_split_reduction): Declare.
29692 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
29693 target requests first reduce vectors by combining low and high
29695 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
29696 (get_vectype_for_scalar_type_and_size): Export.
29697 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
29698 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
29699 * doc/tm.texi: Regenerate.
29700 * config/i386/i386.c (ix86_split_reduction): Implement
29701 TARGET_VECTORIZE_SPLIT_REDUCTION.
29703 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
29706 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
29707 in PIC mode except for TARGET_VXWORKS_RTP.
29708 * config/sparc/sparc.c: Include cfgrtl.h.
29709 (TARGET_INIT_PIC_REG): Define.
29710 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
29711 (sparc_pic_register_p): New predicate.
29712 (sparc_legitimate_address_p): Use it.
29713 (sparc_legitimize_pic_address): Likewise.
29714 (sparc_delegitimize_address): Likewise.
29715 (sparc_mode_dependent_address_p): Likewise.
29716 (gen_load_pcrel_sym): Remove 4th parameter.
29717 (load_got_register): Adjust call to above. Remove obsolete stuff.
29718 (sparc_expand_prologue): Do not call load_got_register here.
29719 (sparc_flat_expand_prologue): Likewise.
29720 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
29721 (sparc_use_pseudo_pic_reg): New function.
29722 (sparc_init_pic_reg): Likewise.
29723 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
29724 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
29726 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
29728 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
29729 Add item for branch_cost.
29731 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
29733 PR rtl-optimization/83565
29734 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
29735 not extend the result to a larger mode for rotate operations.
29736 (num_sign_bit_copies1): Likewise.
29738 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
29741 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
29743 Use values-Xc.o for -pedantic.
29744 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
29746 2018-01-12 Martin Liska <mliska@suse.cz>
29749 * ipa-devirt.c (final_warning_record::grow_type_warnings):
29751 (possible_polymorphic_call_targets): Use it.
29752 (ipa_devirt): Likewise.
29754 2018-01-12 Martin Liska <mliska@suse.cz>
29756 * profile-count.h (enum profile_quality): Use 0 as invalid
29757 enum value of profile_quality.
29759 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
29761 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
29762 -mext-string options.
29764 2018-01-12 Richard Biener <rguenther@suse.de>
29766 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
29767 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
29768 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
29770 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
29772 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
29774 * configure.ac (--with-long-double-format): Add support for the
29775 configuration option to change the default long double format on
29777 * config.gcc (powerpc*-linux*-*): Likewise.
29778 * configure: Regenerate.
29779 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
29780 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
29781 used without modification.
29783 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
29785 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
29786 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
29787 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
29788 MISC_BUILTIN_SPEC_BARRIER.
29789 (rs6000_init_builtins): Likewise.
29790 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
29792 (speculation_barrier): New define_insn.
29793 * doc/extend.texi: Document __builtin_speculation_barrier.
29795 2018-01-11 Jakub Jelinek <jakub@redhat.com>
29798 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
29799 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
29800 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
29802 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
29803 integral modes instead of "ss" and "sd".
29804 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
29805 vectors with 32-bit and 64-bit elements.
29806 (vecdupssescalarmodesuffix): New mode attribute.
29807 (vec_dup<mode>): Use it.
29809 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
29812 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
29813 frame if argument is passed on stack.
29815 2018-01-11 Jakub Jelinek <jakub@redhat.com>
29818 * ree.c (combine_reaching_defs): Optimize also
29819 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
29820 reg2=any_extend(exp); reg1=reg2;, formatting fix.
29822 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
29824 PR middle-end/83189
29825 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
29827 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
29829 PR middle-end/83718
29830 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
29831 after they are computed.
29833 2018-01-11 Bin Cheng <bin.cheng@arm.com>
29835 PR tree-optimization/83695
29836 * gimple-loop-linterchange.cc
29837 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
29838 reset cached scev information after interchange.
29839 (pass_linterchange::execute): Remove call to scev_reset_htab.
29841 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29843 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
29844 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
29845 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
29846 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
29847 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
29848 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
29849 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
29850 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
29851 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
29852 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
29853 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
29854 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
29855 (V_lane_reg): Likewise.
29856 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
29858 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
29859 (vfmal_lane_low<mode>_intrinsic,
29860 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
29861 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
29862 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
29863 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
29864 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
29865 vfmsl_lane_high<mode>_intrinsic): New define_insns.
29867 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29869 * config/arm/arm-cpus.in (fp16fml): New feature.
29870 (ALL_SIMD): Add fp16fml.
29871 (armv8.2-a): Add fp16fml as an option.
29872 (armv8.3-a): Likewise.
29873 (armv8.4-a): Add fp16fml as part of fp16.
29874 * config/arm/arm.h (TARGET_FP16FML): Define.
29875 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
29877 * config/arm/arm-modes.def (V2HF): Define.
29878 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
29879 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
29880 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
29881 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
29882 vfmsl_low, vfmsl_high): New set of builtins.
29883 * config/arm/iterators.md (PLUSMINUS): New code iterator.
29884 (vfml_op): New code attribute.
29885 (VFMLHALVES): New int iterator.
29886 (VFML, VFMLSEL): New mode attributes.
29887 (V_reg): Define mapping for V2HF.
29888 (V_hi, V_lo): New mode attributes.
29889 (VF_constraint): Likewise.
29890 (vfml_half, vfml_half_selector): New int attributes.
29891 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
29893 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
29894 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
29896 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
29897 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
29898 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
29899 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
29901 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
29902 Document new effective target and option set.
29904 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
29906 * config/arm/arm-cpus.in (armv8_4): New feature.
29907 (ARMv8_4a): New fgroup.
29908 (armv8.4-a): New arch.
29909 * config/arm/arm-tables.opt: Regenerate.
29910 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
29911 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
29912 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
29913 Add matching rules for -march=armv8.4-a and extensions.
29914 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
29916 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
29919 * config/rx/rx.md (BW): New mode attribute.
29920 (sync_lock_test_and_setsi): Add mode suffix to insn output.
29922 2018-01-11 Richard Biener <rguenther@suse.de>
29924 PR tree-optimization/83435
29925 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
29926 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
29927 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
29929 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
29930 Alan Hayward <alan.hayward@arm.com>
29931 David Sherwood <david.sherwood@arm.com>
29933 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
29935 (aarch64_classify_address): Initialize it. Track polynomial offsets.
29936 (aarch64_print_address_internal): Use it to check for a zero offset.
29938 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
29939 Alan Hayward <alan.hayward@arm.com>
29940 David Sherwood <david.sherwood@arm.com>
29942 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
29943 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
29944 Return a poly_int64 rather than a HOST_WIDE_INT.
29945 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
29946 rather than a HOST_WIDE_INT.
29947 * config/aarch64/aarch64.h (aarch64_frame): Protect with
29948 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
29949 hard_fp_offset, frame_size, initial_adjust, callee_offset and
29950 final_offset from HOST_WIDE_INT to poly_int64.
29951 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
29952 to_constant when getting the number of units in an Advanced SIMD
29954 (aarch64_builtin_vectorized_function): Check for a constant number
29956 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
29958 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
29959 attribute instead of GET_MODE_NUNITS.
29960 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
29961 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
29962 GET_MODE_SIZE for fixed-size registers.
29963 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
29964 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
29965 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
29966 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
29967 (aarch64_print_operand, aarch64_print_address_internal)
29968 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
29969 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
29970 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
29971 Handle polynomial GET_MODE_SIZE.
29972 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
29973 wider than SImode without modification.
29974 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
29975 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
29976 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
29977 passing and returning SVE modes.
29978 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
29979 rather than GEN_INT.
29980 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
29981 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
29982 (aarch64_allocate_and_probe_stack_space): Likewise.
29983 (aarch64_layout_frame): Cope with polynomial offsets.
29984 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
29985 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
29986 polynomial offsets.
29987 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
29988 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
29989 poly_int64 rather than a HOST_WIDE_INT.
29990 (aarch64_get_separate_components, aarch64_process_components)
29991 (aarch64_expand_prologue, aarch64_expand_epilogue)
29992 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
29993 (aarch64_anchor_offset): New function, split out from...
29994 (aarch64_legitimize_address): ...here.
29995 (aarch64_builtin_vectorization_cost): Handle polynomial
29996 TYPE_VECTOR_SUBPARTS.
29997 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
29999 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
30000 number of elements from the PARALLEL rather than the mode.
30001 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
30002 rather than GET_MODE_BITSIZE.
30003 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
30004 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
30005 (aarch64_expand_vec_perm_const_1): Handle polynomial
30006 d->perm.length () and d->perm elements.
30007 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
30008 Apply to_constant to d->perm elements.
30009 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
30010 polynomial CONST_VECTOR_NUNITS.
30011 (aarch64_move_pointer): Take amount as a poly_int64 rather
30013 (aarch64_progress_pointer): Avoid temporary variable.
30014 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
30015 the mode attribute instead of GET_MODE.
30017 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
30018 Alan Hayward <alan.hayward@arm.com>
30019 David Sherwood <david.sherwood@arm.com>
30021 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
30022 x exists before using it.
30023 (aarch64_add_constant_internal): Rename to...
30024 (aarch64_add_offset_1): ...this. Replace regnum with separate
30025 src and dest rtxes. Handle the case in which they're different,
30026 including when the offset is zero. Replace scratchreg with an rtx.
30027 Use 2 additions if there is no spare register into which we can
30028 move a 16-bit constant.
30029 (aarch64_add_constant): Delete.
30030 (aarch64_add_offset): Replace reg with separate src and dest
30031 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
30032 Use aarch64_add_offset_1.
30033 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
30034 an rtx rather than an int. Take the delta as a poly_int64
30035 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
30036 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
30037 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
30038 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
30039 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
30040 and aarch64_add_sp.
30041 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
30042 aarch64_add_constant.
30044 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
30046 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
30047 Use scalar_float_mode.
30049 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
30051 * config/aarch64/aarch64-simd.md
30052 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
30053 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
30054 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
30055 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
30056 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
30057 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
30058 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
30059 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
30060 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
30061 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
30063 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
30066 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
30067 targ_options->x_arm_arch_string is non NULL.
30069 2018-01-11 Tamar Christina <tamar.christina@arm.com>
30071 * config/aarch64/aarch64.h
30072 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
30074 2018-01-11 Sudakshina Das <sudi.das@arm.com>
30077 * expmed.c (emit_store_flag_force): Swap if const op0
30078 and change VOIDmode to mode of op0.
30080 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
30082 PR rtl-optimization/83761
30083 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
30084 than bytes to mode_for_size.
30086 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
30088 PR middle-end/83189
30089 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
30090 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
30093 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
30095 PR middle-end/83575
30096 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
30097 when in layout mode.
30098 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
30099 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
30102 2018-01-10 Michael Collison <michael.collison@arm.com>
30104 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
30105 * config/aarch64/aarch64-option-extension.def: Add
30106 AARCH64_OPT_EXTENSION of 'fp16fml'.
30107 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30108 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
30109 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
30110 * config/aarch64/constraints.md (Ui7): New constraint.
30111 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
30112 (VFMLA_SEL_W): Ditto.
30115 (VFMLA16_LOW): New int iterator.
30116 (VFMLA16_HIGH): Ditto.
30117 (UNSPEC_FMLAL): New unspec.
30118 (UNSPEC_FMLSL): Ditto.
30119 (UNSPEC_FMLAL2): Ditto.
30120 (UNSPEC_FMLSL2): Ditto.
30121 (f16mac): New code attribute.
30122 * config/aarch64/aarch64-simd-builtins.def
30123 (aarch64_fmlal_lowv2sf): Ditto.
30124 (aarch64_fmlsl_lowv2sf): Ditto.
30125 (aarch64_fmlalq_lowv4sf): Ditto.
30126 (aarch64_fmlslq_lowv4sf): Ditto.
30127 (aarch64_fmlal_highv2sf): Ditto.
30128 (aarch64_fmlsl_highv2sf): Ditto.
30129 (aarch64_fmlalq_highv4sf): Ditto.
30130 (aarch64_fmlslq_highv4sf): Ditto.
30131 (aarch64_fmlal_lane_lowv2sf): Ditto.
30132 (aarch64_fmlsl_lane_lowv2sf): Ditto.
30133 (aarch64_fmlal_laneq_lowv2sf): Ditto.
30134 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
30135 (aarch64_fmlalq_lane_lowv4sf): Ditto.
30136 (aarch64_fmlsl_lane_lowv4sf): Ditto.
30137 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
30138 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
30139 (aarch64_fmlal_lane_highv2sf): Ditto.
30140 (aarch64_fmlsl_lane_highv2sf): Ditto.
30141 (aarch64_fmlal_laneq_highv2sf): Ditto.
30142 (aarch64_fmlsl_laneq_highv2sf): Ditto.
30143 (aarch64_fmlalq_lane_highv4sf): Ditto.
30144 (aarch64_fmlsl_lane_highv4sf): Ditto.
30145 (aarch64_fmlalq_laneq_highv4sf): Ditto.
30146 (aarch64_fmlsl_laneq_highv4sf): Ditto.
30147 * config/aarch64/aarch64-simd.md:
30148 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
30149 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
30150 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
30151 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
30152 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
30153 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
30154 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
30155 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
30156 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
30157 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
30158 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
30159 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
30160 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
30161 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
30162 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
30163 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
30164 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
30165 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
30166 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
30167 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
30168 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
30169 (vfmlsl_low_u32): Ditto.
30170 (vfmlalq_low_u32): Ditto.
30171 (vfmlslq_low_u32): Ditto.
30172 (vfmlal_high_u32): Ditto.
30173 (vfmlsl_high_u32): Ditto.
30174 (vfmlalq_high_u32): Ditto.
30175 (vfmlslq_high_u32): Ditto.
30176 (vfmlal_lane_low_u32): Ditto.
30177 (vfmlsl_lane_low_u32): Ditto.
30178 (vfmlal_laneq_low_u32): Ditto.
30179 (vfmlsl_laneq_low_u32): Ditto.
30180 (vfmlalq_lane_low_u32): Ditto.
30181 (vfmlslq_lane_low_u32): Ditto.
30182 (vfmlalq_laneq_low_u32): Ditto.
30183 (vfmlslq_laneq_low_u32): Ditto.
30184 (vfmlal_lane_high_u32): Ditto.
30185 (vfmlsl_lane_high_u32): Ditto.
30186 (vfmlal_laneq_high_u32): Ditto.
30187 (vfmlsl_laneq_high_u32): Ditto.
30188 (vfmlalq_lane_high_u32): Ditto.
30189 (vfmlslq_lane_high_u32): Ditto.
30190 (vfmlalq_laneq_high_u32): Ditto.
30191 (vfmlslq_laneq_high_u32): Ditto.
30192 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
30193 (AARCH64_FL_FOR_ARCH8_4): New.
30194 (AARCH64_ISA_F16FML): New ISA flag.
30195 (TARGET_F16FML): New feature flag for fp16fml.
30196 (doc/invoke.texi): Document new fp16fml option.
30198 2018-01-10 Michael Collison <michael.collison@arm.com>
30200 * config/aarch64/aarch64-builtins.c:
30201 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
30202 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30203 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
30204 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
30205 (AARCH64_ISA_SHA3): New ISA flag.
30206 (TARGET_SHA3): New feature flag for sha3.
30207 * config/aarch64/iterators.md (sha512_op): New int attribute.
30208 (CRYPTO_SHA512): New int iterator.
30209 (UNSPEC_SHA512H): New unspec.
30210 (UNSPEC_SHA512H2): Ditto.
30211 (UNSPEC_SHA512SU0): Ditto.
30212 (UNSPEC_SHA512SU1): Ditto.
30213 * config/aarch64/aarch64-simd-builtins.def
30214 (aarch64_crypto_sha512hqv2di): New builtin.
30215 (aarch64_crypto_sha512h2qv2di): Ditto.
30216 (aarch64_crypto_sha512su0qv2di): Ditto.
30217 (aarch64_crypto_sha512su1qv2di): Ditto.
30218 (aarch64_eor3qv8hi): Ditto.
30219 (aarch64_rax1qv2di): Ditto.
30220 (aarch64_xarqv2di): Ditto.
30221 (aarch64_bcaxqv8hi): Ditto.
30222 * config/aarch64/aarch64-simd.md:
30223 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
30224 (aarch64_crypto_sha512su0qv2di): Ditto.
30225 (aarch64_crypto_sha512su1qv2di): Ditto.
30226 (aarch64_eor3qv8hi): Ditto.
30227 (aarch64_rax1qv2di): Ditto.
30228 (aarch64_xarqv2di): Ditto.
30229 (aarch64_bcaxqv8hi): Ditto.
30230 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
30231 (vsha512h2q_u64): Ditto.
30232 (vsha512su0q_u64): Ditto.
30233 (vsha512su1q_u64): Ditto.
30234 (veor3q_u16): Ditto.
30235 (vrax1q_u64): Ditto.
30236 (vxarq_u64): Ditto.
30237 (vbcaxq_u16): Ditto.
30238 * config/arm/types.md (crypto_sha512): New type attribute.
30239 (crypto_sha3): Ditto.
30240 (doc/invoke.texi): Document new sha3 option.
30242 2018-01-10 Michael Collison <michael.collison@arm.com>
30244 * config/aarch64/aarch64-builtins.c:
30245 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
30246 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30247 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
30248 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
30249 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
30250 (AARCH64_ISA_SM4): New ISA flag.
30251 (TARGET_SM4): New feature flag for sm4.
30252 * config/aarch64/aarch64-simd-builtins.def
30253 (aarch64_sm3ss1qv4si): Ditto.
30254 (aarch64_sm3tt1aq4si): Ditto.
30255 (aarch64_sm3tt1bq4si): Ditto.
30256 (aarch64_sm3tt2aq4si): Ditto.
30257 (aarch64_sm3tt2bq4si): Ditto.
30258 (aarch64_sm3partw1qv4si): Ditto.
30259 (aarch64_sm3partw2qv4si): Ditto.
30260 (aarch64_sm4eqv4si): Ditto.
30261 (aarch64_sm4ekeyqv4si): Ditto.
30262 * config/aarch64/aarch64-simd.md:
30263 (aarch64_sm3ss1qv4si): Ditto.
30264 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
30265 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
30266 (aarch64_sm4eqv4si): Ditto.
30267 (aarch64_sm4ekeyqv4si): Ditto.
30268 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
30269 (sm3part_op): Ditto.
30270 (CRYPTO_SM3TT): Ditto.
30271 (CRYPTO_SM3PART): Ditto.
30272 (UNSPEC_SM3SS1): New unspec.
30273 (UNSPEC_SM3TT1A): Ditto.
30274 (UNSPEC_SM3TT1B): Ditto.
30275 (UNSPEC_SM3TT2A): Ditto.
30276 (UNSPEC_SM3TT2B): Ditto.
30277 (UNSPEC_SM3PARTW1): Ditto.
30278 (UNSPEC_SM3PARTW2): Ditto.
30279 (UNSPEC_SM4E): Ditto.
30280 (UNSPEC_SM4EKEY): Ditto.
30281 * config/aarch64/constraints.md (Ui2): New constraint.
30282 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
30283 * config/arm/types.md (crypto_sm3): New type attribute.
30284 (crypto_sm4): Ditto.
30285 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
30286 (vsm3tt1aq_u32): Ditto.
30287 (vsm3tt1bq_u32): Ditto.
30288 (vsm3tt2aq_u32): Ditto.
30289 (vsm3tt2bq_u32): Ditto.
30290 (vsm3partw1q_u32): Ditto.
30291 (vsm3partw2q_u32): Ditto.
30292 (vsm4eq_u32): Ditto.
30293 (vsm4ekeyq_u32): Ditto.
30294 (doc/invoke.texi): Document new sm4 option.
30296 2018-01-10 Michael Collison <michael.collison@arm.com>
30298 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
30299 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
30300 (AARCH64_FL_FOR_ARCH8_4): New.
30301 (AARCH64_FL_V8_4): New flag.
30302 (doc/invoke.texi): Document new armv8.4-a option.
30304 2018-01-10 Michael Collison <michael.collison@arm.com>
30306 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30307 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
30308 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
30309 * config/aarch64/aarch64-option-extension.def: Add
30310 AARCH64_OPT_EXTENSION of 'sha2'.
30311 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
30312 (crypto): Disable sha2 and aes if crypto disabled.
30313 (crypto): Enable aes and sha2 if enabled.
30314 (simd): Disable sha2 and aes if simd disabled.
30315 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
30317 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
30318 (TARGET_SHA2): New feature flag for sha2.
30319 (TARGET_AES): New feature flag for aes.
30320 * config/aarch64/aarch64-simd.md:
30321 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
30322 conditional on TARGET_AES.
30323 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
30324 (aarch64_crypto_sha1hsi): Make pattern conditional
30326 (aarch64_crypto_sha1hv4si): Ditto.
30327 (aarch64_be_crypto_sha1hv4si): Ditto.
30328 (aarch64_crypto_sha1su1v4si): Ditto.
30329 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
30330 (aarch64_crypto_sha1su0v4si): Ditto.
30331 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
30332 (aarch64_crypto_sha256su0v4si): Ditto.
30333 (aarch64_crypto_sha256su1v4si): Ditto.
30334 (doc/invoke.texi): Document new aes and sha2 options.
30336 2018-01-10 Martin Sebor <msebor@redhat.com>
30338 PR tree-optimization/83781
30339 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
30342 2018-01-11 Martin Sebor <msebor@gmail.com>
30343 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
30345 PR tree-optimization/83501
30346 PR tree-optimization/81703
30348 * tree-ssa-strlen.c (get_string_cst): Rename...
30349 (get_string_len): ...to this. Handle global constants.
30350 (handle_char_store): Adjust.
30352 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
30353 Jim Wilson <jimw@sifive.com>
30355 * config/riscv/riscv-protos.h (riscv_output_return): New.
30356 * config/riscv/riscv.c (struct machine_function): New naked_p field.
30357 (riscv_attribute_table, riscv_output_return),
30358 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
30359 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
30360 (riscv_compute_frame_info): Only compute frame->mask if not a naked
30362 (riscv_expand_prologue): Add early return for naked function.
30363 (riscv_expand_epilogue): Likewise.
30364 (riscv_function_ok_for_sibcall): Return false for naked function.
30365 (riscv_set_current_function): New.
30366 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
30367 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
30368 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
30369 * doc/extend.texi (RISC-V Function Attributes): New.
30371 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
30373 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
30374 check for 128-bit long double before checking TCmode.
30375 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
30376 128-bit long doubles before checking TFmode or TCmode.
30377 (FLOAT128_IBM_P): Likewise.
30379 2018-01-10 Martin Sebor <msebor@redhat.com>
30381 PR tree-optimization/83671
30382 * builtins.c (c_strlen): Unconditionally return zero for the empty
30384 Use -Warray-bounds for warnings.
30385 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
30386 for non-constant array indices with COMPONENT_REF, arrays of
30387 arrays, and pointers to arrays.
30388 (gimple_fold_builtin_strlen): Determine and set length range for
30389 non-constant character arrays.
30391 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
30393 PR middle-end/81897
30394 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
30397 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
30399 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
30401 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
30404 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
30405 VECTOR_MEM_ALTIVEC_OR_VSX_P.
30406 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
30407 indexed_or_indirect_operand predicate.
30408 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
30409 (*vsx_le_perm_load_v8hi): Likewise.
30410 (*vsx_le_perm_load_v16qi): Likewise.
30411 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
30412 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
30413 (*vsx_le_perm_store_v8hi): Likewise.
30414 (*vsx_le_perm_store_v16qi): Likewise.
30415 (eight unnamed splitters): Likewise.
30417 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
30419 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
30420 * config/rs6000/emmintrin.h: Likewise.
30421 * config/rs6000/mmintrin.h: Likewise.
30422 * config/rs6000/xmmintrin.h: Likewise.
30424 2018-01-10 David Malcolm <dmalcolm@redhat.com>
30427 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
30429 * tree.c (tree_nop_conversion): Return true for location wrapper
30431 (maybe_wrap_with_location): New function.
30432 (selftest::check_strip_nops): New function.
30433 (selftest::test_location_wrappers): New function.
30434 (selftest::tree_c_tests): Call it.
30435 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
30436 (maybe_wrap_with_location): New decl.
30437 (EXPR_LOCATION_WRAPPER_P): New macro.
30438 (location_wrapper_p): New inline function.
30439 (tree_strip_any_location_wrapper): New inline function.
30441 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
30444 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
30445 stack_realign_offset for the largest alignment of stack slot
30447 (ix86_find_max_used_stack_alignment): New function.
30448 (ix86_finalize_stack_frame_flags): Use it. Set
30449 max_used_stack_alignment if we don't realign stack.
30450 * config/i386/i386.h (machine_function): Add
30451 max_used_stack_alignment.
30453 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
30455 * config/arm/arm.opt (-mbranch-cost): New option.
30456 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
30459 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
30462 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
30463 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
30465 2018-01-10 Richard Biener <rguenther@suse.de>
30468 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
30469 early out so it also covers the case where we have a non-NULL
30472 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
30474 PR tree-optimization/83753
30475 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
30476 for non-strided grouped accesses if the number of elements is 1.
30478 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
30481 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
30482 * i386.h (TARGET_USE_GATHER): Define.
30483 * x86-tune.def (X86_TUNE_USE_GATHER): New.
30485 2018-01-10 Martin Liska <mliska@suse.cz>
30488 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
30489 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
30491 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
30492 CLEANUP_NO_PARTITIONING is not set.
30494 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
30496 * doc/rtl.texi: Remove documentation of (const ...) wrappers
30497 for vectors, as a partial revert of r254296.
30498 * rtl.h (const_vec_p): Delete.
30499 (const_vec_duplicate_p): Don't test for vector CONSTs.
30500 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
30501 * expmed.c (make_tree): Likewise.
30504 * common.md (E, F): Use CONSTANT_P instead of checking for
30506 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
30507 checking for CONST_VECTOR.
30509 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
30511 PR middle-end/83575
30512 * predict.c (force_edge_cold): Handle in more sane way edges
30513 with no prediction.
30515 2018-01-09 Carl Love <cel@us.ibm.com>
30517 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
30519 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
30520 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
30521 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
30522 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
30523 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
30524 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
30525 * config/rs6000/rs6000-protos.h: Add extern defition for
30526 rs6000_generate_float2_double_code.
30527 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
30529 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
30530 (float2_v2df): Add define_expand.
30532 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
30535 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
30536 op_mode in the force_to_mode call.
30538 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
30540 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
30541 instead of checking each element individually.
30542 (aarch64_evpc_uzp): Likewise.
30543 (aarch64_evpc_zip): Likewise.
30544 (aarch64_evpc_ext): Likewise.
30545 (aarch64_evpc_rev): Likewise.
30546 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
30547 instead of checking each element individually. Return true without
30549 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
30550 whether all selected elements come from the same input, instead of
30551 checking each element individually. Remove calls to gen_rtx_REG,
30552 start_sequence and end_sequence and instead assert that no rtl is
30555 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
30557 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
30558 order of HIGH and CONST checks.
30560 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
30562 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
30563 if the destination isn't an SSA_NAME.
30565 2018-01-09 Richard Biener <rguenther@suse.de>
30567 PR tree-optimization/83668
30568 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
30570 (canonicalize_loop_form): ... here, renamed from ...
30571 (canonicalize_loop_closed_ssa_form): ... this and amended to
30572 swap successor edges for loop exit blocks to make us use
30573 the RPO order we need for initial schedule generation.
30575 2018-01-09 Joseph Myers <joseph@codesourcery.com>
30577 PR tree-optimization/64811
30578 * match.pd: When optimizing comparisons with Inf, avoid
30579 introducing or losing exceptions from comparisons with NaN.
30581 2018-01-09 Martin Liska <mliska@suse.cz>
30584 * asan.c (shadow_mem_size): Add gcc_assert.
30586 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
30588 Don't save registers in main().
30591 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
30592 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
30593 * config/avr/avr.c (avr_set_current_function): Don't error if
30594 naked, OS_task or OS_main are specified at the same time.
30595 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
30597 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
30599 * common/config/avr/avr-common.c (avr_option_optimization_table):
30600 Switch on -mmain-is-OS_task for optimizing compilations.
30602 2018-01-09 Richard Biener <rguenther@suse.de>
30604 PR tree-optimization/83572
30605 * graphite.c: Include cfganal.h.
30606 (graphite_transform_loops): Connect infinite loops to exit
30607 and remove fake edges at the end.
30609 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
30611 * ipa-inline.c (edge_badness): Revert accidental checkin.
30613 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
30616 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
30617 symbols; not inline clones.
30619 2018-01-09 Jakub Jelinek <jakub@redhat.com>
30622 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
30623 hard registers. Formatting fixes.
30625 PR preprocessor/83722
30626 * gcc.c (try_generate_repro): Pass
30627 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
30628 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
30631 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
30632 Kito Cheng <kito.cheng@gmail.com>
30634 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
30635 (riscv_leaf_function_p): Delete.
30636 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
30638 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
30640 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
30642 (do_ifelse): New function.
30643 (do_isel): New function.
30644 (do_sub3): New function.
30645 (do_add3): New function.
30646 (do_load_mask_compare): New function.
30647 (do_overlap_load_compare): New function.
30648 (expand_compare_loop): New function.
30649 (expand_block_compare): Call expand_compare_loop() when appropriate.
30650 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
30651 option description.
30652 (-mblock-compare-inline-loop-limit): New option.
30654 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
30657 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
30658 Reverse order of second and third operands in first alternative.
30659 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
30660 of first and second elements in UNSPEC_VPERMR vector.
30661 (altivec_expand_vec_perm_le): Likewise.
30663 2018-01-08 Jeff Law <law@redhat.com>
30665 PR rtl-optimizatin/81308
30666 * tree-switch-conversion.c (cfg_altered): New file scoped static.
30667 (process_switch): If group_case_labels makes a change, then set
30669 (pass_convert_switch::execute): If a switch is converted, then
30670 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
30672 PR rtl-optimization/81308
30673 * recog.c (split_all_insns): Conditionally cleanup the CFG after
30676 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
30678 PR target/83663 - Revert r255946
30679 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
30680 generation for cases where splatting a value is not useful.
30681 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
30682 across a vec_duplicate and a paradoxical subreg forming a vector
30683 mode to a vec_concat.
30685 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
30687 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
30688 -march=armv8.3-a variants.
30689 * config/arm/t-multilib: Likewise.
30690 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
30692 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
30694 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
30696 (cceq_ior_compare_complement): Give it a name so I can use it, and
30697 change boolean_or_operator predicate to boolean_operator so it can
30698 be used to generate a crand.
30699 (eqne): New code iterator.
30700 (bd/bd_neg): New code_attrs.
30701 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
30702 a single define_insn.
30703 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
30704 decrement (bdnzt/bdnzf/bdzt/bdzf).
30705 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
30706 with the new names of the branch decrement patterns, and added the
30707 names of the branch decrement conditional patterns.
30709 2018-01-08 Richard Biener <rguenther@suse.de>
30711 PR tree-optimization/83563
30712 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
30715 2018-01-08 Richard Biener <rguenther@suse.de>
30717 PR middle-end/83713
30718 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
30720 2018-01-08 Richard Biener <rguenther@suse.de>
30722 PR tree-optimization/83685
30723 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
30724 references to abnormals.
30726 2018-01-08 Richard Biener <rguenther@suse.de>
30729 * dwarf2out.c (output_indirect_strings): Handle empty
30730 skeleton_debug_str_hash.
30731 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
30733 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
30735 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
30736 (emit_store_direct): Likewise.
30737 (arc_trampoline_adjust_address): Likewise.
30738 (arc_asm_trampoline_template): New function.
30739 (arc_initialize_trampoline): Use asm_trampoline_template.
30740 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
30741 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
30742 * config/arc/arc.md (flush_icache): Delete pattern.
30744 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
30746 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
30747 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
30750 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
30753 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
30754 by not USED_FOR_TARGET.
30755 (make_pass_resolve_sw_modes): Likewise.
30757 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
30759 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
30762 2018-01-08 Richard Biener <rguenther@suse.de>
30764 PR middle-end/83580
30765 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
30767 2018-01-08 Richard Biener <rguenther@suse.de>
30769 PR middle-end/83517
30770 * match.pd ((t * 2) / 2) -> t): Add missing :c.
30772 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
30774 PR middle-end/81897
30775 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
30776 basic blocks with a small number of successors.
30777 (convert_control_dep_chain_into_preds): Improve handling of
30779 (dump_predicates): Split apart into...
30780 (dump_pred_chain): ...here...
30781 (dump_pred_info): ...and here.
30782 (can_one_predicate_be_invalidated_p): Add debugging printfs.
30783 (can_chain_union_be_invalidated_p): Improve check for invalidation
30785 (uninit_uses_cannot_happen): Avoid unnecessary if
30786 convert_control_dep_chain_into_preds yielded nothing.
30788 2018-01-06 Martin Sebor <msebor@redhat.com>
30790 PR tree-optimization/83640
30791 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
30792 subtracting negative offset from size.
30793 (builtin_access::overlap): Adjust offset bounds of the access to fall
30794 within the size of the object if possible.
30796 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
30798 PR rtl-optimization/83699
30799 * expmed.c (extract_bit_field_1): Restrict the vector usage of
30800 extract_bit_field_as_subreg to cases in which the extracted
30801 value is also a vector.
30803 * lra-constraints.c (process_alt_operands): Test for the equivalence
30804 substitutions when detecting a possible reload cycle.
30806 2018-01-06 Jakub Jelinek <jakub@redhat.com>
30809 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
30810 by default if flag_selective_schedling{,2}. Formatting fixes.
30812 PR rtl-optimization/83682
30813 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
30814 if it has non-VECTOR_MODE element mode.
30815 (vec_duplicate_p): Likewise.
30817 PR middle-end/83694
30818 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
30819 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
30821 2018-01-05 Jakub Jelinek <jakub@redhat.com>
30824 * config/i386/i386-builtin.def
30825 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
30826 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
30827 Require also OPTION_MASK_ISA_AVX512F in addition to
30828 OPTION_MASK_ISA_GFNI.
30829 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
30830 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
30831 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
30832 to OPTION_MASK_ISA_GFNI.
30833 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
30834 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
30835 OPTION_MASK_ISA_AVX512BW.
30836 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
30837 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
30838 addition to OPTION_MASK_ISA_GFNI.
30839 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
30840 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
30841 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
30842 to OPTION_MASK_ISA_GFNI.
30843 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
30844 a requirement for all ISAs rather than any of them with a few
30846 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
30848 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
30849 bitmasks to be enabled with 3 exceptions, instead of requiring any
30850 enabled ISA with lots of exceptions.
30851 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
30852 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
30853 Change avx512bw in isa attribute to avx512f.
30854 * config/i386/sgxintrin.h: Add license boilerplate.
30855 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
30856 to __AVX512F__ and __AVX512VL to __AVX512VL__.
30857 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
30858 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
30860 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
30861 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
30862 temporarily sse2 rather than sse if not enabled already.
30865 * config/i386/sse.md (VI248_VLBW): Rename to ...
30866 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
30867 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
30868 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
30869 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
30870 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
30871 mode iterator instead of VI248_VLBW.
30873 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
30875 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
30876 (record_modified): Skip clobbers; add debug output.
30877 (param_change_prob): Use sreal frequencies.
30879 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
30881 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
30882 punt for user-aligned variables.
30884 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
30886 * tree-chrec.c (chrec_contains_symbols): Return true for
30889 2018-01-05 Sudakshina Das <sudi.das@arm.com>
30892 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
30893 of (x|y) == x for BICS pattern.
30895 2018-01-05 Jakub Jelinek <jakub@redhat.com>
30897 PR tree-optimization/83605
30898 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
30899 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
30902 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
30904 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
30905 * config/epiphany/rtems.h: New file.
30907 2018-01-04 Jakub Jelinek <jakub@redhat.com>
30908 Uros Bizjak <ubizjak@gmail.com>
30911 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
30912 QIreg_operand instead of register_operand predicate.
30913 * config/i386/i386.c (ix86_rop_should_change_byte_p,
30914 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
30915 comments instead of -fmitigate[-_]rop.
30917 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
30920 * cgraphunit.c (symbol_table::compile): Switch to text_section
30921 before calling assembly_start debug hook.
30922 * run-rtl-passes.c (run_rtl_passes): Likewise.
30925 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
30927 * tree-vrp.c (extract_range_from_binary_expr_1): Check
30928 range_int_cst_p rather than !symbolic_range_p before calling
30929 extract_range_from_multiplicative_op_1.
30931 2018-01-04 Jeff Law <law@redhat.com>
30933 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
30934 redundant test in assertion.
30936 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
30938 * doc/rtl.texi: Document machine_mode wrapper classes.
30940 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
30942 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
30943 using tree_to_uhwi.
30945 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
30947 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
30948 the VEC_PERM_EXPR fold to fail.
30950 2018-01-04 Jakub Jelinek <jakub@redhat.com>
30953 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
30954 to switched_sections.
30956 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
30959 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
30960 test for d.testing.
30962 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
30965 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
30966 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
30968 2018-01-04 Jakub Jelinek <jakub@redhat.com>
30971 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
30972 is BLKmode and bitpos not zero or mode change is needed.
30974 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
30977 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
30980 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
30983 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
30984 instead of MULT rtx. Update all corresponding splitters.
30985 (*saddl_se): Ditto.
30986 (*ssub<modesuffix>): Ditto.
30987 (*ssubl_se): Ditto.
30988 (*cmp_sadd_di): Update split patterns.
30989 (*cmp_sadd_si): Ditto.
30990 (*cmp_sadd_sidi): Ditto.
30991 (*cmp_ssub_di): Ditto.
30992 (*cmp_ssub_si): Ditto.
30993 (*cmp_ssub_sidi): Ditto.
30994 * config/alpha/predicates.md (const23_operand): New predicate.
30995 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
30996 Look for ASHIFT, not MULT inner operand.
30997 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
30999 2018-01-04 Martin Liska <mliska@suse.cz>
31001 PR gcov-profile/83669
31002 * gcov.c (output_intermediate_file): Add version to intermediate
31004 * doc/gcov.texi: Document new field 'version' in intermediate
31005 file format. Fix location of '-k' option of gcov command.
31007 2018-01-04 Martin Liska <mliska@suse.cz>
31010 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
31012 2018-01-04 Jakub Jelinek <jakub@redhat.com>
31014 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
31016 2018-01-03 Martin Sebor <msebor@redhat.com>
31018 PR tree-optimization/83655
31019 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
31020 checking calls with invalid arguments.
31022 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31024 * tree-vect-stmts.c (vect_get_store_rhs): New function.
31025 (vectorizable_mask_load_store): Delete.
31026 (vectorizable_call): Return false for masked loads and stores.
31027 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
31028 instead of gimple_assign_rhs1.
31029 (vectorizable_load): Handle IFN_MASK_LOAD.
31030 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
31032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31034 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
31036 (vectorizable_mask_load_store): ...here.
31037 (vectorizable_load): ...and here.
31039 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31041 * tree-vect-stmts.c (vect_build_all_ones_mask)
31042 (vect_build_zero_merge_argument): New functions, split out from...
31043 (vectorizable_load): ...here.
31045 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31047 * tree-vect-stmts.c (vect_check_store_rhs): New function,
31049 (vectorizable_mask_load_store): ...here.
31050 (vectorizable_store): ...and here.
31052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31054 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
31056 (vectorizable_mask_load_store): ...here.
31058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31060 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
31061 (vect_model_store_cost): Take a vec_load_store_type instead of a
31063 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
31064 (vect_model_store_cost): Take a vec_load_store_type instead of a
31066 (vectorizable_mask_load_store): Update accordingly.
31067 (vectorizable_store): Likewise.
31068 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
31070 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31072 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
31073 IFN_MASK_LOAD calls here rather than...
31074 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
31076 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31077 Alan Hayward <alan.hayward@arm.com>
31078 David Sherwood <david.sherwood@arm.com>
31080 * expmed.c (extract_bit_field_1): For vector extracts,
31081 fall back to extract_bit_field_as_subreg if vec_extract
31084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31085 Alan Hayward <alan.hayward@arm.com>
31086 David Sherwood <david.sherwood@arm.com>
31088 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
31089 they are variable or constant sized.
31090 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
31091 slots for constant-sized data.
31093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31094 Alan Hayward <alan.hayward@arm.com>
31095 David Sherwood <david.sherwood@arm.com>
31097 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
31098 handling COND_EXPRs with boolean comparisons, try to find a better
31099 basis for the mask type than the boolean itself.
31101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31103 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
31104 is calculated and how it can be overridden.
31105 * genmodes.c (max_bitsize_mode_any_mode): New variable.
31106 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
31108 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
31111 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31112 Alan Hayward <alan.hayward@arm.com>
31113 David Sherwood <david.sherwood@arm.com>
31115 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
31116 Remove the mode argument.
31117 (aarch64_simd_valid_immediate): Remove the mode and inverse
31119 * config/aarch64/iterators.md (bitsize): New iterator.
31120 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
31121 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
31122 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
31123 aarch64_simd_valid_immediate.
31124 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
31125 (aarch64_reg_or_bic_imm): Likewise.
31126 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
31127 with an insn_type enum and msl with a modifier_type enum.
31128 Replace element_width with a scalar_mode. Change the shift
31129 to unsigned int. Add constructors for scalar_float_mode and
31130 scalar_int_mode elements.
31131 (aarch64_vect_float_const_representable_p): Delete.
31132 (aarch64_can_const_movi_rtx_p)
31133 (aarch64_simd_scalar_immediate_valid_for_move)
31134 (aarch64_simd_make_constant): Update call to
31135 aarch64_simd_valid_immediate.
31136 (aarch64_advsimd_valid_immediate_hs): New function.
31137 (aarch64_advsimd_valid_immediate): Likewise.
31138 (aarch64_simd_valid_immediate): Remove mode and inverse
31139 arguments. Rewrite to use the above. Use const_vec_duplicate_p
31140 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
31141 and aarch64_float_const_representable_p on the result.
31142 (aarch64_output_simd_mov_immediate): Remove mode argument.
31143 Update call to aarch64_simd_valid_immediate and use of
31144 simd_immediate_info.
31145 (aarch64_output_scalar_simd_mov_immediate): Update call
31148 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31149 Alan Hayward <alan.hayward@arm.com>
31150 David Sherwood <david.sherwood@arm.com>
31152 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
31153 (mode_nunits): Likewise CONST_MODE_NUNITS.
31154 * machmode.def (ADJUST_NUNITS): Document.
31155 * genmodes.c (mode_data::need_nunits_adj): New field.
31156 (blank_mode): Update accordingly.
31157 (adj_nunits): New variable.
31158 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
31160 (emit_mode_size_inline): Set need_bytesize_adj for all modes
31161 listed in adj_nunits.
31162 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
31163 listed in adj_nunits. Don't emit case statements for such modes.
31164 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
31165 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
31166 nothing if adj_nunits is nonnull.
31167 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
31168 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
31169 (emit_mode_fbit): Update use of print_maybe_const_decl.
31170 (emit_move_size): Likewise. Treat the array as non-const
31172 (emit_mode_adjustments): Handle adj_nunits.
31174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31176 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
31177 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
31178 (VECTOR_MODES): Use it.
31179 (make_vector_modes): Take the prefix as an argument.
31181 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31182 Alan Hayward <alan.hayward@arm.com>
31183 David Sherwood <david.sherwood@arm.com>
31185 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
31186 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
31187 for MODE_VECTOR_BOOL.
31188 * machmode.def (VECTOR_BOOL_MODE): Document.
31189 * genmodes.c (VECTOR_BOOL_MODE): New macro.
31190 (make_vector_bool_mode): New function.
31191 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
31193 * lto-streamer-in.c (lto_input_mode_table): Likewise.
31194 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
31196 * stor-layout.c (int_mode_for_mode): Likewise.
31197 * tree.c (build_vector_type_for_mode): Likewise.
31198 * varasm.c (output_constant_pool_2): Likewise.
31199 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
31200 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
31201 for MODE_VECTOR_BOOL.
31202 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
31203 of mode class checks.
31204 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
31205 instead of a list of mode class checks.
31206 (expand_vector_scalar_condition): Likewise.
31207 (type_for_widest_vector_mode): Handle BImode as an inner mode.
31209 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31210 Alan Hayward <alan.hayward@arm.com>
31211 David Sherwood <david.sherwood@arm.com>
31213 * machmode.h (mode_size): Change from unsigned short to
31215 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
31216 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
31217 or if measurement_type is not polynomial.
31218 (fixed_size_mode::includes_p): Check for constant-sized modes.
31219 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
31220 return a poly_uint16 rather than an unsigned short.
31221 (emit_mode_size): Change the type of mode_size from unsigned short
31222 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
31223 (emit_mode_adjustments): Cope with polynomial vector sizes.
31224 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31226 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31228 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
31229 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
31230 * caller-save.c (setup_save_areas): Likewise.
31231 (replace_reg_with_saved_mem): Likewise.
31232 * calls.c (emit_library_call_value_1): Likewise.
31233 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
31234 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
31235 (gen_lowpart_for_combine): Likewise.
31236 * convert.c (convert_to_integer_1): Likewise.
31237 * cse.c (equiv_constant, cse_insn): Likewise.
31238 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
31239 (cselib_subst_to_values): Likewise.
31240 * dce.c (word_dce_process_block): Likewise.
31241 * df-problems.c (df_word_lr_mark_ref): Likewise.
31242 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
31243 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
31244 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
31245 (rtl_for_decl_location): Likewise.
31246 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
31247 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
31248 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
31249 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
31250 (expand_expr_real_1): Likewise.
31251 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
31252 (pad_below): Likewise.
31253 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
31254 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
31255 * ira.c (get_subreg_tracking_sizes): Likewise.
31256 * ira-build.c (ira_create_allocno_objects): Likewise.
31257 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
31258 (ira_sort_regnos_for_alter_reg): Likewise.
31259 * ira-costs.c (record_operand_costs): Likewise.
31260 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
31261 (resolve_simple_move): Likewise.
31262 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
31263 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
31264 (lra_constraints): Likewise.
31265 (CONST_POOL_OK_P): Reject variable-sized modes.
31266 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
31267 (add_pseudo_to_slot, lra_spill): Likewise.
31268 * omp-low.c (omp_clause_aligned_alignment): Likewise.
31269 * optabs-query.c (get_best_extraction_insn): Likewise.
31270 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
31271 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
31272 (expand_mult_highpart, valid_multiword_target_p): Likewise.
31273 * recog.c (offsettable_address_addr_space_p): Likewise.
31274 * regcprop.c (maybe_mode_change): Likewise.
31275 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
31276 * regrename.c (build_def_use): Likewise.
31277 * regstat.c (dump_reg_info): Likewise.
31278 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
31279 (find_reloads, find_reloads_subreg_address): Likewise.
31280 * reload1.c (eliminate_regs_1): Likewise.
31281 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
31282 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
31283 (simplify_binary_operation_1, simplify_subreg): Likewise.
31284 * targhooks.c (default_function_arg_padding): Likewise.
31285 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
31286 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
31287 (verify_gimple_assign_ternary): Likewise.
31288 * tree-inline.c (estimate_move_cost): Likewise.
31289 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31290 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
31291 (get_address_cost_ainc): Likewise.
31292 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
31293 (vect_supportable_dr_alignment): Likewise.
31294 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
31295 (vectorizable_reduction): Likewise.
31296 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
31297 (vectorizable_operation, vectorizable_load): Likewise.
31298 * tree.c (build_same_sized_truth_vector_type): Likewise.
31299 * valtrack.c (cleanup_auto_inc_dec): Likewise.
31300 * var-tracking.c (emit_note_insn_var_location): Likewise.
31301 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
31302 (ADDR_VEC_ALIGN): Likewise.
31304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31305 Alan Hayward <alan.hayward@arm.com>
31306 David Sherwood <david.sherwood@arm.com>
31308 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
31310 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
31311 or if measurement_type is polynomial.
31312 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
31313 * combine.c (make_extraction): Likewise.
31314 * dse.c (find_shift_sequence): Likewise.
31315 * dwarf2out.c (mem_loc_descriptor): Likewise.
31316 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
31317 (extract_bit_field, extract_low_bits): Likewise.
31318 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
31319 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
31320 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
31321 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
31322 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
31323 * reload.c (find_reloads): Likewise.
31324 * reload1.c (alter_reg): Likewise.
31325 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
31326 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
31327 * tree-if-conv.c (predicate_mem_writes): Likewise.
31328 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
31329 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
31330 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
31331 * valtrack.c (dead_debug_insert_temp): Likewise.
31332 * varasm.c (mergeable_constant_section): Likewise.
31333 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
31335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31336 Alan Hayward <alan.hayward@arm.com>
31337 David Sherwood <david.sherwood@arm.com>
31339 * expr.c (expand_assignment): Cope with polynomial mode sizes
31340 when assigning to a CONCAT.
31342 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31343 Alan Hayward <alan.hayward@arm.com>
31344 David Sherwood <david.sherwood@arm.com>
31346 * machmode.h (mode_precision): Change from unsigned short to
31348 (mode_to_precision): Return a poly_uint16 rather than an unsigned
31350 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
31351 or if measurement_type is not polynomial.
31352 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
31353 in which the mode is already known to be a scalar_int_mode.
31354 * genmodes.c (emit_mode_precision): Change the type of mode_precision
31355 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
31357 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31358 for GET_MODE_PRECISION.
31359 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31360 for GET_MODE_PRECISION.
31361 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
31363 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
31364 (expand_field_assignment, make_extraction): Likewise.
31365 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
31366 (get_last_value): Likewise.
31367 * convert.c (convert_to_integer_1): Likewise.
31368 * cse.c (cse_insn): Likewise.
31369 * expr.c (expand_expr_real_1): Likewise.
31370 * lra-constraints.c (simplify_operand_subreg): Likewise.
31371 * optabs-query.c (can_atomic_load_p): Likewise.
31372 * optabs.c (expand_atomic_load): Likewise.
31373 (expand_atomic_store): Likewise.
31374 * ree.c (combine_reaching_defs): Likewise.
31375 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
31376 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
31377 * tree.h (type_has_mode_precision_p): Likewise.
31378 * ubsan.c (instrument_si_overflow): Likewise.
31380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31381 Alan Hayward <alan.hayward@arm.com>
31382 David Sherwood <david.sherwood@arm.com>
31384 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
31385 polynomial numbers of units.
31386 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
31387 (valid_vector_subparts_p): New function.
31388 (build_vector_type): Remove temporary shim and take the number
31389 of units as a poly_uint64 rather than an int.
31390 (build_opaque_vector_type): Take the number of units as a
31391 poly_uint64 rather than an int.
31392 * tree.c (build_vector_from_ctor): Handle polynomial
31393 TYPE_VECTOR_SUBPARTS.
31394 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
31395 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
31396 (build_vector_from_val): If the number of units is variable,
31397 use build_vec_duplicate_cst for constant operands and
31398 VEC_DUPLICATE_EXPR otherwise.
31399 (make_vector_type): Remove temporary is_constant ().
31400 (build_vector_type, build_opaque_vector_type): Take the number of
31401 units as a poly_uint64 rather than an int.
31402 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
31404 * cfgexpand.c (expand_debug_expr): Likewise.
31405 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
31406 (store_constructor, expand_expr_real_1): Likewise.
31407 (const_scalar_mask_from_tree): Likewise.
31408 * fold-const-call.c (fold_const_reduction): Likewise.
31409 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
31410 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
31411 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
31412 (fold_relational_const): Likewise.
31413 (native_interpret_vector): Likewise. Change the size from an
31414 int to an unsigned int.
31415 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
31416 TYPE_VECTOR_SUBPARTS.
31417 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
31418 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
31419 duplicating a non-constant operand into a variable-length vector.
31420 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
31421 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
31422 * ipa-icf.c (sem_variable::equals): Likewise.
31423 * match.pd: Likewise.
31424 * omp-simd-clone.c (simd_clone_subparts): Likewise.
31425 * print-tree.c (print_node): Likewise.
31426 * stor-layout.c (layout_type): Likewise.
31427 * targhooks.c (default_builtin_vectorization_cost): Likewise.
31428 * tree-cfg.c (verify_gimple_comparison): Likewise.
31429 (verify_gimple_assign_binary): Likewise.
31430 (verify_gimple_assign_ternary): Likewise.
31431 (verify_gimple_assign_single): Likewise.
31432 * tree-pretty-print.c (dump_generic_node): Likewise.
31433 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31434 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
31435 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
31436 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
31437 (vect_shift_permute_load_chain): Likewise.
31438 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
31439 (expand_vector_condition, optimize_vector_constructor): Likewise.
31440 (lower_vec_perm, get_compute_type): Likewise.
31441 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
31442 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
31443 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
31444 (vect_recog_mask_conversion_pattern): Likewise.
31445 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
31446 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
31447 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31448 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
31449 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
31450 (vectorizable_shift, vectorizable_operation, vectorizable_store)
31451 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
31452 (supportable_widening_operation): Likewise.
31453 (supportable_narrowing_operation): Likewise.
31454 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
31456 * varasm.c (output_constant): Likewise.
31458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31459 Alan Hayward <alan.hayward@arm.com>
31460 David Sherwood <david.sherwood@arm.com>
31462 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
31463 so that both the length == 3 and length != 3 cases set up their
31464 own permute vectors. Add comments explaining why we know the
31465 number of elements is constant.
31466 (vect_permute_load_chain): Likewise.
31468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31469 Alan Hayward <alan.hayward@arm.com>
31470 David Sherwood <david.sherwood@arm.com>
31472 * machmode.h (mode_nunits): Change from unsigned char to
31474 (ONLY_FIXED_SIZE_MODES): New macro.
31475 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
31476 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
31477 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
31479 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
31480 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
31481 or if measurement_type is not polynomial.
31482 * genmodes.c (ZERO_COEFFS): New macro.
31483 (emit_mode_nunits_inline): Make mode_nunits_inline return a
31485 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
31486 Use ZERO_COEFFS when emitting initializers.
31487 * data-streamer.h (bp_pack_poly_value): New function.
31488 (bp_unpack_poly_value): Likewise.
31489 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31490 for GET_MODE_NUNITS.
31491 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31492 for GET_MODE_NUNITS.
31493 * tree.c (make_vector_type): Remove temporary shim and make
31494 the real function take the number of units as a poly_uint64
31495 rather than an int.
31496 (build_vector_type_for_mode): Handle polynomial nunits.
31497 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
31498 * emit-rtl.c (const_vec_series_p_1): Likewise.
31499 (gen_rtx_CONST_VECTOR): Likewise.
31500 * fold-const.c (test_vec_duplicate_folding): Likewise.
31501 * genrecog.c (validate_pattern): Likewise.
31502 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
31503 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
31504 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
31505 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
31506 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
31507 * rtlanal.c (subreg_get_info): Likewise.
31508 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31509 (vect_grouped_load_supported): Likewise.
31510 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
31511 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
31512 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
31513 (simplify_const_unary_operation, simplify_binary_operation_1)
31514 (simplify_const_binary_operation, simplify_ternary_operation)
31515 (test_vector_ops_duplicate, test_vector_ops): Likewise.
31516 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
31517 instead of CONST_VECTOR_NUNITS.
31518 * varasm.c (output_constant_pool_2): Likewise.
31519 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
31520 explicit-encoded elements in the XVEC for variable-length vectors.
31522 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31524 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
31526 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31527 Alan Hayward <alan.hayward@arm.com>
31528 David Sherwood <david.sherwood@arm.com>
31530 * coretypes.h (fixed_size_mode): Declare.
31531 (fixed_size_mode_pod): New typedef.
31532 * builtins.h (target_builtins::x_apply_args_mode)
31533 (target_builtins::x_apply_result_mode): Change type to
31534 fixed_size_mode_pod.
31535 * builtins.c (apply_args_size, apply_result_size, result_vector)
31536 (expand_builtin_apply_args_1, expand_builtin_apply)
31537 (expand_builtin_return): Update accordingly.
31539 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31541 * cse.c (hash_rtx_cb): Hash only the encoded elements.
31542 * cselib.c (cselib_hash_rtx): Likewise.
31543 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
31544 CONST_VECTOR encoding.
31546 2018-01-03 Jakub Jelinek <jakub@redhat.com>
31547 Jeff Law <law@redhat.com>
31550 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
31551 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
31552 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
31553 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
31556 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
31557 explicitly probe *sp in a noreturn function if there were any callee
31558 register saves or frame pointer is needed.
31560 2018-01-03 Jakub Jelinek <jakub@redhat.com>
31563 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
31564 BLKmode for ternary, binary or unary expressions.
31567 * var-tracking.c (delete_vta_debug_insn): New inline function.
31568 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
31569 insns from get_insns () to NULL instead of each bb separately.
31570 Use delete_vta_debug_insn. No longer static.
31571 (vt_debug_insns_local, variable_tracking_main_1): Adjust
31572 delete_vta_debug_insns callers.
31573 * rtl.h (delete_vta_debug_insns): Declare.
31574 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
31575 instead of variable_tracking_main.
31577 2018-01-03 Martin Sebor <msebor@redhat.com>
31579 PR tree-optimization/83603
31580 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
31581 arguments past the endof the argument list in functions declared
31582 without a prototype.
31583 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
31584 Avoid checking when arguments are null.
31586 2018-01-03 Martin Sebor <msebor@redhat.com>
31589 * doc/extend.texi (attribute const): Fix a typo.
31590 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
31591 issuing -Wsuggest-attribute for void functions.
31593 2018-01-03 Martin Sebor <msebor@redhat.com>
31595 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
31596 offset_int::from instead of wide_int::to_shwi.
31597 (maybe_diag_overlap): Remove assertion.
31598 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
31599 * gimple-ssa-sprintf.c (format_directive): Same.
31600 (parse_directive): Same.
31601 (sprintf_dom_walker::compute_format_length): Same.
31602 (try_substitute_return_value): Same.
31604 2018-01-03 Jeff Law <law@redhat.com>
31606 PR middle-end/83654
31607 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
31608 non-constant residual for zero at runtime and avoid probing in
31609 that case. Reorganize code for trailing problem to mirror handling
31612 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
31614 PR tree-optimization/83501
31615 * tree-ssa-strlen.c (get_string_cst): New.
31616 (handle_char_store): Call get_string_cst.
31618 2018-01-03 Martin Liska <mliska@suse.cz>
31620 PR tree-optimization/83593
31621 * tree-ssa-strlen.c: Include tree-cfg.h.
31622 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
31623 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
31624 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
31626 (strlen_dom_walker::before_dom_children): Call
31627 gimple_purge_dead_eh_edges. Dump tranformation with details
31629 (strlen_dom_walker::before_dom_children): Update call by adding
31630 new argument cleanup_eh.
31631 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
31633 2018-01-03 Martin Liska <mliska@suse.cz>
31636 * cif-code.def (VARIADIC_THUNK): New enum value.
31637 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
31640 2018-01-03 Jan Beulich <jbeulich@suse.com>
31642 * sse.md (mov<mode>_internal): Tighten condition for when to use
31643 vmovdqu<ssescalarsize> for TI and OI modes.
31645 2018-01-03 Jakub Jelinek <jakub@redhat.com>
31647 Update copyright years.
31649 2018-01-03 Martin Liska <mliska@suse.cz>
31652 * ipa-visibility.c (function_and_variable_visibility): Skip
31653 functions with noipa attribure.
31655 2018-01-03 Jakub Jelinek <jakub@redhat.com>
31657 * gcc.c (process_command): Update copyright notice dates.
31658 * gcov-dump.c (print_version): Ditto.
31659 * gcov.c (print_version): Ditto.
31660 * gcov-tool.c (print_version): Ditto.
31661 * gengtype.c (create_file): Ditto.
31662 * doc/cpp.texi: Bump @copying's copyright year.
31663 * doc/cppinternals.texi: Ditto.
31664 * doc/gcc.texi: Ditto.
31665 * doc/gccint.texi: Ditto.
31666 * doc/gcov.texi: Ditto.
31667 * doc/install.texi: Ditto.
31668 * doc/invoke.texi: Ditto.
31670 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31672 * vector-builder.h (vector_builder::m_full_nelts): Change from
31673 unsigned int to poly_uint64.
31674 (vector_builder::full_nelts): Update prototype accordingly.
31675 (vector_builder::new_vector): Likewise.
31676 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
31677 (vector_builder::operator ==): Likewise.
31678 (vector_builder::finalize): Likewise.
31679 * int-vector-builder.h (int_vector_builder::int_vector_builder):
31680 Take the number of elements as a poly_uint64 rather than an
31682 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
31683 from unsigned int to poly_uint64.
31684 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
31685 (vec_perm_indices::new_vector): Likewise.
31686 (vec_perm_indices::length): Likewise.
31687 (vec_perm_indices::nelts_per_input): Likewise.
31688 (vec_perm_indices::input_nelts): Likewise.
31689 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
31690 number of elements per input as a poly_uint64 rather than an
31691 unsigned int. Use the original encoding for variable-length
31692 vectors, rather than clamping each individual element.
31693 For the second and subsequent elements in each pattern,
31694 clamp the step and base before clamping their sum.
31695 (vec_perm_indices::series_p): Handle polynomial element counts.
31696 (vec_perm_indices::all_in_range_p): Likewise.
31697 (vec_perm_indices_to_tree): Likewise.
31698 (vec_perm_indices_to_rtx): Likewise.
31699 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
31700 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
31701 (tree_vector_builder::new_binary_operation): Handle polynomial
31702 element counts. Return false if we need to know the number
31703 of elements at compile time.
31704 * fold-const.c (fold_vec_perm): Punt if the number of elements
31705 isn't known at compile time.
31707 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31709 * vec-perm-indices.h (vec_perm_builder): Change element type
31710 from HOST_WIDE_INT to poly_int64.
31711 (vec_perm_indices::element_type): Update accordingly.
31712 (vec_perm_indices::clamp): Handle polynomial element_types.
31713 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
31714 (vec_perm_indices::all_in_range_p): Likewise.
31715 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
31717 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
31718 polynomial vec_perm_indices element types.
31719 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
31720 * fold-const.c (fold_vec_perm): Likewise.
31721 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
31722 * tree-vect-generic.c (lower_vec_perm): Likewise.
31723 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31724 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
31725 element type to HOST_WIDE_INT.
31727 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31728 Alan Hayward <alan.hayward@arm.com>
31729 David Sherwood <david.sherwood@arm.com>
31731 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
31732 rather than an int. Use plus_constant.
31733 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
31734 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
31736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31737 Alan Hayward <alan.hayward@arm.com>
31738 David Sherwood <david.sherwood@arm.com>
31740 * calls.c (emit_call_1, expand_call): Change struct_value_size from
31741 a HOST_WIDE_INT to a poly_int64.
31743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31744 Alan Hayward <alan.hayward@arm.com>
31745 David Sherwood <david.sherwood@arm.com>
31747 * calls.c (load_register_parameters): Cope with polynomial
31748 mode sizes. Require a constant size for BLKmode parameters
31749 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
31750 forces a parameter to be padded at the lsb end in order to
31751 fill a complete number of words, require the parameter size
31752 to be ordered wrt UNITS_PER_WORD.
31754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31755 Alan Hayward <alan.hayward@arm.com>
31756 David Sherwood <david.sherwood@arm.com>
31758 * reload1.c (spill_stack_slot_width): Change element type
31759 from unsigned int to poly_uint64_pod.
31760 (alter_reg): Treat mode sizes as polynomial.
31762 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31763 Alan Hayward <alan.hayward@arm.com>
31764 David Sherwood <david.sherwood@arm.com>
31766 * reload.c (complex_word_subreg_p): New function.
31767 (reload_inner_reg_of_subreg, push_reload): Use it.
31769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31770 Alan Hayward <alan.hayward@arm.com>
31771 David Sherwood <david.sherwood@arm.com>
31773 * lra-constraints.c (process_alt_operands): Reject matched
31774 operands whose sizes aren't ordered.
31775 (match_reload): Refer to this check here.
31777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31778 Alan Hayward <alan.hayward@arm.com>
31779 David Sherwood <david.sherwood@arm.com>
31781 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
31782 that the mode size is in the set {1, 2, 4, 8, 16}.
31784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31785 Alan Hayward <alan.hayward@arm.com>
31786 David Sherwood <david.sherwood@arm.com>
31788 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
31789 Use plus_constant instead of gen_rtx_PLUS.
31791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31792 Alan Hayward <alan.hayward@arm.com>
31793 David Sherwood <david.sherwood@arm.com>
31795 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
31796 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
31797 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
31798 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
31799 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
31800 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
31801 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
31802 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
31803 * config/i386/i386.c (ix86_push_rounding): ...this new function.
31804 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
31806 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
31807 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
31808 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
31809 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
31810 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
31811 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
31812 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
31813 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
31814 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
31815 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
31817 * expr.c (emit_move_resolve_push): Treat the input and result
31818 of PUSH_ROUNDING as a poly_int64.
31819 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
31820 (emit_push_insn): Likewise.
31821 * lra-eliminations.c (mark_not_eliminable): Likewise.
31822 * recog.c (push_operand): Likewise.
31823 * reload1.c (elimination_effects): Likewise.
31824 * rtlanal.c (nonzero_bits1): Likewise.
31825 * calls.c (store_one_arg): Likewise. Require the padding to be
31826 known at compile time.
31828 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31829 Alan Hayward <alan.hayward@arm.com>
31830 David Sherwood <david.sherwood@arm.com>
31832 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
31833 Use plus_constant instead of gen_rtx_PLUS.
31835 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31836 Alan Hayward <alan.hayward@arm.com>
31837 David Sherwood <david.sherwood@arm.com>
31839 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
31840 rather than an int.
31842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31843 Alan Hayward <alan.hayward@arm.com>
31844 David Sherwood <david.sherwood@arm.com>
31846 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
31847 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
31848 via stack temporaries. Treat the mode size as polynomial too.
31850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31851 Alan Hayward <alan.hayward@arm.com>
31852 David Sherwood <david.sherwood@arm.com>
31854 * expr.c (expand_expr_real_2): When handling conversions involving
31855 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
31856 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
31857 as a poly_uint64 too.
31859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31860 Alan Hayward <alan.hayward@arm.com>
31861 David Sherwood <david.sherwood@arm.com>
31863 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
31865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31866 Alan Hayward <alan.hayward@arm.com>
31867 David Sherwood <david.sherwood@arm.com>
31869 * combine.c (can_change_dest_mode): Handle polynomial
31870 REGMODE_NATURAL_SIZE.
31871 * expmed.c (store_bit_field_1): Likewise.
31872 * expr.c (store_constructor): Likewise.
31873 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
31874 and polynomial REGMODE_NATURAL_SIZE.
31875 (gen_lowpart_common): Likewise.
31876 * reginfo.c (record_subregs_of_mode): Likewise.
31877 * rtlanal.c (read_modify_subreg_p): Likewise.
31879 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31880 Alan Hayward <alan.hayward@arm.com>
31881 David Sherwood <david.sherwood@arm.com>
31883 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
31884 numbers of elements.
31886 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31887 Alan Hayward <alan.hayward@arm.com>
31888 David Sherwood <david.sherwood@arm.com>
31890 * match.pd: Cope with polynomial numbers of vector elements.
31892 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31893 Alan Hayward <alan.hayward@arm.com>
31894 David Sherwood <david.sherwood@arm.com>
31896 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
31897 in a POINTER_PLUS_EXPR.
31899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31900 Alan Hayward <alan.hayward@arm.com>
31901 David Sherwood <david.sherwood@arm.com>
31903 * omp-simd-clone.c (simd_clone_subparts): New function.
31904 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
31905 (ipa_simd_modify_function_body): Likewise.
31907 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31908 Alan Hayward <alan.hayward@arm.com>
31909 David Sherwood <david.sherwood@arm.com>
31911 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
31912 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
31913 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
31914 (expand_vector_condition, vector_element): Likewise.
31915 (subparts_gt): New function.
31916 (get_compute_type): Use subparts_gt.
31917 (count_type_subparts): Delete.
31918 (expand_vector_operations_1): Use subparts_gt instead of
31919 count_type_subparts.
31921 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31922 Alan Hayward <alan.hayward@arm.com>
31923 David Sherwood <david.sherwood@arm.com>
31925 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
31926 (vect_compile_time_alias): ...this new function. Do the calculation
31927 on poly_ints rather than trees.
31928 (vect_prune_runtime_alias_test_list): Update call accordingly.
31930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31931 Alan Hayward <alan.hayward@arm.com>
31932 David Sherwood <david.sherwood@arm.com>
31934 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
31936 (vect_schedule_slp_instance): Likewise.
31938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31939 Alan Hayward <alan.hayward@arm.com>
31940 David Sherwood <david.sherwood@arm.com>
31942 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
31943 constant and extern definitions for variable-length vectors.
31944 (vect_get_constant_vectors): Note that the number of units
31945 is known to be constant.
31947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31948 Alan Hayward <alan.hayward@arm.com>
31949 David Sherwood <david.sherwood@arm.com>
31951 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
31952 of units as polynomial. Choose between WIDE and NARROW based
31955 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31956 Alan Hayward <alan.hayward@arm.com>
31957 David Sherwood <david.sherwood@arm.com>
31959 * tree-vect-stmts.c (simd_clone_subparts): New function.
31960 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
31962 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31963 Alan Hayward <alan.hayward@arm.com>
31964 David Sherwood <david.sherwood@arm.com>
31966 * tree-vect-stmts.c (vectorizable_call): Treat the number of
31967 vectors as polynomial. Use build_index_vector for
31968 IFN_GOMP_SIMD_LANE.
31970 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31971 Alan Hayward <alan.hayward@arm.com>
31972 David Sherwood <david.sherwood@arm.com>
31974 * tree-vect-stmts.c (get_load_store_type): Treat the number of
31975 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
31976 for variable-length vectors.
31977 (vectorizable_mask_load_store): Treat the number of units as
31978 polynomial, asserting that it is constant if the condition has
31979 already been enforced.
31980 (vectorizable_store, vectorizable_load): Likewise.
31982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31983 Alan Hayward <alan.hayward@arm.com>
31984 David Sherwood <david.sherwood@arm.com>
31986 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
31987 of units as polynomial. Punt if we can't tell at compile time
31988 which vector contains the final result.
31990 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
31991 Alan Hayward <alan.hayward@arm.com>
31992 David Sherwood <david.sherwood@arm.com>
31994 * tree-vect-loop.c (vectorizable_induction): Treat the number
31995 of units as polynomial. Punt on SLP inductions. Use an integer
31996 VEC_SERIES_EXPR for variable-length integer reductions. Use a
31997 cast of such a series for variable-length floating-point
32000 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32001 Alan Hayward <alan.hayward@arm.com>
32002 David Sherwood <david.sherwood@arm.com>
32004 * tree.h (build_index_vector): Declare.
32005 * tree.c (build_index_vector): New function.
32006 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
32007 of units as polynomial, forcibly converting it to a constant if
32008 vectorizable_reduction has already enforced the condition.
32009 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
32010 to create a {1,2,3,...} vector.
32011 (vectorizable_reduction): Treat the number of units as polynomial.
32012 Choose vectype_in based on the largest scalar element size rather
32013 than the smallest number of units. Enforce the restrictions
32016 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32017 Alan Hayward <alan.hayward@arm.com>
32018 David Sherwood <david.sherwood@arm.com>
32020 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
32021 number of units as polynomial.
32023 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32024 Alan Hayward <alan.hayward@arm.com>
32025 David Sherwood <david.sherwood@arm.com>
32027 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
32028 * target.def (autovectorize_vector_sizes): Return the vector sizes
32029 by pointer, using vector_sizes rather than a bitmask.
32030 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
32031 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
32032 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
32034 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
32035 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
32036 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
32037 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
32038 * omp-general.c (omp_max_vf): Likewise.
32039 * omp-low.c (omp_clause_aligned_alignment): Likewise.
32040 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
32041 * tree-vect-loop.c (vect_analyze_loop): Likewise.
32042 * tree-vect-slp.c (vect_slp_bb): Likewise.
32043 * doc/tm.texi: Regenerate.
32044 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
32046 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
32047 the vector size as a poly_uint64 rather than an unsigned int.
32048 (current_vector_size): Change from an unsigned int to a poly_uint64.
32049 (get_vectype_for_scalar_type): Update accordingly.
32050 * tree.h (build_truth_vector_type): Take the size and number of
32051 units as a poly_uint64 rather than an unsigned int.
32052 (build_vector_type): Add a temporary overload that takes
32053 the number of units as a poly_uint64 rather than an unsigned int.
32054 * tree.c (make_vector_type): Likewise.
32055 (build_truth_vector_type): Take the number of units as a poly_uint64
32056 rather than an unsigned int.
32058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32059 Alan Hayward <alan.hayward@arm.com>
32060 David Sherwood <david.sherwood@arm.com>
32062 * target.def (get_mask_mode): Take the number of units and length
32063 as poly_uint64s rather than unsigned ints.
32064 * targhooks.h (default_get_mask_mode): Update accordingly.
32065 * targhooks.c (default_get_mask_mode): Likewise.
32066 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
32067 * doc/tm.texi: Regenerate.
32069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32070 Alan Hayward <alan.hayward@arm.com>
32071 David Sherwood <david.sherwood@arm.com>
32073 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
32074 * omp-general.c (omp_max_vf): Likewise.
32075 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
32076 (expand_omp_simd): Handle polynomial safelen.
32077 * omp-low.c (omplow_simd_context): Add a default constructor.
32078 (omplow_simd_context::max_vf): Change from int to poly_uint64.
32079 (lower_rec_simd_input_clauses): Update accordingly.
32080 (lower_rec_input_clauses): Likewise.
32082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32083 Alan Hayward <alan.hayward@arm.com>
32084 David Sherwood <david.sherwood@arm.com>
32086 * tree-vectorizer.h (vect_nunits_for_cost): New function.
32087 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
32088 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
32089 (vect_analyze_slp_cost): Likewise.
32090 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
32091 (vect_model_load_cost): Likewise.
32093 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32094 Alan Hayward <alan.hayward@arm.com>
32095 David Sherwood <david.sherwood@arm.com>
32097 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
32098 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
32099 from an unsigned int * to a poly_uint64_pod *.
32100 (calculate_unrolling_factor): New function.
32101 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
32103 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32104 Alan Hayward <alan.hayward@arm.com>
32105 David Sherwood <david.sherwood@arm.com>
32107 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
32108 from an unsigned int to a poly_uint64.
32109 (_loop_vec_info::slp_unrolling_factor): Likewise.
32110 (_loop_vec_info::vectorization_factor): Change from an int
32112 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
32113 (vect_get_num_vectors): New function.
32114 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
32115 (vect_get_num_copies): Use vect_get_num_vectors.
32116 (vect_analyze_data_ref_dependences): Change max_vf from an int *
32117 to an unsigned int *.
32118 (vect_analyze_data_refs): Change min_vf from an int * to a
32120 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
32121 than an unsigned HOST_WIDE_INT.
32122 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
32123 (vect_analyze_data_ref_dependence): Change max_vf from an int *
32124 to an unsigned int *.
32125 (vect_analyze_data_ref_dependences): Likewise.
32126 (vect_compute_data_ref_alignment): Handle polynomial vf.
32127 (vect_enhance_data_refs_alignment): Likewise.
32128 (vect_prune_runtime_alias_test_list): Likewise.
32129 (vect_shift_permute_load_chain): Likewise.
32130 (vect_supportable_dr_alignment): Likewise.
32131 (dependence_distance_ge_vf): Take the vectorization factor as a
32132 poly_uint64 rather than an unsigned HOST_WIDE_INT.
32133 (vect_analyze_data_refs): Change min_vf from an int * to a
32135 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
32136 vfm1 as a poly_uint64 rather than an int. Make the same change
32137 for the returned bound_scalar.
32138 (vect_gen_vector_loop_niters): Handle polynomial vf.
32139 (vect_do_peeling): Likewise. Update call to
32140 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
32141 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
32143 * tree-vect-loop.c (vect_determine_vectorization_factor)
32144 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
32145 (vect_get_known_peeling_cost): Likewise.
32146 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
32147 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
32148 (vect_transform_loop): Likewise. Use the lowest possible VF when
32149 updating the upper bounds of the loop.
32150 (vect_min_worthwhile_factor): Make static. Return an unsigned int
32151 rather than an int.
32152 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
32153 polynomial unroll factors.
32154 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
32155 (vect_make_slp_decision): Likewise.
32156 (vect_supported_load_permutation_p): Likewise, and polynomial
32158 (vect_analyze_slp_cost): Handle polynomial vf.
32159 (vect_slp_analyze_node_operations): Likewise.
32160 (vect_slp_analyze_bb_1): Likewise.
32161 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
32162 than an unsigned HOST_WIDE_INT.
32163 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
32164 (vectorizable_load): Handle polynomial vf.
32165 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
32167 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
32169 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32170 Alan Hayward <alan.hayward@arm.com>
32171 David Sherwood <david.sherwood@arm.com>
32173 * match.pd: Handle bit operations involving three constants
32174 and try to fold one pair.
32176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
32178 * tree-vect-loop-manip.c: Include gimple-fold.h.
32179 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
32180 niters_maybe_zero parameters. Handle other cases besides a step of 1.
32181 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
32182 Add a path that uses a step of VF instead of 1, but disable it
32184 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
32185 and niters_no_overflow parameters. Update calls to
32186 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
32187 Create a new SSA name if the latter choses to use a ste other
32188 than zero, and return it via niters_vector_mult_vf_var.
32189 * tree-vect-loop.c (vect_transform_loop): Update calls to
32190 vect_do_peeling, vect_gen_vector_loop_niters and
32191 slpeel_make_loop_iterate_ntimes.
32192 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
32193 (vect_gen_vector_loop_niters): Update declarations after above changes.
32195 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
32197 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
32198 128-bit round to integer instructions.
32199 (ceil<mode>2): Likewise.
32200 (btrunc<mode>2): Likewise.
32201 (round<mode>2): Likewise.
32203 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
32205 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
32206 unaligned VSX load/store on P8/P9.
32207 (expand_block_clear): Allow the use of unaligned VSX
32208 load/store on P8/P9.
32210 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
32212 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
32214 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
32215 swap associated with both a load and a store.
32217 2018-01-02 Andrew Waterman <andrew@sifive.com>
32219 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
32220 * config/riscv/riscv.md (clear_cache): Use it.
32222 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
32224 * web.c: Remove out-of-date comment.
32226 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32228 * expr.c (fixup_args_size_notes): Check that any existing
32229 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
32230 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
32231 (emit_single_push_insn): ...here.
32233 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32235 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
32236 (const_vector_encoded_nelts): New function.
32237 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
32238 (const_vector_int_elt, const_vector_elt): Declare.
32239 * emit-rtl.c (const_vector_int_elt_1): New function.
32240 (const_vector_elt): Likewise.
32241 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
32242 of CONST_VECTOR_ELT.
32244 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32246 * expr.c: Include rtx-vector-builder.h.
32247 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
32248 directly on the tree encoding.
32249 (const_vector_from_tree): Likewise.
32250 * optabs.c: Include rtx-vector-builder.h.
32251 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
32252 sequence of "u" values.
32253 * vec-perm-indices.c: Include rtx-vector-builder.h.
32254 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
32255 directly on the vec_perm_indices encoding.
32257 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32259 * doc/rtl.texi (const_vector): Describe new encoding scheme.
32260 * Makefile.in (OBJS): Add rtx-vector-builder.o.
32261 * rtx-vector-builder.h: New file.
32262 * rtx-vector-builder.c: Likewise.
32263 * rtl.h (rtx_def::u2): Add a const_vector field.
32264 (CONST_VECTOR_NPATTERNS): New macro.
32265 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
32266 (CONST_VECTOR_DUPLICATE_P): Likewise.
32267 (CONST_VECTOR_STEPPED_P): Likewise.
32268 (CONST_VECTOR_ENCODED_ELT): Likewise.
32269 (const_vec_duplicate_p): Check for a duplicated vector encoding.
32270 (unwrap_const_vec_duplicate): Likewise.
32271 (const_vec_series_p): Check for a non-duplicated vector encoding.
32272 Say that the function only returns true for integer vectors.
32273 * emit-rtl.c: Include rtx-vector-builder.h.
32274 (gen_const_vec_duplicate_1): Delete.
32275 (gen_const_vector): Call gen_const_vec_duplicate instead of
32276 gen_const_vec_duplicate_1.
32277 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
32278 (gen_const_vec_duplicate): Use rtx_vector_builder.
32279 (gen_const_vec_series): Likewise.
32280 (gen_rtx_CONST_VECTOR): Likewise.
32281 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
32282 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
32283 Build a new vector rather than modifying a CONST_VECTOR in-place.
32284 (handle_special_swappables): Update call accordingly.
32285 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
32286 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
32287 Build a new vector rather than modifying a CONST_VECTOR in-place.
32288 (handle_special_swappables): Update call accordingly.
32290 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32292 * simplify-rtx.c (simplify_const_binary_operation): Use
32293 CONST_VECTOR_ELT instead of XVECEXP.
32295 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32297 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
32298 the selector elements to be different from the data elements
32299 if the selector is a VECTOR_CST.
32300 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
32301 ssizetype for the selector.
32303 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32305 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
32306 before testing each element individually.
32307 * tree-vect-generic.c (lower_vec_perm): Likewise.
32309 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32311 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
32312 * selftest-run-tests.c (selftest::run_tests): Call it.
32313 * vector-builder.h (vector_builder::operator ==): New function.
32314 (vector_builder::operator !=): Likewise.
32315 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
32316 (vec_perm_indices::all_from_input_p): New function.
32317 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
32318 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
32319 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
32320 instead of reading the VECTOR_CST directly. Detect whether both
32321 vector inputs are the same before constructing the vec_perm_indices,
32322 and update the number of inputs argument accordingly. Use the
32323 utility functions added above. Only construct sel2 if we need to.
32325 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32327 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
32328 the broadcast of the low byte.
32329 (expand_mult_highpart): Use an explicit encoding for the permutes.
32330 * optabs-query.c (can_mult_highpart_p): Likewise.
32331 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
32332 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32333 (vectorizable_bswap): Likewise.
32334 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
32335 explicit encoding for the power-of-2 permutes.
32336 (vect_permute_store_chain): Likewise.
32337 (vect_grouped_load_supported): Likewise.
32338 (vect_permute_load_chain): Likewise.
32340 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32342 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
32343 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
32344 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
32345 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
32346 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
32347 (vect_gen_perm_mask_any): Likewise.
32349 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32351 * int-vector-builder.h: New file.
32352 * vec-perm-indices.h: Include int-vector-builder.h.
32353 (vec_perm_indices): Redefine as an int_vector_builder.
32354 (auto_vec_perm_indices): Delete.
32355 (vec_perm_builder): Redefine as a stand-alone class.
32356 (vec_perm_indices::vec_perm_indices): New function.
32357 (vec_perm_indices::clamp): Likewise.
32358 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
32359 (vec_perm_indices::new_vector): New function.
32360 (vec_perm_indices::new_expanded_vector): Update for new
32361 vec_perm_indices class.
32362 (vec_perm_indices::rotate_inputs): New function.
32363 (vec_perm_indices::all_in_range_p): Operate directly on the
32364 encoded form, without computing elided elements.
32365 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
32366 encoding. Update for new vec_perm_indices class.
32367 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
32368 the given vec_perm_builder.
32369 (expand_vec_perm_var): Update vec_perm_builder constructor.
32370 (expand_mult_highpart): Use vec_perm_builder instead of
32371 auto_vec_perm_indices.
32372 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
32373 vec_perm_indices instead of auto_vec_perm_indices. Use a single
32374 or double series encoding as appropriate.
32375 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
32376 vec_perm_indices instead of auto_vec_perm_indices.
32377 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32378 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32379 (vect_permute_store_chain): Likewise.
32380 (vect_grouped_load_supported): Likewise.
32381 (vect_permute_load_chain): Likewise.
32382 (vect_shift_permute_load_chain): Likewise.
32383 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32384 (vect_transform_slp_perm_load): Likewise.
32385 (vect_schedule_slp_instance): Likewise.
32386 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32387 (vectorizable_mask_load_store): Likewise.
32388 (vectorizable_bswap): Likewise.
32389 (vectorizable_store): Likewise.
32390 (vectorizable_load): Likewise.
32391 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
32392 vec_perm_indices instead of auto_vec_perm_indices. Use
32393 tree_to_vec_perm_builder to read the vector from a tree.
32394 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
32395 vec_perm_builder instead of a vec_perm_indices.
32396 (have_whole_vector_shift): Use vec_perm_builder and
32397 vec_perm_indices instead of auto_vec_perm_indices. Leave the
32398 truncation to calc_vec_perm_mask_for_shift.
32399 (vect_create_epilog_for_reduction): Likewise.
32400 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
32401 from auto_vec_perm_indices to vec_perm_indices.
32402 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
32403 instead of changing individual elements.
32404 (aarch64_vectorize_vec_perm_const): Use new_vector to install
32405 the vector in d.perm.
32406 * config/arm/arm.c (expand_vec_perm_d::perm): Change
32407 from auto_vec_perm_indices to vec_perm_indices.
32408 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
32409 instead of changing individual elements.
32410 (arm_vectorize_vec_perm_const): Use new_vector to install
32411 the vector in d.perm.
32412 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
32413 Update vec_perm_builder constructor.
32414 (rs6000_expand_interleave): Likewise.
32415 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
32416 (rs6000_expand_interleave): Likewise.
32418 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32420 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
32421 to qimode could truncate the indices.
32422 * optabs.c (expand_vec_perm_var): Likewise.
32424 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32426 * Makefile.in (OBJS): Add vec-perm-indices.o.
32427 * vec-perm-indices.h: New file.
32428 * vec-perm-indices.c: Likewise.
32429 * target.h (vec_perm_indices): Replace with a forward class
32431 (auto_vec_perm_indices): Move to vec-perm-indices.h.
32432 * optabs.h: Include vec-perm-indices.h.
32433 (expand_vec_perm): Delete.
32434 (selector_fits_mode_p, expand_vec_perm_var): Declare.
32435 (expand_vec_perm_const): Declare.
32436 * target.def (vec_perm_const_ok): Replace with...
32437 (vec_perm_const): ...this new hook.
32438 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
32439 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
32440 * doc/tm.texi: Regenerate.
32441 * optabs.def (vec_perm_const): Delete.
32442 * doc/md.texi (vec_perm_const): Likewise.
32443 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
32444 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
32445 expand_vec_perm for constant permutation vectors. Assert that
32446 the mode of variable permutation vectors is the integer equivalent
32447 of the mode that is being permuted.
32448 * optabs-query.h (selector_fits_mode_p): Declare.
32449 * optabs-query.c: Include vec-perm-indices.h.
32450 (selector_fits_mode_p): New function.
32451 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
32452 is defined, instead of checking whether the vec_perm_const_optab
32453 exists. Use targetm.vectorize.vec_perm_const instead of
32454 targetm.vectorize.vec_perm_const_ok. Check whether the indices
32455 fit in the vector mode before using a variable permute.
32456 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
32457 vec_perm_indices instead of an rtx.
32458 (expand_vec_perm): Replace with...
32459 (expand_vec_perm_const): ...this new function. Take the selector
32460 as a vec_perm_indices rather than an rtx. Also take the mode of
32461 the selector. Update call to shift_amt_for_vec_perm_mask.
32462 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
32463 Use vec_perm_indices::new_expanded_vector to expand the original
32464 selector into bytes. Check whether the indices fit in the vector
32465 mode before using a variable permute.
32466 (expand_vec_perm_var): Make global.
32467 (expand_mult_highpart): Use expand_vec_perm_const.
32468 * fold-const.c: Includes vec-perm-indices.h.
32469 * tree-ssa-forwprop.c: Likewise.
32470 * tree-vect-data-refs.c: Likewise.
32471 * tree-vect-generic.c: Likewise.
32472 * tree-vect-loop.c: Likewise.
32473 * tree-vect-slp.c: Likewise.
32474 * tree-vect-stmts.c: Likewise.
32475 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
32477 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
32478 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
32479 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
32480 (aarch64_vectorize_vec_perm_const): ...this new function.
32481 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32482 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32483 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
32484 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
32485 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32486 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32487 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
32489 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
32490 check for NEON modes.
32491 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
32492 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
32493 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
32494 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
32496 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
32497 the old VEC_PERM_CONST conditions.
32498 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
32499 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
32500 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
32501 (ia64_vectorize_vec_perm_const_ok): Merge into...
32502 (ia64_vectorize_vec_perm_const): ...this new function.
32503 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
32504 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
32505 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
32506 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
32507 * config/mips/mips.c (mips_expand_vec_perm_const)
32508 (mips_vectorize_vec_perm_const_ok): Merge into...
32509 (mips_vectorize_vec_perm_const): ...this new function.
32510 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
32511 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
32512 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
32513 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
32514 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
32515 (rs6000_expand_vec_perm_const): Delete.
32516 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
32518 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32519 (altivec_expand_vec_perm_const_le): Take each operand individually.
32520 Operate on constant selectors rather than rtxes.
32521 (altivec_expand_vec_perm_const): Likewise. Update call to
32522 altivec_expand_vec_perm_const_le.
32523 (rs6000_expand_vec_perm_const): Delete.
32524 (rs6000_vectorize_vec_perm_const_ok): Delete.
32525 (rs6000_vectorize_vec_perm_const): New function.
32526 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32527 an element count and rtx array.
32528 (rs6000_expand_extract_even): Update call accordingly.
32529 (rs6000_expand_interleave): Likewise.
32530 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
32531 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
32532 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
32533 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
32534 (rs6000_expand_vec_perm_const): Delete.
32535 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32536 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32537 (altivec_expand_vec_perm_const_le): Take each operand individually.
32538 Operate on constant selectors rather than rtxes.
32539 (altivec_expand_vec_perm_const): Likewise. Update call to
32540 altivec_expand_vec_perm_const_le.
32541 (rs6000_expand_vec_perm_const): Delete.
32542 (rs6000_vectorize_vec_perm_const_ok): Delete.
32543 (rs6000_vectorize_vec_perm_const): New function. Remove stray
32544 reference to the SPE evmerge intructions.
32545 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32546 an element count and rtx array.
32547 (rs6000_expand_extract_even): Update call accordingly.
32548 (rs6000_expand_interleave): Likewise.
32549 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
32550 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
32552 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32554 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32556 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
32557 vector mode and that that mode matches the mode of the data
32559 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
32560 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
32561 directly using expand_vec_perm_1 when forcing selectors into
32563 (expand_vec_perm_var): New function, split out from expand_vec_perm.
32565 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32567 * optabs-query.h (can_vec_perm_p): Delete.
32568 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
32569 * optabs-query.c (can_vec_perm_p): Split into...
32570 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
32571 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
32572 particular selector is valid.
32573 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32574 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32575 (vect_grouped_load_supported): Likewise.
32576 (vect_shift_permute_load_chain): Likewise.
32577 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32578 (vect_transform_slp_perm_load): Likewise.
32579 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32580 (vectorizable_bswap): Likewise.
32581 (vect_gen_perm_mask_checked): Likewise.
32582 * fold-const.c (fold_ternary_loc): Likewise. Don't take
32583 implementations of variable permutation vectors into account
32584 when deciding which selector to use.
32585 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
32586 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
32587 with a false third argument.
32588 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
32589 to test whether the constant selector is valid and can_vec_perm_var_p
32590 to test whether a variable selector is valid.
32592 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32594 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
32595 * optabs-query.c (can_vec_perm_p): Likewise.
32596 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
32597 instead of vec_perm_indices.
32598 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
32599 (vect_gen_perm_mask_checked): Likewise,
32600 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
32601 (vect_gen_perm_mask_checked): Likewise,
32603 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
32605 * optabs-query.h (qimode_for_vec_perm): Declare.
32606 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
32607 (qimode_for_vec_perm): ...this new function.
32608 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
32610 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
32612 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
32613 does not have a conditional at the top.
32615 2018-01-02 Richard Biener <rguenther@suse.de>
32617 * ipa-inline.c (big_speedup_p): Fix expression.
32619 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
32622 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
32625 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
32629 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
32630 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
32631 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
32632 cond_taken_branch_cost 3->4.
32634 2018-01-01 Jakub Jelinek <jakub@redhat.com>
32636 PR tree-optimization/83581
32637 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
32638 TODO_cleanup_cfg if any changes have been made.
32640 PR middle-end/83608
32641 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
32642 convert_modes if target mode has the right side, but different mode
32645 PR middle-end/83609
32646 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
32647 last argument when extracting from CONCAT. If either from_real or
32648 from_imag is NULL, use expansion through memory. If result is not
32649 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
32650 the parts directly to inner mode, if even that fails, use expansion
32653 PR middle-end/83623
32654 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
32655 check for bswap in mode rather than HImode and use that in expand_unop
32658 Copyright (C) 2018 Free Software Foundation, Inc.
32660 Copying and distribution of this file, with or without modification,
32661 are permitted in any medium without royalty provided the copyright
32662 notice and this notice are preserved.